>temp_0|source:-*- coding: utf-8 -*-|start:5624|stop:6124
 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s /
 mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.mean
(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_lengths),\max(sequ
ence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == None:des
t = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + 
'.png')try:plt.show()except:passreturnclass QualSource:def __init__(self, quals_
file_path, lazy_init
>temp_1|source:-*- coding: utf-8 -*-|start:5244|stop:5744
on) + 1, ytickstep),[y for y in range(0, max(seq_len_distribution) + 1, ytickste
p)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, yma
x = max(seq_len_distribution) + (max(seq_len_distribution) / 20.0))plt.figtext(0
.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-large', ha = 'center')ax1
 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)
plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %
.2f / min: %s / max:
>temp_2|source:-*- coding: utf-8 -*-|start:4017|stop:4517
matplotlib.gridspec as gridspecsequence_lengths = []self.reset()while self.next(
):if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fastalib] Readi
ng: %s' % (self.pos))sys.stderr.flush()sequence_lengths.append(len(self.seq))sel
f.reset()sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(sequence_len
gths) + (int(max(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (
max_seq_len + 1)for l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.
figure(figsize = (16
>temp_3|source:-*- coding: utf-8 -*-|start:6776|stop:7276
plit()]self.pos += 1return Trueelse:return Falseif line.startswith('>'):self.fil
e_pointer.seek(self.file_pointer.tell() - len(line))breakqualscores += ' ' + lin
e.strip()self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.q
uals.split()]self.pos += 1return Truedef close(self):self.file_pointer.close()de
f reset(self):self.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Nones
elf.ids = []self.file_pointer.seek(0)if __name__ == '__main__':fasta = SequenceS
ource(sys.argv[1])fa
>temp_4|source:-*- coding: utf-8 -*-|start:4142|stop:4642
ys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()sequen
ce_lengths.append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_seq
_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) o
r 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq
_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update
({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewidth=
0.1)gs = gridspec.Gr
>temp_5|source:-*- coding: utf-8 -*-|start:6202|stop:6702
= lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself
.ids = []self.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)
if self.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l for l in
 self.file_pointer.readlines() if l.startswith('>')])self.reset()def next(self):
self.id = self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_in
t = Nonequalscores = ''while 1:line = self.file_pointer.readline()if not line:if
 len(qualscores):sel
>temp_6|source:-*- coding: utf-8 -*-|start:2084|stop:2584
 for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()if sel
f.unique:self.init_unique_hash()def init_unique_hash(self):while self.next_regul
ar():hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_has
h_dict:self.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[h
ash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [sel
f.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([
(self.unique_hash_di
>temp_7|source:-*- coding: utf-8 -*-|start:6069|stop:6569
ualSource:def __init__(self, quals_file_path, lazy_init = True):self.quals_file_
path = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self
.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer 
= open(self.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.tota
l_quals = Noneelse:self.total_quals = len([l for l in self.file_pointer.readline
s() if l.startswith('>')])self.reset()def next(self):self.id = self.file_pointer
.readline()[1:].stri
>temp_8|source:-*- coding: utf-8 -*-|start:974|stop:1474
(sequence), piece_length) + [len(sequence)]return '\n'.join([sequence[ticks[x]:t
icks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.output_file
_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = []self.sequenc
es = []self.fasta = SequenceSource(f_name)while self.fasta.next():if self.fasta.
pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FA
STA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.f
asta.id)self.sequenc
>temp_9|source:-*- coding: utf-8 -*-|start:3960|stop:4460
 ytickstep = None):import matplotlib.pyplot as pltimport matplotlib.gridspec as 
gridspecsequence_lengths = []self.reset()while self.next():if self.pos % 10000 =
= 0 or self.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sy
s.stderr.flush()sequence_lengths.append(len(self.seq))self.reset()sys.stderr.wri
te('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequen
ce_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l i
n sequence_lengths:s
>temp_10|source:-*- coding: utf-8 -*-|start:6709|stop:7209
 = qualscores.strip()self.quals_int = [int(q) for q in self.quals.split()]self.p
os += 1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.see
k(self.file_pointer.tell() - len(line))breakqualscores += ' ' + line.strip()self
.quals = qualscores.strip()self.quals_int = [int(q) for q in self.quals.split()]
self.pos += 1return Truedef close(self):self.file_pointer.close()def reset(self)
:self.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []s
elf.file_pointer.see
>temp_11|source:-*- coding: utf-8 -*-|start:6187|stop:6687
self.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals
_int = Noneself.ids = []self.file_pointer = open(self.quals_file_path)self.file_
pointer.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = 
len([l for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()
def next(self):self.id = self.file_pointer.readline()[1:].strip()self.quals = No
neself.quals_int = Nonequalscores = ''while 1:line = self.file_pointer.readline(
)if not line:if len(
>temp_12|source:-*- coding: utf-8 -*-|start:2857|stop:3357
nique > 0 and self.pos < self.total_unique:hash_entry = self.unique_hash_dict[se
lf.unique_hash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.
allow_mixed_case else hash_entry['seq'].upper()self.id  = hash_entry['id']self.i
ds = hash_entry['ids']return Trueelse:return Falseelse:return Falsedef next_regu
lar(self):self.seq = Noneself.id = self.file_pointer.readline()[1:].strip()seque
nce = ''while 1:line = self.file_pointer.readline()if not line:if len(sequence):
self.seq = sequences
>temp_13|source:-*- coding: utf-8 -*-|start:536|stop:1036
 = True, store_frequencies = True):if entry.unique and store_frequencies:self.wr
ite_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(
entry.id)self.write_seq(entry.seq, split)def write_id(self, id):self.output_file
_obj.write('>%s\n' % id)def write_seq(self, seq, split = True):if split:seq = se
lf.split(seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequence, p
iece_length = 80):ticks = range(0, len(sequence), piece_length) + [len(sequence)
]return '\n'.join([s
>temp_14|source:-*- coding: utf-8 -*-|start:5061|stop:5561
== None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickste
p, max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0,
 max(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_dist
ribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len
)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_distribution
) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-la
rge', ha = 'center')
>temp_15|source:-*- coding: utf-8 -*-|start:2309|stop:2809
self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].append(self.id)self.uni
que_hash_dict[hash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : self
.id,'ids': [self.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] fo
r i in sorted([(self.unique_hash_dict[hash]['count'], hash)\for hash in self.uni
que_hash_dict], reverse = True)]self.total_unique = len(self.unique_hash_dict)se
lf.reset()def next(self):if self.unique:return self.next_unique()else:return sel
f.next_regular()def 
>temp_16|source:-*- coding: utf-8 -*-|start:1770|stop:2270
_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []s
elf.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list = []self.uniq
ue_next_hash = 0self.file_pointer = open(self.fasta_file_path)self.file_pointer.
seek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for l
 in self.file_pointer.readlines() if l.startswith('>')])self.reset()if self.uniq
ue:self.init_unique_hash()def init_unique_hash(self):while self.next_regular():h
ash = hashlib.sha1(s
>temp_17|source:-*- coding: utf-8 -*-|start:5189|stop:5689
e='xx-small')plt.yticks(range(0, max(seq_len_distribution) + 1, ytickstep),[y fo
r y in range(0, max(seq_len_distribution) + 1, ytickstep)],size='xx-small')plt.x
lim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribut
ion) + (max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title),
 weight = 'black', size = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.
rcParams.update({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])plt.xticks(
[])plt.text(0.02, 0.
>temp_18|source:-*- coding: utf-8 -*-|start:2427|stop:2927
self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'
count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.unique_hash_dict[
hash]['count'], hash)\for hash in self.unique_hash_dict], reverse = True)]self.t
otal_unique = len(self.unique_hash_dict)self.reset()def next(self):if self.uniqu
e:return self.next_unique()else:return self.next_regular()def next_unique(self):
if self.unique:if self.total_unique > 0 and self.pos < self.total_unique:hash_en
try = self.unique_ha
>temp_19|source:-*- coding: utf-8 -*-|start:3664|stop:4164
t():if self.id == read_id:return self.seqreturn Falsedef close(self):self.file_p
ointer.close()def reset(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids
 = []self.file_pointer.seek(0)def visualize_sequence_length_distribution(self, t
itle, dest = None, max_seq_len = None, xtickstep = None, ytickstep = None):impor
t matplotlib.pyplot as pltimport matplotlib.gridspec as gridspecsequence_lengths
 = []self.reset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:sys
.stderr.write('\r[fa
>temp_20|source:-*- coding: utf-8 -*-|start:57|stop:557
ram is free software; you can redistribute it and/or modify it under# the terms 
of the GNU General Public License as published by the Free# Software Foundation;
 either version 2 of the License, or (at your option)# any later version.## Plea
se read the docs/COPYING file.import sysimport numpyimport hashlibclass FastaOut
put:def __init__(self, output_file_path):self.output_file_path = output_file_pat
hself.output_file_obj = open(output_file_path, 'w')def store(self, entry, split 
= True, store_freque
>temp_21|source:-*- coding: utf-8 -*-|start:4538|stop:5038
date({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewi
dth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt
.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq
_len_distribution, color = 'black', alpha = 0.3)plt.fill_between(range(0, max_se
q_len + 1), seq_len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.ylab
el('number of sequences')plt.xlabel('sequence length')if xtickstep == None:xtick
step = (max_seq_len 
>temp_22|source:-*- coding: utf-8 -*-|start:1074|stop:1574
n range(0, len(ticks) - 1)])def close(self):self.output_file_obj.close()class Re
adFasta:def __init__(self, f_name):self.ids = []self.sequences = []self.fasta = 
SequenceSource(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0 or s
elf.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s'
 % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequenc
es.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close(
)class SequenceSourc
>temp_23|source:-*- coding: utf-8 -*-|start:44|stop:544
n## This program is free software; you can redistribute it and/or modify it unde
r# the terms of the GNU General Public License as published by the Free# Softwar
e Foundation; either version 2 of the License, or (at your option)# any later ve
rsion.## Please read the docs/COPYING file.import sysimport numpyimport hashlibc
lass FastaOutput:def __init__(self, output_file_path):self.output_file_path = ou
tput_file_pathself.output_file_obj = open(output_file_path, 'w')def store(self, 
entry, split = True,
>temp_24|source:-*- coding: utf-8 -*-|start:5830|stop:6330
),\min(sequence_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, si
ze = 'x-large')if dest == None:dest = self.fasta_file_pathtry:plt.savefig(dest +
 '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passreturnclass Qu
alSource:def __init__(self, quals_file_path, lazy_init = True):self.quals_file_p
ath = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.
id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer =
 open(self.quals_fil
>temp_25|source:-*- coding: utf-8 -*-|start:2806|stop:3306
ef next_unique(self):if self.unique:if self.total_unique > 0 and self.pos < self
.total_unique:hash_entry = self.unique_hash_dict[self.unique_hash_list[self.pos]
]self.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_case else hash_en
try['seq'].upper()self.id  = hash_entry['id']self.ids = hash_entry['ids']return 
Trueelse:return Falseelse:return Falsedef next_regular(self):self.seq = Noneself
.id = self.file_pointer.readline()[1:].strip()sequence = ''while 1:line = self.f
ile_pointer.readline
>temp_26|source:-*- coding: utf-8 -*-|start:5759|stop:6259
uence_lengths),numpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min(se
quence_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, size = 'x-l
arge')if dest == None:dest = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')e
xcept:plt.savefig(dest + '.png')try:plt.show()except:passreturnclass QualSource:
def __init__(self, quals_file_path, lazy_init = True):self.quals_file_path = qua
ls_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Non
eself.quals = Nonese
>temp_27|source:-*- coding: utf-8 -*-|start:4441|stop:4941
 sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12
))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', line
style='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])
plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 
0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.fill_betwee
n(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black', alph
a = 0.15)plt.ylabel(
>temp_28|source:-*- coding: utf-8 -*-|start:6353|stop:6853
r.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l
 for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()def ne
xt(self):self.id = self.file_pointer.readline()[1:].strip()self.quals = Noneself
.quals_int = Nonequalscores = ''while 1:line = self.file_pointer.readline()if no
t line:if len(qualscores):self.quals = qualscores.strip()self.quals_int = [int(q
) for q in self.quals.split()]self.pos += 1return Trueelse:return Falseif line.s
tartswith('>'):self.
>temp_29|source:-*- coding: utf-8 -*-|start:4447|stop:4947
nce_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.
rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle=
'-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.gr
id(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)p
lt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.fill_between(rang
e(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black', alpha = 0.
15)plt.ylabel('numbe
>temp_30|source:-*- coding: utf-8 -*-|start:5498|stop:5998
 % (title), weight = 'black', size = 'xx-large', ha = 'center')ax1 = plt.subplot
(gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])
plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s 
/ max: %s'\% (len(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(sequ
ence_lengths),\min(sequence_lengths),\max(sequence_lengths)),\va = 'center', alp
ha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_pathtry:plt.sa
vefig(dest + '.pdf')
>temp_31|source:-*- coding: utf-8 -*-|start:1875|stop:2375
nique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0self.file
_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy_init:
self.total_seq = Noneelse:self.total_seq = len([l for l in self.file_pointer.rea
dlines() if l.startswith('>')])self.reset()if self.unique:self.init_unique_hash(
)def init_unique_hash(self):while self.next_regular():hash = hashlib.sha1(self.s
eq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_hash_dict[ha
sh]['ids'].append(se
>temp_32|source:-*- coding: utf-8 -*-|start:6083|stop:6583
__init__(self, quals_file_path, lazy_init = True):self.quals_file_path = quals_f
ile_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Nonesel
f.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer = open(self.qu
als_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_quals = None
else:self.total_quals = len([l for l in self.file_pointer.readlines() if l.start
swith('>')])self.reset()def next(self):self.id = self.file_pointer.readline()[1:
].strip()self.quals 
>temp_33|source:-*- coding: utf-8 -*-|start:4790|stop:5290
ution, color = 'black', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), 
seq_len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number o
f sequences')plt.xlabel('sequence length')if xtickstep == None:xtickstep = (max_
seq_len / 50) or 1if ytickstep == None:ytickstep = max(seq_len_distribution) / 2
0 or 1plt.xticks(range(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size
='xx-small')plt.yticks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for
 y in range(0, max(s
>temp_34|source:-*- coding: utf-8 -*-|start:5675|stop:6175
.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len
(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(sequence_lengths),\mi
n(sequence_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, size = 
'x-large')if dest == None:dest = self.fasta_file_pathtry:plt.savefig(dest + '.pd
f')except:plt.savefig(dest + '.png')try:plt.show()except:passreturnclass QualSou
rce:def __init__(self, quals_file_path, lazy_init = True):self.quals_file_path =
 quals_file_pathself
>temp_35|source:-*- coding: utf-8 -*-|start:6187|stop:6687
self.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals
_int = Noneself.ids = []self.file_pointer = open(self.quals_file_path)self.file_
pointer.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = 
len([l for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()
def next(self):self.id = self.file_pointer.readline()[1:].strip()self.quals = No
neself.quals_int = Nonequalscores = ''while 1:line = self.file_pointer.readline(
)if not line:if len(
>temp_36|source:-*- coding: utf-8 -*-|start:2189|stop:2689
_hash()def init_unique_hash(self):while self.next_regular():hash = hashlib.sha1(
self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_hash_d
ict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] += 1else:se
lf.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'co
unt': 1}self.unique_hash_list = [i[1] for i in sorted([(self.unique_hash_dict[ha
sh]['count'], hash)\for hash in self.unique_hash_dict], reverse = True)]self.tot
al_unique = len(self
>temp_37|source:-*- coding: utf-8 -*-|start:5092|stop:5592
_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1, xtickstep),
 rotation=90, size='xx-small')plt.yticks(range(0, max(seq_len_distribution) + 1,
 ytickstep),[y for y in range(0, max(seq_len_distribution) + 1, ytickstep)],size
='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(
seq_len_distribution) + (max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96
, '%s' % (title), weight = 'black', size = 'xx-large', ha = 'center')ax1 = plt.s
ubplot(gs[9])plt.rcP
>temp_38|source:-*- coding: utf-8 -*-|start:90|stop:590
istribute it and/or modify it under# the terms of the GNU General Public License
 as published by the Free# Software Foundation; either version 2 of the License,
 or (at your option)# any later version.## Please read the docs/COPYING file.imp
ort sysimport numpyimport hashlibclass FastaOutput:def __init__(self, output_fil
e_path):self.output_file_path = output_file_pathself.output_file_obj = open(outp
ut_file_path, 'w')def store(self, entry, split = True, store_frequencies = True)
:if entry.unique and
>temp_39|source:-*- coding: utf-8 -*-|start:3991|stop:4491
otlib.pyplot as pltimport matplotlib.gridspec as gridspecsequence_lengths = []se
lf.reset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.stderr
.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_length
s.append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_seq_len:max_
seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or 10)seq_
len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_dist
ribution[l] += 1fig 
>temp_40|source:-*- coding: utf-8 -*-|start:1281|stop:1781
f self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastali
b] Reading FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.
append(self.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')
def close(self):self.fasta.close()class SequenceSource():def __init__(self, fast
a_file_path, lazy_init = True, unique = False, allow_mixed_case = False):self.fa
sta_file_path = fasta_file_pathself.name = Noneself.lazy_init = lazy_initself.al
low_mixed_case = all
>temp_41|source:-*- coding: utf-8 -*-|start:782|stop:1282
rite('>%s\n' % id)def write_seq(self, seq, split = True):if split:seq = self.spl
it(seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequence, piece_l
ength = 80):ticks = range(0, len(sequence), piece_length) + [len(sequence)]retur
n '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])
def close(self):self.output_file_obj.close()class ReadFasta:def __init__(self, f
_name):self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)while 
self.fasta.next():if
>temp_42|source:-*- coding: utf-8 -*-|start:4270|stop:4770
stderr.write('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(
max(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len +
 1)for l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsiz
e = (16, 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='
0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subpl
ot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.9
5, right = 0.98)plt.
>temp_43|source:-*- coding: utf-8 -*-|start:5234|stop:5734
distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_distribution) + 1
, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymi
n = 0, ymax = max(seq_len_distribution) + (max(seq_len_distribution) / 20.0))plt
.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-large', ha = 'c
enter')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.g
rid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2
f / std: %.2f / min:
>temp_44|source:-*- coding: utf-8 -*-|start:840|stop:1340
f split:seq = self.split(seq)self.output_file_obj.write('%s\n' % seq)def split(s
elf, sequence, piece_length = 80):ticks = range(0, len(sequence), piece_length) 
+ [len(sequence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in rang
e(0, len(ticks) - 1)])def close(self):self.output_file_obj.close()class ReadFast
a:def __init__(self, f_name):self.ids = []self.sequences = []self.fasta = Sequen
ceSource(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fa
sta.pos == 1:sys.std
>temp_45|source:-*- coding: utf-8 -*-|start:4229|stop:4729
ths.append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_seq_len:ma
x_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or 10)se
q_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_di
stribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update({'axes
.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewidth=0.1)gs 
= gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_a
djust(left=0.05, bot
>temp_46|source:-*- coding: utf-8 -*-|start:2920|stop:3420
ique_hash_dict[self.unique_hash_list[self.pos]]self.pos += 1self.seq = hash_entr
y['seq'] if self.allow_mixed_case else hash_entry['seq'].upper()self.id  = hash_
entry['id']self.ids = hash_entry['ids']return Trueelse:return Falseelse:return F
alsedef next_regular(self):self.seq = Noneself.id = self.file_pointer.readline()
[1:].strip()sequence = ''while 1:line = self.file_pointer.readline()if not line:
if len(sequence):self.seq = sequenceself.pos += 1return Trueelse:return Falseif 
line.startswith('>')
>temp_47|source:-*- coding: utf-8 -*-|start:5146|stop:5646
x_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, max(
seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_distribut
ion) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.
ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_distribution) / 2
0.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-large',
 ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' : 2
0})plt.grid(False)pl
>temp_48|source:-*- coding: utf-8 -*-|start:3055|stop:3555
].upper()self.id  = hash_entry['id']self.ids = hash_entry['ids']return Trueelse:
return Falseelse:return Falsedef next_regular(self):self.seq = Noneself.id = sel
f.file_pointer.readline()[1:].strip()sequence = ''while 1:line = self.file_point
er.readline()if not line:if len(sequence):self.seq = sequenceself.pos += 1return
 Trueelse:return Falseif line.startswith('>'):self.file_pointer.seek(self.file_p
ointer.tell() - len(line))breaksequence += line.strip()self.seq = sequence if se
lf.allow_mixed_case 
>temp_49|source:-*- coding: utf-8 -*-|start:1812|stop:2312
d  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict =
 {}self.unique_hash_list = []self.unique_next_hash = 0self.file_pointer = open(s
elf.fasta_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = 
Noneelse:self.total_seq = len([l for l in self.file_pointer.readlines() if l.sta
rtswith('>')])self.reset()if self.unique:self.init_unique_hash()def init_unique_
hash(self):while self.next_regular():hash = hashlib.sha1(self.seq.upper()).hexdi
gest()if hash in sel
>temp_50|source:-*- coding: utf-8 -*-|start:3562|stop:4062
quence.upper()self.pos += 1return Truedef get_seq_by_read_id(self, read_id):self
.reset()while self.next():if self.id == read_id:return self.seqreturn Falsedef c
lose(self):self.file_pointer.close()def reset(self):self.pos = 0self.id  = Nones
elf.seq = Noneself.ids = []self.file_pointer.seek(0)def visualize_sequence_lengt
h_distribution(self, title, dest = None, max_seq_len = None, xtickstep = None, y
tickstep = None):import matplotlib.pyplot as pltimport matplotlib.gridspec as gr
idspecsequence_lengt
>temp_51|source:-*- coding: utf-8 -*-|start:4223|stop:4723
e_lengths.append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_seq_
len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or
 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_
len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update(
{'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewidth=0
.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subp
lots_adjust(left=0.0
>temp_52|source:-*- coding: utf-8 -*-|start:627|stop:1127
%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)se
lf.write_seq(entry.seq, split)def write_id(self, id):self.output_file_obj.write(
'>%s\n' % id)def write_seq(self, seq, split = True):if split:seq = self.split(se
q)self.output_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length
 = 80):ticks = range(0, len(sequence), piece_length) + [len(sequence)]return '\n
'.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def c
lose(self):self.outp
>temp_53|source:-*- coding: utf-8 -*-|start:4408|stop:4908
= [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_distribution[l] += 1f
ig = plt.figure(figsize = (16, 12))plt.rcParams.update({'axes.linewidth' : 0.9})
plt.rc('grid', color='0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec
(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bo
ttom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_distribution, color = 'bl
ack', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distributi
on, y2 = 0, color = 
>temp_54|source:-*- coding: utf-8 -*-|start:96|stop:596
ute it and/or modify it under# the terms of the GNU General Public License as pu
blished by the Free# Software Foundation; either version 2 of the License, or (a
t your option)# any later version.## Please read the docs/COPYING file.import sy
simport numpyimport hashlibclass FastaOutput:def __init__(self, output_file_path
):self.output_file_path = output_file_pathself.output_file_obj = open(output_fil
e_path, 'w')def store(self, entry, split = True, store_frequencies = True):if en
try.unique and store
>temp_55|source:-*- coding: utf-8 -*-|start:891|stop:1391
rite('%s\n' % seq)def split(self, sequence, piece_length = 80):ticks = range(0, 
len(sequence), piece_length) + [len(sequence)]return '\n'.join([sequence[ticks[x
]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.output_f
ile_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = []self.sequ
ences = []self.fasta = SequenceSource(f_name)while self.fasta.next():if self.fas
ta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading
 FASTA into memory: 
>temp_56|source:-*- coding: utf-8 -*-|start:2012|stop:2512
ek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for l i
n self.file_pointer.readlines() if l.startswith('>')])self.reset()if self.unique
:self.init_unique_hash()def init_unique_hash(self):while self.next_regular():has
h = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:s
elf.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['co
unt'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'s
eq': self.seq,'count
>temp_57|source:-*- coding: utf-8 -*-|start:3312|stop:3812
ot line:if len(sequence):self.seq = sequenceself.pos += 1return Trueelse:return 
Falseif line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - l
en(line))breaksequence += line.strip()self.seq = sequence if self.allow_mixed_ca
se else sequence.upper()self.pos += 1return Truedef get_seq_by_read_id(self, rea
d_id):self.reset()while self.next():if self.id == read_id:return self.seqreturn 
Falsedef close(self):self.file_pointer.close()def reset(self):self.pos = 0self.i
d  = Noneself.seq = 
>temp_58|source:-*- coding: utf-8 -*-|start:1762|stop:2262
ow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.i
ds = []self.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list = []s
elf.unique_next_hash = 0self.file_pointer = open(self.fasta_file_path)self.file_
pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len(
[l for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()if s
elf.unique:self.init_unique_hash()def init_unique_hash(self):while self.next_reg
ular():hash = hashli
>temp_59|source:-*- coding: utf-8 -*-|start:2874|stop:3374
f.pos < self.total_unique:hash_entry = self.unique_hash_dict[self.unique_hash_li
st[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_case 
else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_entry['
ids']return Trueelse:return Falseelse:return Falsedef next_regular(self):self.se
q = Noneself.id = self.file_pointer.readline()[1:].strip()sequence = ''while 1:l
ine = self.file_pointer.readline()if not line:if len(sequence):self.seq = sequen
ceself.pos += 1retur
>temp_60|source:-*- coding: utf-8 -*-|start:5888|stop:6388
nter', alpha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_path
try:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()exc
ept:passreturnclass QualSource:def __init__(self, quals_file_path, lazy_init = T
rue):self.quals_file_path = quals_file_pathself.name = Noneself.lazy_init = lazy
_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids =
 []self.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if sel
f.lazy_init:self.tot
>temp_61|source:-*- coding: utf-8 -*-|start:1387|stop:1887
ry: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.
sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta
.close()class SequenceSource():def __init__(self, fasta_file_path, lazy_init = T
rue, unique = False, allow_mixed_case = False):self.fasta_file_path = fasta_file
_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mix
ed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = uniq
ueself.unique_hash_d
>temp_62|source:-*- coding: utf-8 -*-|start:4120|stop:4620
= 0 or self.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sy
s.stderr.flush()sequence_lengths.append(len(self.seq))self.reset()sys.stderr.wri
te('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequen
ce_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l i
n sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 1
2))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', lin
estyle='-', linewidt
>temp_63|source:-*- coding: utf-8 -*-|start:1752|stop:2252
itself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = 
Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_hash_
list = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_file_path)
self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_
seq = len([l for l in self.file_pointer.readlines() if l.startswith('>')])self.r
eset()if self.unique:self.init_unique_hash()def init_unique_hash(self):while sel
f.next_regular():has
>temp_64|source:-*- coding: utf-8 -*-|start:2489|stop:2989
,'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self
.unique_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_dict], reve
rse = True)]self.total_unique = len(self.unique_hash_dict)self.reset()def next(s
elf):if self.unique:return self.next_unique()else:return self.next_regular()def 
next_unique(self):if self.unique:if self.total_unique > 0 and self.pos < self.to
tal_unique:hash_entry = self.unique_hash_dict[self.unique_hash_list[self.pos]]se
lf.pos += 1self.seq 
>temp_65|source:-*- coding: utf-8 -*-|start:1417|stop:1917
.stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.se
q)sys.stderr.write('\n')def close(self):self.fasta.close()class SequenceSource()
:def __init__(self, fasta_file_path, lazy_init = True, unique = False, allow_mix
ed_case = False):self.fasta_file_path = fasta_file_pathself.name = Noneself.lazy
_init = lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = 
Noneself.seq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}se
lf.unique_hash_list 
>temp_66|source:-*- coding: utf-8 -*-|start:2591|stop:3091
]['count'], hash)\for hash in self.unique_hash_dict], reverse = True)]self.total
_unique = len(self.unique_hash_dict)self.reset()def next(self):if self.unique:re
turn self.next_unique()else:return self.next_regular()def next_unique(self):if s
elf.unique:if self.total_unique > 0 and self.pos < self.total_unique:hash_entry 
= self.unique_hash_dict[self.unique_hash_list[self.pos]]self.pos += 1self.seq = 
hash_entry['seq'] if self.allow_mixed_case else hash_entry['seq'].upper()self.id
  = hash_entry['id']
>temp_67|source:-*- coding: utf-8 -*-|start:3100|stop:3600
= hash_entry['ids']return Trueelse:return Falseelse:return Falsedef next_regular
(self):self.seq = Noneself.id = self.file_pointer.readline()[1:].strip()sequence
 = ''while 1:line = self.file_pointer.readline()if not line:if len(sequence):sel
f.seq = sequenceself.pos += 1return Trueelse:return Falseif line.startswith('>')
:self.file_pointer.seek(self.file_pointer.tell() - len(line))breaksequence += li
ne.strip()self.seq = sequence if self.allow_mixed_case else sequence.upper()self
.pos += 1return True
>temp_68|source:-*- coding: utf-8 -*-|start:5429|stop:5929
ion) + (max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title),
 weight = 'black', size = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.
rcParams.update({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])plt.xticks(
[])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\
% (len(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(sequence_length
s),\min(sequence_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, s
ize = 'x-large')if d
>temp_69|source:-*- coding: utf-8 -*-|start:2262|stop:2762
b.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique
_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] += 1
else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.
seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.unique_hash_
dict[hash]['count'], hash)\for hash in self.unique_hash_dict], reverse = True)]s
elf.total_unique = len(self.unique_hash_dict)self.reset()def next(self):if self.
unique:return self.n
>temp_70|source:-*- coding: utf-8 -*-|start:1446|stop:1946
d(self.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def c
lose(self):self.fasta.close()class SequenceSource():def __init__(self, fasta_fil
e_path, lazy_init = True, unique = False, allow_mixed_case = False):self.fasta_f
ile_path = fasta_file_pathself.name = Noneself.lazy_init = lazy_initself.allow_m
ixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids =
 []self.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list = []self.
unique_next_hash = 0
>temp_71|source:-*- coding: utf-8 -*-|start:3762|stop:4262
reset(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.file_poi
nter.seek(0)def visualize_sequence_length_distribution(self, title, dest = None,
 max_seq_len = None, xtickstep = None, ytickstep = None):import matplotlib.pyplo
t as pltimport matplotlib.gridspec as gridspecsequence_lengths = []self.reset()w
hile self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[
fastalib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_lengths.append(le
n(self.seq))self.res
>temp_72|source:-*- coding: utf-8 -*-|start:334|stop:834
sysimport numpyimport hashlibclass FastaOutput:def __init__(self, output_file_pa
th):self.output_file_path = output_file_pathself.output_file_obj = open(output_f
ile_path, 'w')def store(self, entry, split = True, store_frequencies = True):if 
entry.unique and store_frequencies:self.write_id('%s|%s' % (entry.id, 'frequency
:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_seq(entry.seq, spl
it)def write_id(self, id):self.output_file_obj.write('>%s\n' % id)def write_seq(
self, seq, split = T
>temp_73|source:-*- coding: utf-8 -*-|start:2024|stop:2524
.lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for l in self.file_
pointer.readlines() if l.startswith('>')])self.reset()if self.unique:self.init_u
nique_hash()def init_unique_hash(self):while self.next_regular():hash = hashlib.
sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_h
ash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] += 1el
se:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.se
q,'count': 1}self.un
>temp_74|source:-*- coding: utf-8 -*-|start:289|stop:789
.## Please read the docs/COPYING file.import sysimport numpyimport hashlibclass 
FastaOutput:def __init__(self, output_file_path):self.output_file_path = output_
file_pathself.output_file_obj = open(output_file_path, 'w')def store(self, entry
, split = True, store_frequencies = True):if entry.unique and store_frequencies:
self.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.wr
ite_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, id):self.outp
ut_file_obj.write('>
>temp_75|source:-*- coding: utf-8 -*-|start:3852|stop:4352
0)def visualize_sequence_length_distribution(self, title, dest = None, max_seq_l
en = None, xtickstep = None, ytickstep = None):import matplotlib.pyplot as pltim
port matplotlib.gridspec as gridspecsequence_lengths = []self.reset()while self.
next():if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fastalib] 
Reading: %s' % (self.pos))sys.stderr.flush()sequence_lengths.append(len(self.seq
))self.reset()sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(sequenc
e_lengths) + (int(ma
>temp_76|source:-*- coding: utf-8 -*-|start:1074|stop:1574
n range(0, len(ticks) - 1)])def close(self):self.output_file_obj.close()class Re
adFasta:def __init__(self, f_name):self.ids = []self.sequences = []self.fasta = 
SequenceSource(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0 or s
elf.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s'
 % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequenc
es.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close(
)class SequenceSourc
>temp_77|source:-*- coding: utf-8 -*-|start:5772|stop:6272
),numpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_length
s),\max(sequence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest
 == None:dest = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.sav
efig(dest + '.png')try:plt.show()except:passreturnclass QualSource:def __init__(
self, quals_file_path, lazy_init = True):self.quals_file_path = quals_file_paths
elf.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals =
 Noneself.quals_int 
>temp_78|source:-*- coding: utf-8 -*-|start:4220|stop:4720
ence_lengths.append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_s
eq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0)
 or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:s
eq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.upda
te({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewidt
h=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.s
ubplots_adjust(left=
>temp_79|source:-*- coding: utf-8 -*-|start:3507|stop:4007
p()self.seq = sequence if self.allow_mixed_case else sequence.upper()self.pos +=
 1return Truedef get_seq_by_read_id(self, read_id):self.reset()while self.next()
:if self.id == read_id:return self.seqreturn Falsedef close(self):self.file_poin
ter.close()def reset(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids = 
[]self.file_pointer.seek(0)def visualize_sequence_length_distribution(self, titl
e, dest = None, max_seq_len = None, xtickstep = None, ytickstep = None):import m
atplotlib.pyplot as 
>temp_80|source:-*- coding: utf-8 -*-|start:6535|stop:7035
f.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_int = Nonequals
cores = ''while 1:line = self.file_pointer.readline()if not line:if len(qualscor
es):self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.quals.
split()]self.pos += 1return Trueelse:return Falseif line.startswith('>'):self.fi
le_pointer.seek(self.file_pointer.tell() - len(line))breakqualscores += ' ' + li
ne.strip()self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.
quals.split()]self.p
>temp_81|source:-*- coding: utf-8 -*-|start:3474|stop:3974
(line))breaksequence += line.strip()self.seq = sequence if self.allow_mixed_case
 else sequence.upper()self.pos += 1return Truedef get_seq_by_read_id(self, read_
id):self.reset()while self.next():if self.id == read_id:return self.seqreturn Fa
lsedef close(self):self.file_pointer.close()def reset(self):self.pos = 0self.id 
 = Noneself.seq = Noneself.ids = []self.file_pointer.seek(0)def visualize_sequen
ce_length_distribution(self, title, dest = None, max_seq_len = None, xtickstep =
 None, ytickstep = N
>temp_82|source:-*- coding: utf-8 -*-|start:576|stop:1076
try.unique and store_frequencies:self.write_id('%s|%s' % (entry.id, 'frequency:%
d' % len(entry.ids)))else:self.write_id(entry.id)self.write_seq(entry.seq, split
)def write_id(self, id):self.output_file_obj.write('>%s\n' % id)def write_seq(se
lf, seq, split = True):if split:seq = self.split(seq)self.output_file_obj.write(
'%s\n' % seq)def split(self, sequence, piece_length = 80):ticks = range(0, len(s
equence), piece_length) + [len(sequence)]return '\n'.join([sequence[ticks[x]:tic
ks[x + 1]] for x in 
>temp_83|source:-*- coding: utf-8 -*-|start:653|stop:1153
y:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_seq(entry.seq, sp
lit)def write_id(self, id):self.output_file_obj.write('>%s\n' % id)def write_seq
(self, seq, split = True):if split:seq = self.split(seq)self.output_file_obj.wri
te('%s\n' % seq)def split(self, sequence, piece_length = 80):ticks = range(0, le
n(sequence), piece_length) + [len(sequence)]return '\n'.join([sequence[ticks[x]:
ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.output_fil
e_obj.close()class R
>temp_84|source:-*- coding: utf-8 -*-|start:1488|stop:1988
.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close()class Sequenc
eSource():def __init__(self, fasta_file_path, lazy_init = True, unique = False, 
allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.name = None
self.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0se
lf.id  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.unique_hash_di
ct = {}self.unique_hash_list = []self.unique_next_hash = 0self.file_pointer = op
en(self.fasta_file_p
>temp_85|source:-*- coding: utf-8 -*-|start:3572|stop:4072
er()self.pos += 1return Truedef get_seq_by_read_id(self, read_id):self.reset()wh
ile self.next():if self.id == read_id:return self.seqreturn Falsedef close(self)
:self.file_pointer.close()def reset(self):self.pos = 0self.id  = Noneself.seq = 
Noneself.ids = []self.file_pointer.seek(0)def visualize_sequence_length_distribu
tion(self, title, dest = None, max_seq_len = None, xtickstep = None, ytickstep =
 None):import matplotlib.pyplot as pltimport matplotlib.gridspec as gridspecsequ
ence_lengths = []sel
>temp_86|source:-*- coding: utf-8 -*-|start:2470|stop:2970
id,'ids': [self.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for
 i in sorted([(self.unique_hash_dict[hash]['count'], hash)\for hash in self.uniq
ue_hash_dict], reverse = True)]self.total_unique = len(self.unique_hash_dict)sel
f.reset()def next(self):if self.unique:return self.next_unique()else:return self
.next_regular()def next_unique(self):if self.unique:if self.total_unique > 0 and
 self.pos < self.total_unique:hash_entry = self.unique_hash_dict[self.unique_has
h_list[self.pos]]sel
>temp_87|source:-*- coding: utf-8 -*-|start:1816|stop:2316
 Noneself.seq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}s
elf.unique_hash_list = []self.unique_next_hash = 0self.file_pointer = open(self.
fasta_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = None
else:self.total_seq = len([l for l in self.file_pointer.readlines() if l.startsw
ith('>')])self.reset()if self.unique:self.init_unique_hash()def init_unique_hash
(self):while self.next_regular():hash = hashlib.sha1(self.seq.upper()).hexdigest
()if hash in self.un
>temp_88|source:-*- coding: utf-8 -*-|start:2179|stop:2679
nit_unique_hash()def init_unique_hash(self):while self.next_regular():hash = has
hlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.uni
que_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] +
= 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': se
lf.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.unique_ha
sh_dict[hash]['count'], hash)\for hash in self.unique_hash_dict], reverse = True
)]self.total_unique 
>temp_89|source:-*- coding: utf-8 -*-|start:994|stop:1494
ngth) + [len(sequence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x i
n range(0, len(ticks) - 1)])def close(self):self.output_file_obj.close()class Re
adFasta:def __init__(self, f_name):self.ids = []self.sequences = []self.fasta = 
SequenceSource(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0 or s
elf.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s'
 % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequenc
es.append(self.fasta
>temp_90|source:-*- coding: utf-8 -*-|start:6189|stop:6689
lf.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_i
nt = Noneself.ids = []self.file_pointer = open(self.quals_file_path)self.file_po
inter.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = le
n([l for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()de
f next(self):self.id = self.file_pointer.readline()[1:].strip()self.quals = None
self.quals_int = Nonequalscores = ''while 1:line = self.file_pointer.readline()i
f not line:if len(qu
>temp_91|source:-*- coding: utf-8 -*-|start:3887|stop:4387
tribution(self, title, dest = None, max_seq_len = None, xtickstep = None, yticks
tep = None):import matplotlib.pyplot as pltimport matplotlib.gridspec as gridspe
csequence_lengths = []self.reset()while self.next():if self.pos % 10000 == 0 or 
self.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.stder
r.flush()sequence_lengths.append(len(self.seq))self.reset()sys.stderr.write('\n'
)if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_leng
ths) / 100.0) or 10)
>temp_92|source:-*- coding: utf-8 -*-|start:652|stop:1152
cy:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_seq(entry.seq, s
plit)def write_id(self, id):self.output_file_obj.write('>%s\n' % id)def write_se
q(self, seq, split = True):if split:seq = self.split(seq)self.output_file_obj.wr
ite('%s\n' % seq)def split(self, sequence, piece_length = 80):ticks = range(0, l
en(sequence), piece_length) + [len(sequence)]return '\n'.join([sequence[ticks[x]
:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.output_fi
le_obj.close()class 
>temp_93|source:-*- coding: utf-8 -*-|start:6086|stop:6586
nit__(self, quals_file_path, lazy_init = True):self.quals_file_path = quals_file
_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.q
uals = Noneself.quals_int = Noneself.ids = []self.file_pointer = open(self.quals
_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_quals = Noneels
e:self.total_quals = len([l for l in self.file_pointer.readlines() if l.startswi
th('>')])self.reset()def next(self):self.id = self.file_pointer.readline()[1:].s
trip()self.quals = N
>temp_94|source:-*- coding: utf-8 -*-|start:77|stop:577
; you can redistribute it and/or modify it under# the terms of the GNU General P
ublic License as published by the Free# Software Foundation; either version 2 of
 the License, or (at your option)# any later version.## Please read the docs/COP
YING file.import sysimport numpyimport hashlibclass FastaOutput:def __init__(sel
f, output_file_path):self.output_file_path = output_file_pathself.output_file_ob
j = open(output_file_path, 'w')def store(self, entry, split = True, store_freque
ncies = True):if ent
>temp_95|source:-*- coding: utf-8 -*-|start:4799|stop:5299
lor = 'black', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_d
istribution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequenc
es')plt.xlabel('sequence length')if xtickstep == None:xtickstep = (max_seq_len /
 50) or 1if ytickstep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt
.xticks(range(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-smal
l')plt.yticks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y in ran
ge(0, max(seq_len_di
>temp_96|source:-*- coding: utf-8 -*-|start:4534|stop:5034
s.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', li
newidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True
)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot
(seq_len_distribution, color = 'black', alpha = 0.3)plt.fill_between(range(0, ma
x_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.
ylabel('number of sequences')plt.xlabel('sequence length')if xtickstep == None:x
tickstep = (max_seq_
>temp_97|source:-*- coding: utf-8 -*-|start:2400|stop:2900
ct[hash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': 
[self.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sort
ed([(self.unique_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_di
ct], reverse = True)]self.total_unique = len(self.unique_hash_dict)self.reset()d
ef next(self):if self.unique:return self.next_unique()else:return self.next_regu
lar()def next_unique(self):if self.unique:if self.total_unique > 0 and self.pos 
< self.total_unique:
>temp_98|source:-*- coding: utf-8 -*-|start:4075|stop:4575
eset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.wri
te('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_lengths.ap
pend(len(self.seq))self.reset()sys.stderr.write('\n')if not max_seq_len:max_seq_
len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or 10)seq_len_
distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_distribu
tion[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update({'axes.linew
idth' : 0.9})plt.rc(
>temp_99|source:-*- coding: utf-8 -*-|start:4082|stop:4582
hile self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[
fastalib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_lengths.append(le
n(self.seq))self.reset()sys.stderr.write('\n')if not max_seq_len:max_seq_len = m
ax(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or 10)seq_len_distrib
ution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_distribution[l]
 += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update({'axes.linewidth' :
 0.9})plt.rc('grid',
>temp_100|source:-*- coding: utf-8 -*-|start:5738|stop:6238
/ max: %s'\% (len(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(sequ
ence_lengths),\min(sequence_lengths),\max(sequence_lengths)),\va = 'center', alp
ha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_pathtry:plt.sa
vefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passre
turnclass QualSource:def __init__(self, quals_file_path, lazy_init = True):self.
quals_file_path = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.
pos = 0self.id  = No
>temp_101|source:-*- coding: utf-8 -*-|start:3633|stop:4133
_id):self.reset()while self.next():if self.id == read_id:return self.seqreturn F
alsedef close(self):self.file_pointer.close()def reset(self):self.pos = 0self.id
  = Noneself.seq = Noneself.ids = []self.file_pointer.seek(0)def visualize_seque
nce_length_distribution(self, title, dest = None, max_seq_len = None, xtickstep 
= None, ytickstep = None):import matplotlib.pyplot as pltimport matplotlib.grids
pec as gridspecsequence_lengths = []self.reset()while self.next():if self.pos % 
10000 == 0 or self.p
>temp_102|source:-*- coding: utf-8 -*-|start:5203|stop:5703
lt.yticks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y in range(0
, max(seq_len_distribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, 
xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(se
q_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'bla
ck', size = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.updat
e({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(0.
02, 0.5, 'total: %s 
>temp_103|source:-*- coding: utf-8 -*-|start:1177|stop:1677
lf, f_name):self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)w
hile self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.
stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))sys
.stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.se
q)sys.stderr.write('\n')def close(self):self.fasta.close()class SequenceSource()
:def __init__(self, fasta_file_path, lazy_init = True, unique = False, allow_mix
ed_case = False):sel
>temp_104|source:-*- coding: utf-8 -*-|start:4937|stop:5437
bel('number of sequences')plt.xlabel('sequence length')if xtickstep == None:xtic
kstep = (max_seq_len / 50) or 1if ytickstep == None:ytickstep = max(seq_len_dist
ribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1, xtickstep), rota
tion=90, size='xx-small')plt.yticks(range(0, max(seq_len_distribution) + 1, ytic
kstep),[y for y in range(0, max(seq_len_distribution) + 1, ytickstep)],size='xx-
small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_l
en_distribution) + (
>temp_105|source:-*- coding: utf-8 -*-|start:6413|stop:6913
total_quals = len([l for l in self.file_pointer.readlines() if l.startswith('>')
])self.reset()def next(self):self.id = self.file_pointer.readline()[1:].strip()s
elf.quals = Noneself.quals_int = Nonequalscores = ''while 1:line = self.file_poi
nter.readline()if not line:if len(qualscores):self.quals = qualscores.strip()sel
f.quals_int = [int(q) for q in self.quals.split()]self.pos += 1return Trueelse:r
eturn Falseif line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell
() - len(line))break
>temp_106|source:-*- coding: utf-8 -*-|start:2939|stop:3439
.unique_hash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.al
low_mixed_case else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids
 = hash_entry['ids']return Trueelse:return Falseelse:return Falsedef next_regula
r(self):self.seq = Noneself.id = self.file_pointer.readline()[1:].strip()sequenc
e = ''while 1:line = self.file_pointer.readline()if not line:if len(sequence):se
lf.seq = sequenceself.pos += 1return Trueelse:return Falseif line.startswith('>'
):self.file_pointer.
>temp_107|source:-*- coding: utf-8 -*-|start:1069|stop:1569
r x in range(0, len(ticks) - 1)])def close(self):self.output_file_obj.close()cla
ss ReadFasta:def __init__(self, f_name):self.ids = []self.sequences = []self.fas
ta = SequenceSource(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0
 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory
: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.se
quences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.c
lose()class Sequence
>temp_108|source:-*- coding: utf-8 -*-|start:5357|stop:5857
 = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (
max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight 
= 'black', size = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams
.update({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.t
ext(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(s
equence_lengths),numpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min(
sequence_lengths),\m
>temp_109|source:-*- coding: utf-8 -*-|start:60|stop:560
 is free software; you can redistribute it and/or modify it under# the terms of 
the GNU General Public License as published by the Free# Software Foundation; ei
ther version 2 of the License, or (at your option)# any later version.## Please 
read the docs/COPYING file.import sysimport numpyimport hashlibclass FastaOutput
:def __init__(self, output_file_path):self.output_file_path = output_file_pathse
lf.output_file_obj = open(output_file_path, 'w')def store(self, entry, split = T
rue, store_frequenci
>temp_110|source:-*- coding: utf-8 -*-|start:2570|stop:3070
unique_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_dict], rever
se = True)]self.total_unique = len(self.unique_hash_dict)self.reset()def next(se
lf):if self.unique:return self.next_unique()else:return self.next_regular()def n
ext_unique(self):if self.unique:if self.total_unique > 0 and self.pos < self.tot
al_unique:hash_entry = self.unique_hash_dict[self.unique_hash_list[self.pos]]sel
f.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_case else hash_entry[
'seq'].upper()self.i
>temp_111|source:-*- coding: utf-8 -*-|start:404|stop:904
ut_file_path):self.output_file_path = output_file_pathself.output_file_obj = ope
n(output_file_path, 'w')def store(self, entry, split = True, store_frequencies =
 True):if entry.unique and store_frequencies:self.write_id('%s|%s' % (entry.id, 
'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_seq(entr
y.seq, split)def write_id(self, id):self.output_file_obj.write('>%s\n' % id)def 
write_seq(self, seq, split = True):if split:seq = self.split(seq)self.output_fil
e_obj.write('%s\n' %
>temp_112|source:-*- coding: utf-8 -*-|start:6211|stop:6711
itself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []
self.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if self.l
azy_init:self.total_quals = Noneelse:self.total_quals = len([l for l in self.fil
e_pointer.readlines() if l.startswith('>')])self.reset()def next(self):self.id =
 self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_int = Noneq
ualscores = ''while 1:line = self.file_pointer.readline()if not line:if len(qual
scores):self.quals =
>temp_113|source:-*- coding: utf-8 -*-|start:2037|stop:2537
lf.total_seq = Noneelse:self.total_seq = len([l for l in self.file_pointer.readl
ines() if l.startswith('>')])self.reset()if self.unique:self.init_unique_hash()d
ef init_unique_hash(self):while self.next_regular():hash = hashlib.sha1(self.seq
.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_hash_dict[hash
]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] += 1else:self.uniqu
e_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'count': 1}
self.unique_hash_lis
>temp_114|source:-*- coding: utf-8 -*-|start:1716|stop:2216
.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_casese
lf.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.un
ique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0self.file_
pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy_init:s
elf.total_seq = Noneelse:self.total_seq = len([l for l in self.file_pointer.read
lines() if l.startswith('>')])self.reset()if self.unique:self.init_unique_hash()
def init_unique_hash
>temp_115|source:-*- coding: utf-8 -*-|start:178|stop:678
ished by the Free# Software Foundation; either version 2 of the License, or (at 
your option)# any later version.## Please read the docs/COPYING file.import sysi
mport numpyimport hashlibclass FastaOutput:def __init__(self, output_file_path):
self.output_file_path = output_file_pathself.output_file_obj = open(output_file_
path, 'w')def store(self, entry, split = True, store_frequencies = True):if entr
y.unique and store_frequencies:self.write_id('%s|%s' % (entry.id, 'frequency:%d'
 % len(entry.ids)))e
>temp_116|source:-*- coding: utf-8 -*-|start:2939|stop:3439
.unique_hash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.al
low_mixed_case else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids
 = hash_entry['ids']return Trueelse:return Falseelse:return Falsedef next_regula
r(self):self.seq = Noneself.id = self.file_pointer.readline()[1:].strip()sequenc
e = ''while 1:line = self.file_pointer.readline()if not line:if len(sequence):se
lf.seq = sequenceself.pos += 1return Trueelse:return Falseif line.startswith('>'
):self.file_pointer.
>temp_117|source:-*- coding: utf-8 -*-|start:6292|stop:6792
elf.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if self.la
zy_init:self.total_quals = Noneelse:self.total_quals = len([l for l in self.file
_pointer.readlines() if l.startswith('>')])self.reset()def next(self):self.id = 
self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_int = Nonequ
alscores = ''while 1:line = self.file_pointer.readline()if not line:if len(quals
cores):self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.qua
ls.split()]self.pos 
>temp_118|source:-*- coding: utf-8 -*-|start:1211|stop:1711
ences = []self.fasta = SequenceSource(f_name)while self.fasta.next():if self.fas
ta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading
 FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(sel
f.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(
self):self.fasta.close()class SequenceSource():def __init__(self, fasta_file_pat
h, lazy_init = True, unique = False, allow_mixed_case = False):self.fasta_file_p
ath = fasta_file_pat
>temp_119|source:-*- coding: utf-8 -*-|start:4751|stop:5251
, right = 0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.f
ill_between(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'bl
ack', alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence length'
)if xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:yt
ickstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq
_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, max(seq_l
en_distribution) + 1
>temp_120|source:-*- coding: utf-8 -*-|start:2611|stop:3111
r hash in self.unique_hash_dict], reverse = True)]self.total_unique = len(self.u
nique_hash_dict)self.reset()def next(self):if self.unique:return self.next_uniqu
e()else:return self.next_regular()def next_unique(self):if self.unique:if self.t
otal_unique > 0 and self.pos < self.total_unique:hash_entry = self.unique_hash_d
ict[self.unique_hash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if
 self.allow_mixed_case else hash_entry['seq'].upper()self.id  = hash_entry['id']
self.ids = hash_entr
>temp_121|source:-*- coding: utf-8 -*-|start:6523|stop:7023
elf.id = self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_int
 = Nonequalscores = ''while 1:line = self.file_pointer.readline()if not line:if 
len(qualscores):self.quals = qualscores.strip()self.quals_int = [int(q) for q in
 self.quals.split()]self.pos += 1return Trueelse:return Falseif line.startswith(
'>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))breakqualscores
 += ' ' + line.strip()self.quals = qualscores.strip()self.quals_int = [int(q) fo
r q in self.quals.sp
>temp_122|source:-*- coding: utf-8 -*-|start:492|stop:992
_file_path, 'w')def store(self, entry, split = True, store_frequencies = True):i
f entry.unique and store_frequencies:self.write_id('%s|%s' % (entry.id, 'frequen
cy:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_seq(entry.seq, s
plit)def write_id(self, id):self.output_file_obj.write('>%s\n' % id)def write_se
q(self, seq, split = True):if split:seq = self.split(seq)self.output_file_obj.wr
ite('%s\n' % seq)def split(self, sequence, piece_length = 80):ticks = range(0, l
en(sequence), piece_
>temp_123|source:-*- coding: utf-8 -*-|start:2287|stop:2787
hexdigest()if hash in self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].a
ppend(self.id)self.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_dic
t[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'count': 1}self.uniqu
e_hash_list = [i[1] for i in sorted([(self.unique_hash_dict[hash]['count'], hash
)\for hash in self.unique_hash_dict], reverse = True)]self.total_unique = len(se
lf.unique_hash_dict)self.reset()def next(self):if self.unique:return self.next_u
nique()else:return s
>temp_124|source:-*- coding: utf-8 -*-|start:3540|stop:4040
low_mixed_case else sequence.upper()self.pos += 1return Truedef get_seq_by_read_
id(self, read_id):self.reset()while self.next():if self.id == read_id:return sel
f.seqreturn Falsedef close(self):self.file_pointer.close()def reset(self):self.p
os = 0self.id  = Noneself.seq = Noneself.ids = []self.file_pointer.seek(0)def vi
sualize_sequence_length_distribution(self, title, dest = None, max_seq_len = Non
e, xtickstep = None, ytickstep = None):import matplotlib.pyplot as pltimport mat
plotlib.gridspec as 
>temp_125|source:-*- coding: utf-8 -*-|start:5977|stop:6477
avefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passr
eturnclass QualSource:def __init__(self, quals_file_path, lazy_init = True):self
.quals_file_path = quals_file_pathself.name = Noneself.lazy_init = lazy_initself
.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.f
ile_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if self.lazy_in
it:self.total_quals = Noneelse:self.total_quals = len([l for l in self.file_poin
ter.readlines() if l
>temp_126|source:-*- coding: utf-8 -*-|start:4044|stop:4544
specsequence_lengths = []self.reset()while self.next():if self.pos % 10000 == 0 
or self.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.st
derr.flush()sequence_lengths.append(len(self.seq))self.reset()sys.stderr.write('
\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_l
engths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in se
quence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))p
lt.rcParams.update({
>temp_127|source:-*- coding: utf-8 -*-|start:3849|stop:4349
ek(0)def visualize_sequence_length_distribution(self, title, dest = None, max_se
q_len = None, xtickstep = None, ytickstep = None):import matplotlib.pyplot as pl
timport matplotlib.gridspec as gridspecsequence_lengths = []self.reset()while se
lf.next():if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fastali
b] Reading: %s' % (self.pos))sys.stderr.flush()sequence_lengths.append(len(self.
seq))self.reset()sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(sequ
ence_lengths) + (int
>temp_128|source:-*- coding: utf-8 -*-|start:5560|stop:6060
)ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(Fa
lse)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / st
d: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence_length
s), numpy.std(sequence_lengths),\min(sequence_lengths),\max(sequence_lengths)),\
va = 'center', alpha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_f
ile_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.s
how()except:passretu
>temp_129|source:-*- coding: utf-8 -*-|start:5185|stop:5685
 size='xx-small')plt.yticks(range(0, max(seq_len_distribution) + 1, ytickstep),[
y for y in range(0, max(seq_len_distribution) + 1, ytickstep)],size='xx-small')p
lt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distr
ibution) + (max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (tit
le), weight = 'black', size = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])
plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])plt.xti
cks([])plt.text(0.02
>temp_130|source:-*- coding: utf-8 -*-|start:896|stop:1396
'%s\n' % seq)def split(self, sequence, piece_length = 80):ticks = range(0, len(s
equence), piece_length) + [len(sequence)]return '\n'.join([sequence[ticks[x]:tic
ks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.output_file_o
bj.close()class ReadFasta:def __init__(self, f_name):self.ids = []self.sequences
 = []self.fasta = SequenceSource(f_name)while self.fasta.next():if self.fasta.po
s % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FAST
A into memory: %s' %
>temp_131|source:-*- coding: utf-8 -*-|start:1102|stop:1602
def close(self):self.output_file_obj.close()class ReadFasta:def __init__(self, f
_name):self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)while 
self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stder
r.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))sys.stde
rr.flush()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.seq)sys
.stderr.write('\n')def close(self):self.fasta.close()class SequenceSource():def 
__init__(self, fasta
>temp_132|source:-*- coding: utf-8 -*-|start:6026|stop:6526
png')try:plt.show()except:passreturnclass QualSource:def __init__(self, quals_fi
le_path, lazy_init = True):self.quals_file_path = quals_file_pathself.name = Non
eself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.qual
s_int = Noneself.ids = []self.file_pointer = open(self.quals_file_path)self.file
_pointer.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals =
 len([l for l in self.file_pointer.readlines() if l.startswith('>')])self.reset(
)def next(self):self
>temp_133|source:-*- coding: utf-8 -*-|start:179|stop:679
shed by the Free# Software Foundation; either version 2 of the License, or (at y
our option)# any later version.## Please read the docs/COPYING file.import sysim
port numpyimport hashlibclass FastaOutput:def __init__(self, output_file_path):s
elf.output_file_path = output_file_pathself.output_file_obj = open(output_file_p
ath, 'w')def store(self, entry, split = True, store_frequencies = True):if entry
.unique and store_frequencies:self.write_id('%s|%s' % (entry.id, 'frequency:%d' 
% len(entry.ids)))el
>temp_134|source:-*- coding: utf-8 -*-|start:1091|stop:1591
cks) - 1)])def close(self):self.output_file_obj.close()class ReadFasta:def __ini
t__(self, f_name):self.ids = []self.sequences = []self.fasta = SequenceSource(f_
name)while self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 
1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.po
s))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(self.fa
sta.seq)sys.stderr.write('\n')def close(self):self.fasta.close()class SequenceSo
urce():def __init__(
>temp_135|source:-*- coding: utf-8 -*-|start:1048|stop:1548
s[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.outpu
t_file_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = []self.s
equences = []self.fasta = SequenceSource(f_name)while self.fasta.next():if self.
fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Read
ing FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(
self.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def clo
se(self):self.fasta.
>temp_136|source:-*- coding: utf-8 -*-|start:5376|stop:5876
q_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_distrib
ution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = '
xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edge
color' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'tot
al: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths),nu
mpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_lengths),\
max(sequence_lengths
>temp_137|source:-*- coding: utf-8 -*-|start:6658|stop:7158
readline()if not line:if len(qualscores):self.quals = qualscores.strip()self.qua
ls_int = [int(q) for q in self.quals.split()]self.pos += 1return Trueelse:return
 Falseif line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - 
len(line))breakqualscores += ' ' + line.strip()self.quals = qualscores.strip()se
lf.quals_int = [int(q) for q in self.quals.split()]self.pos += 1return Truedef c
lose(self):self.file_pointer.close()def reset(self):self.pos = 0self.id  = Nones
elf.quals = Noneself
>temp_138|source:-*- coding: utf-8 -*-|start:720|stop:1220
entry.seq, split)def write_id(self, id):self.output_file_obj.write('>%s\n' % id)
def write_seq(self, seq, split = True):if split:seq = self.split(seq)self.output
_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length = 80):ticks 
= range(0, len(sequence), piece_length) + [len(sequence)]return '\n'.join([seque
nce[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):se
lf.output_file_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = 
[]self.sequences = [
>temp_139|source:-*- coding: utf-8 -*-|start:3913|stop:4413
t = None, max_seq_len = None, xtickstep = None, ytickstep = None):import matplot
lib.pyplot as pltimport matplotlib.gridspec as gridspecsequence_lengths = []self
.reset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.w
rite('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_lengths.
append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_seq_len:max_se
q_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or 10)seq_le
n_distribution = [0]
>temp_140|source:-*- coding: utf-8 -*-|start:5714|stop:6214
f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence_
lengths), numpy.std(sequence_lengths),\min(sequence_lengths),\max(sequence_lengt
hs)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == None:dest = self.f
asta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try
:plt.show()except:passreturnclass QualSource:def __init__(self, quals_file_path,
 lazy_init = True):self.quals_file_path = quals_file_pathself.name = Noneself.la
zy_init = lazy_inits
>temp_141|source:-*- coding: utf-8 -*-|start:3043|stop:3543
_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_entry['ids']retu
rn Trueelse:return Falseelse:return Falsedef next_regular(self):self.seq = Nones
elf.id = self.file_pointer.readline()[1:].strip()sequence = ''while 1:line = sel
f.file_pointer.readline()if not line:if len(sequence):self.seq = sequenceself.po
s += 1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.seek
(self.file_pointer.tell() - len(line))breaksequence += line.strip()self.seq = se
quence if self.allow
>temp_142|source:-*- coding: utf-8 -*-|start:5672|stop:6172
plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (
len(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(sequence_lengths),
\min(sequence_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, size
 = 'x-large')if dest == None:dest = self.fasta_file_pathtry:plt.savefig(dest + '
.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passreturnclass Qual
Source:def __init__(self, quals_file_path, lazy_init = True):self.quals_file_pat
h = quals_file_paths
>temp_143|source:-*- coding: utf-8 -*-|start:4999|stop:5499
kstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:ytickstep 
= max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1
, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, max(seq_len_distr
ibution) + 1, ytickstep),[y for y in range(0, max(seq_len_distribution) + 1, yti
ckstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0
, ymax = max(seq_len_distribution) + (max(seq_len_distribution) / 20.0))plt.figt
ext(0.5, 0.96, '%s' 
>temp_144|source:-*- coding: utf-8 -*-|start:550|stop:1050
_frequencies = True):if entry.unique and store_frequencies:self.write_id('%s|%s'
 % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)self.
write_seq(entry.seq, split)def write_id(self, id):self.output_file_obj.write('>%
s\n' % id)def write_seq(self, seq, split = True):if split:seq = self.split(seq)s
elf.output_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length = 
80):ticks = range(0, len(sequence), piece_length) + [len(sequence)]return '\n'.j
oin([sequence[ticks[
>temp_145|source:-*- coding: utf-8 -*-|start:6519|stop:7019
f):self.id = self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals
_int = Nonequalscores = ''while 1:line = self.file_pointer.readline()if not line
:if len(qualscores):self.quals = qualscores.strip()self.quals_int = [int(q) for 
q in self.quals.split()]self.pos += 1return Trueelse:return Falseif line.startsw
ith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))breakqualsc
ores += ' ' + line.strip()self.quals = qualscores.strip()self.quals_int = [int(q
) for q in self.qual
>temp_146|source:-*- coding: utf-8 -*-|start:6156|stop:6656
quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = 
Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer = open(
self.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_quals
 = Noneelse:self.total_quals = len([l for l in self.file_pointer.readlines() if 
l.startswith('>')])self.reset()def next(self):self.id = self.file_pointer.readli
ne()[1:].strip()self.quals = Noneself.quals_int = Nonequalscores = ''while 1:lin
e = self.file_pointe
>temp_147|source:-*- coding: utf-8 -*-|start:6387|stop:6887
tal_quals = Noneelse:self.total_quals = len([l for l in self.file_pointer.readli
nes() if l.startswith('>')])self.reset()def next(self):self.id = self.file_point
er.readline()[1:].strip()self.quals = Noneself.quals_int = Nonequalscores = ''wh
ile 1:line = self.file_pointer.readline()if not line:if len(qualscores):self.qua
ls = qualscores.strip()self.quals_int = [int(q) for q in self.quals.split()]self
.pos += 1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.s
eek(self.file_pointe
>temp_148|source:-*- coding: utf-8 -*-|start:6350|stop:6850
nter.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = len
([l for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()def
 next(self):self.id = self.file_pointer.readline()[1:].strip()self.quals = Nones
elf.quals_int = Nonequalscores = ''while 1:line = self.file_pointer.readline()if
 not line:if len(qualscores):self.quals = qualscores.strip()self.quals_int = [in
t(q) for q in self.quals.split()]self.pos += 1return Trueelse:return Falseif lin
e.startswith('>'):se
>temp_149|source:-*- coding: utf-8 -*-|start:6099|stop:6599
uals_file_path, lazy_init = True):self.quals_file_path = quals_file_pathself.nam
e = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Nonese
lf.quals_int = Noneself.ids = []self.file_pointer = open(self.quals_file_path)se
lf.file_pointer.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_
quals = len([l for l in self.file_pointer.readlines() if l.startswith('>')])self
.reset()def next(self):self.id = self.file_pointer.readline()[1:].strip()self.qu
als = Noneself.quals
>temp_150|source:-*- coding: utf-8 -*-|start:4070|stop:4570
elf.reset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.stder
r.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_lengt
hs.append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_seq_len:max
_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or 10)seq
_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_dis
tribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update({'axes.
linewidth' : 0.9})pl
>temp_151|source:-*- coding: utf-8 -*-|start:544|stop:1044
 store_frequencies = True):if entry.unique and store_frequencies:self.write_id('
%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id
)self.write_seq(entry.seq, split)def write_id(self, id):self.output_file_obj.wri
te('>%s\n' % id)def write_seq(self, seq, split = True):if split:seq = self.split
(seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequence, piece_len
gth = 80):ticks = range(0, len(sequence), piece_length) + [len(sequence)]return 
'\n'.join([sequence[
>temp_152|source:-*- coding: utf-8 -*-|start:773|stop:1273
ile_obj.write('>%s\n' % id)def write_seq(self, seq, split = True):if split:seq =
 self.split(seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequence
, piece_length = 80):ticks = range(0, len(sequence), piece_length) + [len(sequen
ce)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(tick
s) - 1)])def close(self):self.output_file_obj.close()class ReadFasta:def __init_
_(self, f_name):self.ids = []self.sequences = []self.fasta = SequenceSource(f_na
me)while self.fasta.
>temp_153|source:-*- coding: utf-8 -*-|start:4184|stop:4684
 % (self.pos))sys.stderr.flush()sequence_lengths.append(len(self.seq))self.reset
()sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) +
 (int(max(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq
_len + 1)for l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(
figsize = (16, 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', c
olor='0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt
.subplot(gs[0:8])plt
>temp_154|source:-*- coding: utf-8 -*-|start:189|stop:689
e Free# Software Foundation; either version 2 of the License, or (at your option
)# any later version.## Please read the docs/COPYING file.import sysimport numpy
import hashlibclass FastaOutput:def __init__(self, output_file_path):self.output
_file_path = output_file_pathself.output_file_obj = open(output_file_path, 'w')d
ef store(self, entry, split = True, store_frequencies = True):if entry.unique an
d store_frequencies:self.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entr
y.ids)))else:self.wr
>temp_155|source:-*- coding: utf-8 -*-|start:6586|stop:7086
oneself.quals_int = Nonequalscores = ''while 1:line = self.file_pointer.readline
()if not line:if len(qualscores):self.quals = qualscores.strip()self.quals_int =
 [int(q) for q in self.quals.split()]self.pos += 1return Trueelse:return Falseif
 line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line
))breakqualscores += ' ' + line.strip()self.quals = qualscores.strip()self.quals
_int = [int(q) for q in self.quals.split()]self.pos += 1return Truedef close(sel
f):self.file_pointer
>temp_156|source:-*- coding: utf-8 -*-|start:4386|stop:4886
)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len
_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update({'a
xes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewidth=0.1)
gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplot
s_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_dis
tribution, color = 'black', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 
1), seq_len_distribu
>temp_157|source:-*- coding: utf-8 -*-|start:3523|stop:4023
quence if self.allow_mixed_case else sequence.upper()self.pos += 1return Truedef
 get_seq_by_read_id(self, read_id):self.reset()while self.next():if self.id == r
ead_id:return self.seqreturn Falsedef close(self):self.file_pointer.close()def r
eset(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.file_poin
ter.seek(0)def visualize_sequence_length_distribution(self, title, dest = None, 
max_seq_len = None, xtickstep = None, ytickstep = None):import matplotlib.pyplot
 as pltimport matplo
>temp_158|source:-*- coding: utf-8 -*-|start:1826|stop:2326
seq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique
_hash_list = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_file
_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.
total_seq = len([l for l in self.file_pointer.readlines() if l.startswith('>')])
self.reset()if self.unique:self.init_unique_hash()def init_unique_hash(self):whi
le self.next_regular():hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash 
in self.unique_hash_
>temp_159|source:-*- coding: utf-8 -*-|start:5618|stop:6118
lor' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total
: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths),nump
y.mean(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_lengths),\ma
x(sequence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == No
ne:dest = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(d
est + '.png')try:plt.show()except:passreturnclass QualSource:def __init__(self, 
quals_file_path, laz
>temp_160|source:-*- coding: utf-8 -*-|start:5436|stop:5936
(max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight
 = 'black', size = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParam
s.update({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.
text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(
sequence_lengths),numpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min
(sequence_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, size = '
x-large')if dest == 
>temp_161|source:-*- coding: utf-8 -*-|start:3343|stop:3843
eq = sequenceself.pos += 1return Trueelse:return Falseif line.startswith('>'):se
lf.file_pointer.seek(self.file_pointer.tell() - len(line))breaksequence += line.
strip()self.seq = sequence if self.allow_mixed_case else sequence.upper()self.po
s += 1return Truedef get_seq_by_read_id(self, read_id):self.reset()while self.ne
xt():if self.id == read_id:return self.seqreturn Falsedef close(self):self.file_
pointer.close()def reset(self):self.pos = 0self.id  = Noneself.seq = Noneself.id
s = []self.file_poin
>temp_162|source:-*- coding: utf-8 -*-|start:554|stop:1054
quencies = True):if entry.unique and store_frequencies:self.write_id('%s|%s' % (
entry.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)self.writ
e_seq(entry.seq, split)def write_id(self, id):self.output_file_obj.write('>%s\n'
 % id)def write_seq(self, seq, split = True):if split:seq = self.split(seq)self.
output_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length = 80):
ticks = range(0, len(sequence), piece_length) + [len(sequence)]return '\n'.join(
[sequence[ticks[x]:t
>temp_163|source:-*- coding: utf-8 -*-|start:970|stop:1470
 len(sequence), piece_length) + [len(sequence)]return '\n'.join([sequence[ticks[
x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.output_
file_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = []self.seq
uences = []self.fasta = SequenceSource(f_name)while self.fasta.next():if self.fa
sta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Readin
g FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(se
lf.fasta.id)self.seq
>temp_164|source:-*- coding: utf-8 -*-|start:2|stop:502
v.140216# Copyright (C) 2014, A. Murat Eren## This program is free software; you
 can redistribute it and/or modify it under# the terms of the GNU General Public
 License as published by the Free# Software Foundation; either version 2 of the 
License, or (at your option)# any later version.## Please read the docs/COPYING 
file.import sysimport numpyimport hashlibclass FastaOutput:def __init__(self, ou
tput_file_path):self.output_file_path = output_file_pathself.output_file_obj = o
pen(output_file_path
>temp_165|source:-*- coding: utf-8 -*-|start:2772|stop:3272
()else:return self.next_regular()def next_unique(self):if self.unique:if self.to
tal_unique > 0 and self.pos < self.total_unique:hash_entry = self.unique_hash_di
ct[self.unique_hash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if 
self.allow_mixed_case else hash_entry['seq'].upper()self.id  = hash_entry['id']s
elf.ids = hash_entry['ids']return Trueelse:return Falseelse:return Falsedef next
_regular(self):self.seq = Noneself.id = self.file_pointer.readline()[1:].strip()
sequence = ''while 1
>temp_166|source:-*- coding: utf-8 -*-|start:2811|stop:3311
xt_unique(self):if self.unique:if self.total_unique > 0 and self.pos < self.tota
l_unique:hash_entry = self.unique_hash_dict[self.unique_hash_list[self.pos]]self
.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_case else hash_entry['
seq'].upper()self.id  = hash_entry['id']self.ids = hash_entry['ids']return Truee
lse:return Falseelse:return Falsedef next_regular(self):self.seq = Noneself.id =
 self.file_pointer.readline()[1:].strip()sequence = ''while 1:line = self.file_p
ointer.readline()if 
>temp_167|source:-*- coding: utf-8 -*-|start:1058|stop:1558
[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.output_file_obj
.close()class ReadFasta:def __init__(self, f_name):self.ids = []self.sequences =
 []self.fasta = SequenceSource(f_name)while self.fasta.next():if self.fasta.pos 
% 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA 
into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta
.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):s
elf.fasta.close()cla
>temp_168|source:-*- coding: utf-8 -*-|start:6840|stop:7340
th('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))breakqualsco
res += ' ' + line.strip()self.quals = qualscores.strip()self.quals_int = [int(q)
 for q in self.quals.split()]self.pos += 1return Truedef close(self):self.file_p
ointer.close()def reset(self):self.pos = 0self.id  = Noneself.quals = Noneself.q
uals_int = Noneself.ids = []self.file_pointer.seek(0)if __name__ == '__main__':f
asta = SequenceSource(sys.argv[1])fasta.visualize_sequence_length_distribution(t
itle = sys.argv[2] i
>temp_169|source:-*- coding: utf-8 -*-|start:5118|stop:5618
t.xticks(range(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-sma
ll')plt.yticks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y in ra
nge(0, max(seq_len_distribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin 
= 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (m
ax(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight =
 'black', size = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.
update({'axes.edgeco
>temp_170|source:-*- coding: utf-8 -*-|start:1303|stop:1803
0 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into 
memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)s
elf.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.f
asta.close()class SequenceSource():def __init__(self, fasta_file_path, lazy_init
 = True, unique = False, allow_mixed_case = False):self.fasta_file_path = fasta_
file_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow
_mixed_caseself.pos 
>temp_171|source:-*- coding: utf-8 -*-|start:3153|stop:3653
eturn Falsedef next_regular(self):self.seq = Noneself.id = self.file_pointer.rea
dline()[1:].strip()sequence = ''while 1:line = self.file_pointer.readline()if no
t line:if len(sequence):self.seq = sequenceself.pos += 1return Trueelse:return F
alseif line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - le
n(line))breaksequence += line.strip()self.seq = sequence if self.allow_mixed_cas
e else sequence.upper()self.pos += 1return Truedef get_seq_by_read_id(self, read
_id):self.reset()whi
>temp_172|source:-*- coding: utf-8 -*-|start:1984|stop:2484
le_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:sel
f.total_seq = len([l for l in self.file_pointer.readlines() if l.startswith('>')
])self.reset()if self.unique:self.init_unique_hash()def init_unique_hash(self):w
hile self.next_regular():hash = hashlib.sha1(self.seq.upper()).hexdigest()if has
h in self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].append(self.id)sel
f.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' :
 self.id,'ids': [sel
>temp_173|source:-*- coding: utf-8 -*-|start:6666|stop:7166
()if not line:if len(qualscores):self.quals = qualscores.strip()self.quals_int =
 [int(q) for q in self.quals.split()]self.pos += 1return Trueelse:return Falseif
 line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line
))breakqualscores += ' ' + line.strip()self.quals = qualscores.strip()self.quals
_int = [int(q) for q in self.quals.split()]self.pos += 1return Truedef close(sel
f):self.file_pointer.close()def reset(self):self.pos = 0self.id  = Noneself.qual
s = Noneself.quals_i
>temp_174|source:-*- coding: utf-8 -*-|start:6071|stop:6571
lSource:def __init__(self, quals_file_path, lazy_init = True):self.quals_file_pa
th = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.i
d  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer = 
open(self.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_
quals = Noneelse:self.total_quals = len([l for l in self.file_pointer.readlines(
) if l.startswith('>')])self.reset()def next(self):self.id = self.file_pointer.r
eadline()[1:].strip(
>temp_175|source:-*- coding: utf-8 -*-|start:1794|stop:2294
self.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.
unique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0self.fil
e_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy_init
:self.total_seq = Noneelse:self.total_seq = len([l for l in self.file_pointer.re
adlines() if l.startswith('>')])self.reset()if self.unique:self.init_unique_hash
()def init_unique_hash(self):while self.next_regular():hash = hashlib.sha1(self.
seq.upper()).hexdige
>temp_176|source:-*- coding: utf-8 -*-|start:2933|stop:3433
t[self.unique_hash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if s
elf.allow_mixed_case else hash_entry['seq'].upper()self.id  = hash_entry['id']se
lf.ids = hash_entry['ids']return Trueelse:return Falseelse:return Falsedef next_
regular(self):self.seq = Noneself.id = self.file_pointer.readline()[1:].strip()s
equence = ''while 1:line = self.file_pointer.readline()if not line:if len(sequen
ce):self.seq = sequenceself.pos += 1return Trueelse:return Falseif line.startswi
th('>'):self.file_po
>temp_177|source:-*- coding: utf-8 -*-|start:1515|stop:2015
('\n')def close(self):self.fasta.close()class SequenceSource():def __init__(self
, fasta_file_path, lazy_init = True, unique = False, allow_mixed_case = False):s
elf.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init = lazy_inits
elf.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Non
eself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_hash_lis
t = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_file_path)sel
f.file_pointer.seek(
>temp_178|source:-*- coding: utf-8 -*-|start:3029|stop:3529
case else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_en
try['ids']return Trueelse:return Falseelse:return Falsedef next_regular(self):se
lf.seq = Noneself.id = self.file_pointer.readline()[1:].strip()sequence = ''whil
e 1:line = self.file_pointer.readline()if not line:if len(sequence):self.seq = s
equenceself.pos += 1return Trueelse:return Falseif line.startswith('>'):self.fil
e_pointer.seek(self.file_pointer.tell() - len(line))breaksequence += line.strip(
)self.seq = sequence
>temp_179|source:-*- coding: utf-8 -*-|start:1300|stop:1800
1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA in
to memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.i
d)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):sel
f.fasta.close()class SequenceSource():def __init__(self, fasta_file_path, lazy_i
nit = True, unique = False, allow_mixed_case = False):self.fasta_file_path = fas
ta_file_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = al
low_mixed_caseself.p
>temp_180|source:-*- coding: utf-8 -*-|start:3666|stop:4166
):if self.id == read_id:return self.seqreturn Falsedef close(self):self.file_poi
nter.close()def reset(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids =
 []self.file_pointer.seek(0)def visualize_sequence_length_distribution(self, tit
le, dest = None, max_seq_len = None, xtickstep = None, ytickstep = None):import 
matplotlib.pyplot as pltimport matplotlib.gridspec as gridspecsequence_lengths =
 []self.reset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.s
tderr.write('\r[fast
>temp_181|source:-*- coding: utf-8 -*-|start:6549|stop:7049
.readline()[1:].strip()self.quals = Noneself.quals_int = Nonequalscores = ''whil
e 1:line = self.file_pointer.readline()if not line:if len(qualscores):self.quals
 = qualscores.strip()self.quals_int = [int(q) for q in self.quals.split()]self.p
os += 1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.see
k(self.file_pointer.tell() - len(line))breakqualscores += ' ' + line.strip()self
.quals = qualscores.strip()self.quals_int = [int(q) for q in self.quals.split()]
self.pos += 1return 
>temp_182|source:-*- coding: utf-8 -*-|start:1852|stop:2352
f.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list = []self.unique
_next_hash = 0self.file_pointer = open(self.fasta_file_path)self.file_pointer.se
ek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for l i
n self.file_pointer.readlines() if l.startswith('>')])self.reset()if self.unique
:self.init_unique_hash()def init_unique_hash(self):while self.next_regular():has
h = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:s
elf.unique_hash_dict
>temp_183|source:-*- coding: utf-8 -*-|start:6292|stop:6792
elf.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if self.la
zy_init:self.total_quals = Noneelse:self.total_quals = len([l for l in self.file
_pointer.readlines() if l.startswith('>')])self.reset()def next(self):self.id = 
self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_int = Nonequ
alscores = ''while 1:line = self.file_pointer.readline()if not line:if len(quals
cores):self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.qua
ls.split()]self.pos 
>temp_184|source:-*- coding: utf-8 -*-|start:5401|stop:5901
ymax = max(seq_len_distribution) + (max(seq_len_distribution) / 20.0))plt.figtex
t(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-large', ha = 'center')
ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(Fal
se)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std
: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence_lengths
), numpy.std(sequence_lengths),\min(sequence_lengths),\max(sequence_lengths)),\v
a = 'center', alpha 
>temp_185|source:-*- coding: utf-8 -*-|start:3252|stop:3752
sequence = ''while 1:line = self.file_pointer.readline()if not line:if len(seque
nce):self.seq = sequenceself.pos += 1return Trueelse:return Falseif line.startsw
ith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))breaksequen
ce += line.strip()self.seq = sequence if self.allow_mixed_case else sequence.upp
er()self.pos += 1return Truedef get_seq_by_read_id(self, read_id):self.reset()wh
ile self.next():if self.id == read_id:return self.seqreturn Falsedef close(self)
:self.file_pointer.c
>temp_186|source:-*- coding: utf-8 -*-|start:2212|stop:2712
hash(self):while self.next_regular():hash = hashlib.sha1(self.seq.upper()).hexdi
gest()if hash in self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].append
(self.id)self.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_dict[has
h] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'count': 1}self.unique_has
h_list = [i[1] for i in sorted([(self.unique_hash_dict[hash]['count'], hash)\for
 hash in self.unique_hash_dict], reverse = True)]self.total_unique = len(self.un
ique_hash_dict)self.
>temp_187|source:-*- coding: utf-8 -*-|start:1084|stop:1584
 len(ticks) - 1)])def close(self):self.output_file_obj.close()class ReadFasta:de
f __init__(self, f_name):self.ids = []self.sequences = []self.fasta = SequenceSo
urce(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.
pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.f
asta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(
self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close()class Seq
uenceSource():def __
>temp_188|source:-*- coding: utf-8 -*-|start:5564|stop:6064
 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)
plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %
.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence_lengths), 
numpy.std(sequence_lengths),\min(sequence_lengths),\max(sequence_lengths)),\va =
 'center', alpha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_
pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show(
)except:passreturncl
>temp_189|source:-*- coding: utf-8 -*-|start:5863|stop:6363
uence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == None:de
st = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest +
 '.png')try:plt.show()except:passreturnclass QualSource:def __init__(self, quals
_file_path, lazy_init = True):self.quals_file_path = quals_file_pathself.name = 
Noneself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.q
uals_int = Noneself.ids = []self.file_pointer = open(self.quals_file_path)self.f
ile_pointer.seek(0)i
>temp_190|source:-*- coding: utf-8 -*-|start:1051|stop:1551
]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.output_f
ile_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = []self.sequ
ences = []self.fasta = SequenceSource(f_name)while self.fasta.next():if self.fas
ta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading
 FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(sel
f.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(
self):self.fasta.clo
>temp_191|source:-*- coding: utf-8 -*-|start:40|stop:540
 Eren## This program is free software; you can redistribute it and/or modify it 
under# the terms of the GNU General Public License as published by the Free# Sof
tware Foundation; either version 2 of the License, or (at your option)# any late
r version.## Please read the docs/COPYING file.import sysimport numpyimport hash
libclass FastaOutput:def __init__(self, output_file_path):self.output_file_path 
= output_file_pathself.output_file_obj = open(output_file_path, 'w')def store(se
lf, entry, split = T
>temp_192|source:-*- coding: utf-8 -*-|start:1373|stop:1873
ASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.
fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(se
lf):self.fasta.close()class SequenceSource():def __init__(self, fasta_file_path,
 lazy_init = True, unique = False, allow_mixed_case = False):self.fasta_file_pat
h = fasta_file_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_ca
se = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []self
.unique = uniqueself
>temp_193|source:-*- coding: utf-8 -*-|start:4516|stop:5016
6, 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50',
 linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[
0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, rig
ht = 0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.fill_b
etween(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black',
 alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence length')if x
tickstep == None:xti
>temp_194|source:-*- coding: utf-8 -*-|start:402|stop:902
tput_file_path):self.output_file_path = output_file_pathself.output_file_obj = o
pen(output_file_path, 'w')def store(self, entry, split = True, store_frequencies
 = True):if entry.unique and store_frequencies:self.write_id('%s|%s' % (entry.id
, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_seq(en
try.seq, split)def write_id(self, id):self.output_file_obj.write('>%s\n' % id)de
f write_seq(self, seq, split = True):if split:seq = self.split(seq)self.output_f
ile_obj.write('%s\n'
>temp_195|source:-*- coding: utf-8 -*-|start:5825|stop:6325
ngths),\min(sequence_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.
8, size = 'x-large')if dest == None:dest = self.fasta_file_pathtry:plt.savefig(d
est + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passreturncla
ss QualSource:def __init__(self, quals_file_path, lazy_init = True):self.quals_f
ile_path = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0
self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_poin
ter = open(self.qual
>temp_196|source:-*- coding: utf-8 -*-|start:2080|stop:2580
n([l for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()if
 self.unique:self.init_unique_hash()def init_unique_hash(self):while self.next_r
egular():hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique
_hash_dict:self.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_di
ct[hash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': 
[self.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sort
ed([(self.unique_has
>temp_197|source:-*- coding: utf-8 -*-|start:2509|stop:3009
unt': 1}self.unique_hash_list = [i[1] for i in sorted([(self.unique_hash_dict[ha
sh]['count'], hash)\for hash in self.unique_hash_dict], reverse = True)]self.tot
al_unique = len(self.unique_hash_dict)self.reset()def next(self):if self.unique:
return self.next_unique()else:return self.next_regular()def next_unique(self):if
 self.unique:if self.total_unique > 0 and self.pos < self.total_unique:hash_entr
y = self.unique_hash_dict[self.unique_hash_list[self.pos]]self.pos += 1self.seq 
= hash_entry['seq'] 
>temp_198|source:-*- coding: utf-8 -*-|start:281|stop:781
 version.## Please read the docs/COPYING file.import sysimport numpyimport hashl
ibclass FastaOutput:def __init__(self, output_file_path):self.output_file_path =
 output_file_pathself.output_file_obj = open(output_file_path, 'w')def store(sel
f, entry, split = True, store_frequencies = True):if entry.unique and store_freq
uencies:self.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else
:self.write_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, id):s
elf.output_file_obj.
>temp_199|source:-*- coding: utf-8 -*-|start:3945|stop:4445
ickstep = None, ytickstep = None):import matplotlib.pyplot as pltimport matplotl
ib.gridspec as gridspecsequence_lengths = []self.reset()while self.next():if sel
f.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' 
% (self.pos))sys.stderr.flush()sequence_lengths.append(len(self.seq))self.reset(
)sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + 
(int(max(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_
len + 1)for l in seq
>temp_200|source:-*- coding: utf-8 -*-|start:1602|stop:2102
_file_path, lazy_init = True, unique = False, allow_mixed_case = False):self.fas
ta_file_path = fasta_file_pathself.name = Noneself.lazy_init = lazy_initself.all
ow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.i
ds = []self.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list = []s
elf.unique_next_hash = 0self.file_pointer = open(self.fasta_file_path)self.file_
pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len(
[l for l in self.fil
>temp_201|source:-*- coding: utf-8 -*-|start:5863|stop:6363
uence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == None:de
st = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest +
 '.png')try:plt.show()except:passreturnclass QualSource:def __init__(self, quals
_file_path, lazy_init = True):self.quals_file_path = quals_file_pathself.name = 
Noneself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.q
uals_int = Noneself.ids = []self.file_pointer = open(self.quals_file_path)self.f
ile_pointer.seek(0)i
>temp_202|source:-*- coding: utf-8 -*-|start:3319|stop:3819
:if len(sequence):self.seq = sequenceself.pos += 1return Trueelse:return Falseif
 line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line
))breaksequence += line.strip()self.seq = sequence if self.allow_mixed_case else
 sequence.upper()self.pos += 1return Truedef get_seq_by_read_id(self, read_id):s
elf.reset()while self.next():if self.id == read_id:return self.seqreturn Falsede
f close(self):self.file_pointer.close()def reset(self):self.pos = 0self.id  = No
neself.seq = Nonesel
>temp_203|source:-*- coding: utf-8 -*-|start:2200|stop:2700
init_unique_hash(self):while self.next_regular():hash = hashlib.sha1(self.seq.up
per()).hexdigest()if hash in self.unique_hash_dict:self.unique_hash_dict[hash]['
ids'].append(self.id)self.unique_hash_dict[hash]['count'] += 1else:self.unique_h
ash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'count': 1}sel
f.unique_hash_list = [i[1] for i in sorted([(self.unique_hash_dict[hash]['count'
], hash)\for hash in self.unique_hash_dict], reverse = True)]self.total_unique =
 len(self.unique_has
>temp_204|source:-*- coding: utf-8 -*-|start:2082|stop:2582
[l for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()if s
elf.unique:self.init_unique_hash()def init_unique_hash(self):while self.next_reg
ular():hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_h
ash_dict:self.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict
[hash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [s
elf.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted
([(self.unique_hash_
>temp_205|source:-*- coding: utf-8 -*-|start:4289|stop:4789
f not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_length
s) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequenc
e_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rc
Params.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-
', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid
(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt
.plot(seq_len_distri
>temp_206|source:-*- coding: utf-8 -*-|start:6498|stop:6998
f.reset()def next(self):self.id = self.file_pointer.readline()[1:].strip()self.q
uals = Noneself.quals_int = Nonequalscores = ''while 1:line = self.file_pointer.
readline()if not line:if len(qualscores):self.quals = qualscores.strip()self.qua
ls_int = [int(q) for q in self.quals.split()]self.pos += 1return Trueelse:return
 Falseif line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - 
len(line))breakqualscores += ' ' + line.strip()self.quals = qualscores.strip()se
lf.quals_int = [int(
>temp_207|source:-*- coding: utf-8 -*-|start:5633|stop:6133
grid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.
2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence
_lengths), numpy.std(sequence_lengths),\min(sequence_lengths),\max(sequence_leng
ths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == None:dest = self.
fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')tr
y:plt.show()except:passreturnclass QualSource:def __init__(self, quals_file_path
, lazy_init = True):
>temp_208|source:-*- coding: utf-8 -*-|start:1380|stop:1880
to memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.i
d)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):sel
f.fasta.close()class SequenceSource():def __init__(self, fasta_file_path, lazy_i
nit = True, unique = False, allow_mixed_case = False):self.fasta_file_path = fas
ta_file_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = al
low_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.unique
 = uniqueself.unique
>temp_209|source:-*- coding: utf-8 -*-|start:3981|stop:4481
port matplotlib.pyplot as pltimport matplotlib.gridspec as gridspecsequence_leng
ths = []self.reset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:
sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()seque
nce_lengths.append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_se
q_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) 
or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:se
q_len_distribution[l
>temp_210|source:-*- coding: utf-8 -*-|start:1887|stop:2387
ict = {}self.unique_hash_list = []self.unique_next_hash = 0self.file_pointer = o
pen(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_s
eq = Noneelse:self.total_seq = len([l for l in self.file_pointer.readlines() if 
l.startswith('>')])self.reset()if self.unique:self.init_unique_hash()def init_un
ique_hash(self):while self.next_regular():hash = hashlib.sha1(self.seq.upper()).
hexdigest()if hash in self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].a
ppend(self.id)self.u
>temp_211|source:-*- coding: utf-8 -*-|start:6506|stop:7006
)def next(self):self.id = self.file_pointer.readline()[1:].strip()self.quals = N
oneself.quals_int = Nonequalscores = ''while 1:line = self.file_pointer.readline
()if not line:if len(qualscores):self.quals = qualscores.strip()self.quals_int =
 [int(q) for q in self.quals.split()]self.pos += 1return Trueelse:return Falseif
 line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line
))breakqualscores += ' ' + line.strip()self.quals = qualscores.strip()self.quals
_int = [int(q) for q
>temp_212|source:-*- coding: utf-8 -*-|start:2409|stop:2909
'count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id]
,'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self
.unique_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_dict], reve
rse = True)]self.total_unique = len(self.unique_hash_dict)self.reset()def next(s
elf):if self.unique:return self.next_unique()else:return self.next_regular()def 
next_unique(self):if self.unique:if self.total_unique > 0 and self.pos < self.to
tal_unique:hash_entr
>temp_213|source:-*- coding: utf-8 -*-|start:4671|stop:5171
t(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95
, right = 0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.f
ill_between(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'bl
ack', alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence length'
)if xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:yt
ickstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq
_len + 1, xtickstep)
>temp_214|source:-*- coding: utf-8 -*-|start:4726|stop:5226
bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_distribution, color = '
black', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distribu
tion, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequences')plt
.xlabel('sequence length')if xtickstep == None:xtickstep = (max_seq_len / 50) or
 1if ytickstep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks
(range(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.
yticks(range(0, max(
>temp_215|source:-*- coding: utf-8 -*-|start:4674|stop:5174
s[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, r
ight = 0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.fill
_between(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black
', alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence length')if
 xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:ytick
step = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_le
n + 1, xtickstep), r
>temp_216|source:-*- coding: utf-8 -*-|start:1262|stop:1762
self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stder
r.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))sys.stde
rr.flush()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.seq)sys
.stderr.write('\n')def close(self):self.fasta.close()class SequenceSource():def 
__init__(self, fasta_file_path, lazy_init = True, unique = False, allow_mixed_ca
se = False):self.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init
 = lazy_initself.all
>temp_217|source:-*- coding: utf-8 -*-|start:5015|stop:5515
ickstep = (max_seq_len / 50) or 1if ytickstep == None:ytickstep = max(seq_len_di
stribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1, xtickstep), ro
tation=90, size='xx-small')plt.yticks(range(0, max(seq_len_distribution) + 1, yt
ickstep),[y for y in range(0, max(seq_len_distribution) + 1, ytickstep)],size='x
x-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq
_len_distribution) + (max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '
%s' % (title), weigh
>temp_218|source:-*- coding: utf-8 -*-|start:5845|stop:6345
_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')
if dest == None:dest = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:
plt.savefig(dest + '.png')try:plt.show()except:passreturnclass QualSource:def __
init__(self, quals_file_path, lazy_init = True):self.quals_file_path = quals_fil
e_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.
quals = Noneself.quals_int = Noneself.ids = []self.file_pointer = open(self.qual
s_file_path)self.fil
>temp_219|source:-*- coding: utf-8 -*-|start:2788|stop:3288
lf.next_regular()def next_unique(self):if self.unique:if self.total_unique > 0 a
nd self.pos < self.total_unique:hash_entry = self.unique_hash_dict[self.unique_h
ash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed
_case else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_e
ntry['ids']return Trueelse:return Falseelse:return Falsedef next_regular(self):s
elf.seq = Noneself.id = self.file_pointer.readline()[1:].strip()sequence = ''whi
le 1:line = self.fil
>temp_220|source:-*- coding: utf-8 -*-|start:917|stop:1417
t(self, sequence, piece_length = 80):ticks = range(0, len(sequence), piece_lengt
h) + [len(sequence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in r
ange(0, len(ticks) - 1)])def close(self):self.output_file_obj.close()class ReadF
asta:def __init__(self, f_name):self.ids = []self.sequences = []self.fasta = Seq
uenceSource(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0 or self
.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % 
(self.fasta.pos))sys
>temp_221|source:-*- coding: utf-8 -*-|start:684|stop:1184
lf.write_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, id):self
.output_file_obj.write('>%s\n' % id)def write_seq(self, seq, split = True):if sp
lit:seq = self.split(seq)self.output_file_obj.write('%s\n' % seq)def split(self,
 sequence, piece_length = 80):ticks = range(0, len(sequence), piece_length) + [l
en(sequence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0,
 len(ticks) - 1)])def close(self):self.output_file_obj.close()class ReadFasta:de
f __init__(self, f_n
>temp_222|source:-*- coding: utf-8 -*-|start:5763|stop:6263
e_lengths),numpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min(sequen
ce_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large
')if dest == None:dest = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')excep
t:plt.savefig(dest + '.png')try:plt.show()except:passreturnclass QualSource:def 
__init__(self, quals_file_path, lazy_init = True):self.quals_file_path = quals_f
ile_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Nonesel
f.quals = Noneself.q
>temp_223|source:-*- coding: utf-8 -*-|start:1873|stop:2373
.unique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0self.fi
le_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy_ini
t:self.total_seq = Noneelse:self.total_seq = len([l for l in self.file_pointer.r
eadlines() if l.startswith('>')])self.reset()if self.unique:self.init_unique_has
h()def init_unique_hash(self):while self.next_regular():hash = hashlib.sha1(self
.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_hash_dict[
hash]['ids'].append(
>temp_224|source:-*- coding: utf-8 -*-|start:1866|stop:2366
queself.unique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0
self.file_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if self.l
azy_init:self.total_seq = Noneelse:self.total_seq = len([l for l in self.file_po
inter.readlines() if l.startswith('>')])self.reset()if self.unique:self.init_uni
que_hash()def init_unique_hash(self):while self.next_regular():hash = hashlib.sh
a1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_has
h_dict[hash]['ids'].
>temp_225|source:-*- coding: utf-8 -*-|start:6079|stop:6579
def __init__(self, quals_file_path, lazy_init = True):self.quals_file_path = qua
ls_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Non
eself.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer = open(sel
f.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_quals = 
Noneelse:self.total_quals = len([l for l in self.file_pointer.readlines() if l.s
tartswith('>')])self.reset()def next(self):self.id = self.file_pointer.readline(
)[1:].strip()self.qu
>temp_226|source:-*- coding: utf-8 -*-|start:4709|stop:5209
djust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_distri
bution, color = 'black', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1),
 seq_len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number 
of sequences')plt.xlabel('sequence length')if xtickstep == None:xtickstep = (max
_seq_len / 50) or 1if ytickstep == None:ytickstep = max(seq_len_distribution) / 
20 or 1plt.xticks(range(xtickstep, max_seq_len + 1, xtickstep), rotation=90, siz
e='xx-small')plt.yti
>temp_227|source:-*- coding: utf-8 -*-|start:1461|stop:1961
)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self
.fasta.close()class SequenceSource():def __init__(self, fasta_file_path, lazy_in
it = True, unique = False, allow_mixed_case = False):self.fasta_file_path = fast
a_file_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = all
ow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.unique 
= uniqueself.unique_hash_dict = {}self.unique_hash_list = []self.unique_next_has
h = 0self.file_point
>temp_228|source:-*- coding: utf-8 -*-|start:5973|stop:6473
lt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:p
assreturnclass QualSource:def __init__(self, quals_file_path, lazy_init = True):
self.quals_file_path = quals_file_pathself.name = Noneself.lazy_init = lazy_init
self.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []se
lf.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if self.laz
y_init:self.total_quals = Noneelse:self.total_quals = len([l for l in self.file_
pointer.readlines() 
>temp_229|source:-*- coding: utf-8 -*-|start:4253|stop:4753
)self.reset()sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(sequence
_lengths) + (int(max(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0]
 * (max_seq_len + 1)for l in sequence_lengths:seq_len_distribution[l] += 1fig = 
plt.figure(figsize = (16, 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.r
c('grid', color='0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 
1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom 
= 0.03, top = 0.95, 
>temp_230|source:-*- coding: utf-8 -*-|start:3650|stop:4150
while self.next():if self.id == read_id:return self.seqreturn Falsedef close(sel
f):self.file_pointer.close()def reset(self):self.pos = 0self.id  = Noneself.seq 
= Noneself.ids = []self.file_pointer.seek(0)def visualize_sequence_length_distri
bution(self, title, dest = None, max_seq_len = None, xtickstep = None, ytickstep
 = None):import matplotlib.pyplot as pltimport matplotlib.gridspec as gridspecse
quence_lengths = []self.reset()while self.next():if self.pos % 10000 == 0 or sel
f.pos == 1:sys.stder
>temp_231|source:-*- coding: utf-8 -*-|start:2429|stop:2929
lf.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'co
unt': 1}self.unique_hash_list = [i[1] for i in sorted([(self.unique_hash_dict[ha
sh]['count'], hash)\for hash in self.unique_hash_dict], reverse = True)]self.tot
al_unique = len(self.unique_hash_dict)self.reset()def next(self):if self.unique:
return self.next_unique()else:return self.next_regular()def next_unique(self):if
 self.unique:if self.total_unique > 0 and self.pos < self.total_unique:hash_entr
y = self.unique_hash
>temp_232|source:-*- coding: utf-8 -*-|start:4934|stop:5434
ylabel('number of sequences')plt.xlabel('sequence length')if xtickstep == None:x
tickstep = (max_seq_len / 50) or 1if ytickstep == None:ytickstep = max(seq_len_d
istribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1, xtickstep), r
otation=90, size='xx-small')plt.yticks(range(0, max(seq_len_distribution) + 1, y
tickstep),[y for y in range(0, max(seq_len_distribution) + 1, ytickstep)],size='
xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(se
q_len_distribution) 
>temp_233|source:-*- coding: utf-8 -*-|start:3195|stop:3695
 = Noneself.id = self.file_pointer.readline()[1:].strip()sequence = ''while 1:li
ne = self.file_pointer.readline()if not line:if len(sequence):self.seq = sequenc
eself.pos += 1return Trueelse:return Falseif line.startswith('>'):self.file_poin
ter.seek(self.file_pointer.tell() - len(line))breaksequence += line.strip()self.
seq = sequence if self.allow_mixed_case else sequence.upper()self.pos += 1return
 Truedef get_seq_by_read_id(self, read_id):self.reset()while self.next():if self
.id == read_id:retur
>temp_234|source:-*- coding: utf-8 -*-|start:1242|stop:1742
Source(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fast
a.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self
.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.appen
d(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close()class S
equenceSource():def __init__(self, fasta_file_path, lazy_init = True, unique = F
alse, allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.name 
= Noneself.lazy_init
>temp_235|source:-*- coding: utf-8 -*-|start:6612|stop:7112
alscores = ''while 1:line = self.file_pointer.readline()if not line:if len(quals
cores):self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.qua
ls.split()]self.pos += 1return Trueelse:return Falseif line.startswith('>'):self
.file_pointer.seek(self.file_pointer.tell() - len(line))breakqualscores += ' ' +
 line.strip()self.quals = qualscores.strip()self.quals_int = [int(q) for q in se
lf.quals.split()]self.pos += 1return Truedef close(self):self.file_pointer.close
()def reset(self):se
>temp_236|source:-*- coding: utf-8 -*-|start:1229|stop:1729
ta = SequenceSource(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0
 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory
: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.se
quences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.c
lose()class SequenceSource():def __init__(self, fasta_file_path, lazy_init = Tru
e, unique = False, allow_mixed_case = False):self.fasta_file_path = fasta_file_p
athself.name = Nones
>temp_237|source:-*- coding: utf-8 -*-|start:3739|stop:4239
ile_pointer.close()def reset(self):self.pos = 0self.id  = Noneself.seq = Nonesel
f.ids = []self.file_pointer.seek(0)def visualize_sequence_length_distribution(se
lf, title, dest = None, max_seq_len = None, xtickstep = None, ytickstep = None):
import matplotlib.pyplot as pltimport matplotlib.gridspec as gridspecsequence_le
ngths = []self.reset()while self.next():if self.pos % 10000 == 0 or self.pos == 
1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()seq
uence_lengths.append
>temp_238|source:-*- coding: utf-8 -*-|start:4410|stop:4910
[0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_distribution[l] += 1fig
 = plt.figure(figsize = (16, 12))plt.rcParams.update({'axes.linewidth' : 0.9})pl
t.rc('grid', color='0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(1
0, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bott
om = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_distribution, color = 'blac
k', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distribution
, y2 = 0, color = 'b
>temp_239|source:-*- coding: utf-8 -*-|start:6355|stop:6855
seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l f
or l in self.file_pointer.readlines() if l.startswith('>')])self.reset()def next
(self):self.id = self.file_pointer.readline()[1:].strip()self.quals = Noneself.q
uals_int = Nonequalscores = ''while 1:line = self.file_pointer.readline()if not 
line:if len(qualscores):self.quals = qualscores.strip()self.quals_int = [int(q) 
for q in self.quals.split()]self.pos += 1return Trueelse:return Falseif line.sta
rtswith('>'):self.fi
>temp_240|source:-*- coding: utf-8 -*-|start:1507|stop:2007
rr.write('\n')def close(self):self.fasta.close()class SequenceSource():def __ini
t__(self, fasta_file_path, lazy_init = True, unique = False, allow_mixed_case = 
False):self.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init = la
zy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.s
eq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_
hash_list = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_file_
path)self.file_point
>temp_241|source:-*- coding: utf-8 -*-|start:2285|stop:2785
).hexdigest()if hash in self.unique_hash_dict:self.unique_hash_dict[hash]['ids']
.append(self.id)self.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_d
ict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'count': 1}self.uni
que_hash_list = [i[1] for i in sorted([(self.unique_hash_dict[hash]['count'], ha
sh)\for hash in self.unique_hash_dict], reverse = True)]self.total_unique = len(
self.unique_hash_dict)self.reset()def next(self):if self.unique:return self.next
_unique()else:return
>temp_242|source:-*- coding: utf-8 -*-|start:2331|stop:2831
self.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['c
ount'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'
seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.u
nique_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_dict], revers
e = True)]self.total_unique = len(self.unique_hash_dict)self.reset()def next(sel
f):if self.unique:return self.next_unique()else:return self.next_regular()def ne
xt_unique(self):if s
>temp_243|source:-*- coding: utf-8 -*-|start:1940|stop:2440
sh = 0self.file_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if 
self.lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for l in self.f
ile_pointer.readlines() if l.startswith('>')])self.reset()if self.unique:self.in
it_unique_hash()def init_unique_hash(self):while self.next_regular():hash = hash
lib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.uniq
ue_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] +=
 1else:self.unique_h
>temp_244|source:-*- coding: utf-8 -*-|start:1708|stop:2208
pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mixe
d_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = uniqu
eself.unique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0se
lf.file_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if self.laz
y_init:self.total_seq = Noneelse:self.total_seq = len([l for l in self.file_poin
ter.readlines() if l.startswith('>')])self.reset()if self.unique:self.init_uniqu
e_hash()def init_uni
>temp_245|source:-*- coding: utf-8 -*-|start:4062|stop:4562
hs = []self.reset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:s
ys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()sequen
ce_lengths.append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_seq
_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) o
r 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq
_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update
({'axes.linewidth' :
>temp_246|source:-*- coding: utf-8 -*-|start:3068|stop:3568
.id  = hash_entry['id']self.ids = hash_entry['ids']return Trueelse:return Falsee
lse:return Falsedef next_regular(self):self.seq = Noneself.id = self.file_pointe
r.readline()[1:].strip()sequence = ''while 1:line = self.file_pointer.readline()
if not line:if len(sequence):self.seq = sequenceself.pos += 1return Trueelse:ret
urn Falseif line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell()
 - len(line))breaksequence += line.strip()self.seq = sequence if self.allow_mixe
d_case else sequence
>temp_247|source:-*- coding: utf-8 -*-|start:868|stop:1368
)self.output_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length 
= 80):ticks = range(0, len(sequence), piece_length) + [len(sequence)]return '\n'
.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def cl
ose(self):self.output_file_obj.close()class ReadFasta:def __init__(self, f_name)
:self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)while self.f
asta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.writ
e('\r[fastalib] Read
>temp_248|source:-*- coding: utf-8 -*-|start:1400|stop:1900
lf.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.app
end(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close()class
 SequenceSource():def __init__(self, fasta_file_path, lazy_init = True, unique =
 False, allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.nam
e = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself.p
os = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.unique
_hash_dict = {}self.
>temp_249|source:-*- coding: utf-8 -*-|start:6457|stop:6957
ter.readlines() if l.startswith('>')])self.reset()def next(self):self.id = self.
file_pointer.readline()[1:].strip()self.quals = Noneself.quals_int = Nonequalsco
res = ''while 1:line = self.file_pointer.readline()if not line:if len(qualscores
):self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.quals.sp
lit()]self.pos += 1return Trueelse:return Falseif line.startswith('>'):self.file
_pointer.seek(self.file_pointer.tell() - len(line))breakqualscores += ' ' + line
.strip()self.quals =
>temp_250|source:-*- coding: utf-8 -*-|start:4640|stop:5140
GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=
0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_distribution, col
or = 'black', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_di
stribution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequence
s')plt.xlabel('sequence length')if xtickstep == None:xtickstep = (max_seq_len / 
50) or 1if ytickstep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt.
xticks(range(xtickst
>temp_251|source:-*- coding: utf-8 -*-|start:6787|stop:7287
.pos += 1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.s
eek(self.file_pointer.tell() - len(line))breakqualscores += ' ' + line.strip()se
lf.quals = qualscores.strip()self.quals_int = [int(q) for q in self.quals.split(
)]self.pos += 1return Truedef close(self):self.file_pointer.close()def reset(sel
f):self.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = [
]self.file_pointer.seek(0)if __name__ == '__main__':fasta = SequenceSource(sys.a
rgv[1])fasta.visuali
>temp_252|source:-*- coding: utf-8 -*-|start:1923|stop:2423
lf.unique_next_hash = 0self.file_pointer = open(self.fasta_file_path)self.file_p
ointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len([
l for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()if se
lf.unique:self.init_unique_hash()def init_unique_hash(self):while self.next_regu
lar():hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_ha
sh_dict:self.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[
hash]['count'] += 1e
>temp_253|source:-*- coding: utf-8 -*-|start:1221|stop:1721
self.fasta = SequenceSource(f_name)while self.fasta.next():if self.fasta.pos % 1
000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA int
o memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id
)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self
.fasta.close()class SequenceSource():def __init__(self, fasta_file_path, lazy_in
it = True, unique = False, allow_mixed_case = False):self.fasta_file_path = fast
a_file_pathself.name
>temp_254|source:-*- coding: utf-8 -*-|start:3386|stop:3886
eturn Falseif line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell
() - len(line))breaksequence += line.strip()self.seq = sequence if self.allow_mi
xed_case else sequence.upper()self.pos += 1return Truedef get_seq_by_read_id(sel
f, read_id):self.reset()while self.next():if self.id == read_id:return self.seqr
eturn Falsedef close(self):self.file_pointer.close()def reset(self):self.pos = 0
self.id  = Noneself.seq = Noneself.ids = []self.file_pointer.seek(0)def visualiz
e_sequence_length_di
>temp_255|source:-*- coding: utf-8 -*-|start:3774|stop:4274
self.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.file_pointer.seek(0)
def visualize_sequence_length_distribution(self, title, dest = None, max_seq_len
 = None, xtickstep = None, ytickstep = None):import matplotlib.pyplot as pltimpo
rt matplotlib.gridspec as gridspecsequence_lengths = []self.reset()while self.ne
xt():if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fastalib] Re
ading: %s' % (self.pos))sys.stderr.flush()sequence_lengths.append(len(self.seq))
self.reset()sys.stde
>temp_256|source:-*- coding: utf-8 -*-|start:1183|stop:1683
name):self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)while s
elf.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr
.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))sys.stder
r.flush()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.seq)sys.
stderr.write('\n')def close(self):self.fasta.close()class SequenceSource():def _
_init__(self, fasta_file_path, lazy_init = True, unique = False, allow_mixed_cas
e = False):self.fast
>temp_257|source:-*- coding: utf-8 -*-|start:4459|stop:4959
seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.upd
ate({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewid
th=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.
subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_
len_distribution, color = 'black', alpha = 0.3)plt.fill_between(range(0, max_seq
_len + 1), seq_len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabe
l('number of sequenc
>temp_258|source:-*- coding: utf-8 -*-|start:4202|stop:4702
stderr.flush()sequence_lengths.append(len(self.seq))self.reset()sys.stderr.write
('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence
_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in 
sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12)
)plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', lines
tyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])p
lt.grid(True)plt.sub
>temp_259|source:-*- coding: utf-8 -*-|start:4278|stop:4778
rite('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequ
ence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l
 in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16,
 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', l
inestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:
8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, right
 = 0.98)plt.plot(seq
>temp_260|source:-*- coding: utf-8 -*-|start:2375|stop:2875
lf.id)self.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_dict[hash] 
= {'id' : self.id,'ids': [self.id],'seq': self.seq,'count': 1}self.unique_hash_l
ist = [i[1] for i in sorted([(self.unique_hash_dict[hash]['count'], hash)\for ha
sh in self.unique_hash_dict], reverse = True)]self.total_unique = len(self.uniqu
e_hash_dict)self.reset()def next(self):if self.unique:return self.next_unique()e
lse:return self.next_regular()def next_unique(self):if self.unique:if self.total
_unique > 0 and self
>temp_261|source:-*- coding: utf-8 -*-|start:1709|stop:2209
athself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed
_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = unique
self.unique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0sel
f.file_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy
_init:self.total_seq = Noneelse:self.total_seq = len([l for l in self.file_point
er.readlines() if l.startswith('>')])self.reset()if self.unique:self.init_unique
_hash()def init_uniq
>temp_262|source:-*- coding: utf-8 -*-|start:3141|stop:3641
 Falseelse:return Falsedef next_regular(self):self.seq = Noneself.id = self.file
_pointer.readline()[1:].strip()sequence = ''while 1:line = self.file_pointer.rea
dline()if not line:if len(sequence):self.seq = sequenceself.pos += 1return Truee
lse:return Falseif line.startswith('>'):self.file_pointer.seek(self.file_pointer
.tell() - len(line))breaksequence += line.strip()self.seq = sequence if self.all
ow_mixed_case else sequence.upper()self.pos += 1return Truedef get_seq_by_read_i
d(self, read_id):sel
>temp_263|source:-*- coding: utf-8 -*-|start:1016|stop:1516
]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) 
- 1)])def close(self):self.output_file_obj.close()class ReadFasta:def __init__(s
elf, f_name):self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)
while self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys
.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))sy
s.stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.s
eq)sys.stderr.write(
>temp_264|source:-*- coding: utf-8 -*-|start:3129|stop:3629
eelse:return Falseelse:return Falsedef next_regular(self):self.seq = Noneself.id
 = self.file_pointer.readline()[1:].strip()sequence = ''while 1:line = self.file
_pointer.readline()if not line:if len(sequence):self.seq = sequenceself.pos += 1
return Trueelse:return Falseif line.startswith('>'):self.file_pointer.seek(self.
file_pointer.tell() - len(line))breaksequence += line.strip()self.seq = sequence
 if self.allow_mixed_case else sequence.upper()self.pos += 1return Truedef get_s
eq_by_read_id(self, 
>temp_265|source:-*- coding: utf-8 -*-|start:2536|stop:3036
st = [i[1] for i in sorted([(self.unique_hash_dict[hash]['count'], hash)\for has
h in self.unique_hash_dict], reverse = True)]self.total_unique = len(self.unique
_hash_dict)self.reset()def next(self):if self.unique:return self.next_unique()el
se:return self.next_regular()def next_unique(self):if self.unique:if self.total_
unique > 0 and self.pos < self.total_unique:hash_entry = self.unique_hash_dict[s
elf.unique_hash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self
.allow_mixed_case el
>temp_266|source:-*- coding: utf-8 -*-|start:5141|stop:5641
p, max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0,
 max(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_dist
ribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len
)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_distribution
) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-la
rge', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor
' : 20})plt.grid(Fal
>temp_267|source:-*- coding: utf-8 -*-|start:324|stop:824
le.import sysimport numpyimport hashlibclass FastaOutput:def __init__(self, outp
ut_file_path):self.output_file_path = output_file_pathself.output_file_obj = ope
n(output_file_path, 'w')def store(self, entry, split = True, store_frequencies =
 True):if entry.unique and store_frequencies:self.write_id('%s|%s' % (entry.id, 
'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_seq(entr
y.seq, split)def write_id(self, id):self.output_file_obj.write('>%s\n' % id)def 
write_seq(self, seq,
>temp_268|source:-*- coding: utf-8 -*-|start:2098|stop:2598
.file_pointer.readlines() if l.startswith('>')])self.reset()if self.unique:self.
init_unique_hash()def init_unique_hash(self):while self.next_regular():hash = ha
shlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.un
ique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] 
+= 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': s
elf.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.unique_h
ash_dict[hash]['coun
>temp_269|source:-*- coding: utf-8 -*-|start:3780|stop:4280
os = 0self.id  = Noneself.seq = Noneself.ids = []self.file_pointer.seek(0)def vi
sualize_sequence_length_distribution(self, title, dest = None, max_seq_len = Non
e, xtickstep = None, ytickstep = None):import matplotlib.pyplot as pltimport mat
plotlib.gridspec as gridspecsequence_lengths = []self.reset()while self.next():i
f self.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fastalib] Reading:
 %s' % (self.pos))sys.stderr.flush()sequence_lengths.append(len(self.seq))self.r
eset()sys.stderr.wri
>temp_270|source:-*- coding: utf-8 -*-|start:6744|stop:7244
 = [int(q) for q in self.quals.split()]self.pos += 1return Trueelse:return False
if line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(li
ne))breakqualscores += ' ' + line.strip()self.quals = qualscores.strip()self.qua
ls_int = [int(q) for q in self.quals.split()]self.pos += 1return Truedef close(s
elf):self.file_pointer.close()def reset(self):self.pos = 0self.id  = Noneself.qu
als = Noneself.quals_int = Noneself.ids = []self.file_pointer.seek(0)if __name__
 == '__main__':fasta
>temp_271|source:-*- coding: utf-8 -*-|start:1951|stop:2451
file_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy_i
nit:self.total_seq = Noneelse:self.total_seq = len([l for l in self.file_pointer
.readlines() if l.startswith('>')])self.reset()if self.unique:self.init_unique_h
ash()def init_unique_hash(self):while self.next_regular():hash = hashlib.sha1(se
lf.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_hash_dic
t[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] += 1else:self
.unique_hash_dict[ha
>temp_272|source:-*- coding: utf-8 -*-|start:2712|stop:3212
reset()def next(self):if self.unique:return self.next_unique()else:return self.n
ext_regular()def next_unique(self):if self.unique:if self.total_unique > 0 and s
elf.pos < self.total_unique:hash_entry = self.unique_hash_dict[self.unique_hash_
list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_cas
e else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_entry
['ids']return Trueelse:return Falseelse:return Falsedef next_regular(self):self.
seq = Noneself.id = 
>temp_273|source:-*- coding: utf-8 -*-|start:2713|stop:3213
eset()def next(self):if self.unique:return self.next_unique()else:return self.ne
xt_regular()def next_unique(self):if self.unique:if self.total_unique > 0 and se
lf.pos < self.total_unique:hash_entry = self.unique_hash_dict[self.unique_hash_l
ist[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_case
 else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_entry[
'ids']return Trueelse:return Falseelse:return Falsedef next_regular(self):self.s
eq = Noneself.id = s
>temp_274|source:-*- coding: utf-8 -*-|start:6370|stop:6870
lazy_init:self.total_quals = Noneelse:self.total_quals = len([l for l in self.fi
le_pointer.readlines() if l.startswith('>')])self.reset()def next(self):self.id 
= self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_int = None
qualscores = ''while 1:line = self.file_pointer.readline()if not line:if len(qua
lscores):self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.q
uals.split()]self.pos += 1return Trueelse:return Falseif line.startswith('>'):se
lf.file_pointer.seek
>temp_275|source:-*- coding: utf-8 -*-|start:5250|stop:5750
1, ytickstep),[y for y in range(0, max(seq_len_distribution) + 1, ytickstep)],si
ze='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = ma
x(seq_len_distribution) + (max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.
96, '%s' % (title), weight = 'black', size = 'xx-large', ha = 'center')ax1 = plt
.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)plt.yt
icks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / 
min: %s / max: %s'\%
>temp_276|source:-*- coding: utf-8 -*-|start:532|stop:1032
plit = True, store_frequencies = True):if entry.unique and store_frequencies:sel
f.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write
_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, id):self.output_
file_obj.write('>%s\n' % id)def write_seq(self, seq, split = True):if split:seq 
= self.split(seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequenc
e, piece_length = 80):ticks = range(0, len(sequence), piece_length) + [len(seque
nce)]return '\n'.joi
>temp_277|source:-*- coding: utf-8 -*-|start:2391|stop:2891
e_hash_dict[hash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.i
d,'ids': [self.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for 
i in sorted([(self.unique_hash_dict[hash]['count'], hash)\for hash in self.uniqu
e_hash_dict], reverse = True)]self.total_unique = len(self.unique_hash_dict)self
.reset()def next(self):if self.unique:return self.next_unique()else:return self.
next_regular()def next_unique(self):if self.unique:if self.total_unique > 0 and 
self.pos < self.tota
>temp_278|source:-*- coding: utf-8 -*-|start:5577|stop:6077
t(gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([]
)plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s
 / max: %s'\% (len(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(seq
uence_lengths),\min(sequence_lengths),\max(sequence_lengths)),\va = 'center', al
pha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_pathtry:plt.s
avefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passr
eturnclass QualSourc
>temp_279|source:-*- coding: utf-8 -*-|start:6215|stop:6715
lf.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self
.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if self.lazy_
init:self.total_quals = Noneelse:self.total_quals = len([l for l in self.file_po
inter.readlines() if l.startswith('>')])self.reset()def next(self):self.id = sel
f.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_int = Nonequals
cores = ''while 1:line = self.file_pointer.readline()if not line:if len(qualscor
es):self.quals = qua
>temp_280|source:-*- coding: utf-8 -*-|start:905|stop:1405
seq)def split(self, sequence, piece_length = 80):ticks = range(0, len(sequence),
 piece_length) + [len(sequence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]
] for x in range(0, len(ticks) - 1)])def close(self):self.output_file_obj.close(
)class ReadFasta:def __init__(self, f_name):self.ids = []self.sequences = []self
.fasta = SequenceSource(f_name)while self.fasta.next():if self.fasta.pos % 1000 
== 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into me
mory: %s' % (self.fa
>temp_281|source:-*- coding: utf-8 -*-|start:3510|stop:4010
self.seq = sequence if self.allow_mixed_case else sequence.upper()self.pos += 1r
eturn Truedef get_seq_by_read_id(self, read_id):self.reset()while self.next():if
 self.id == read_id:return self.seqreturn Falsedef close(self):self.file_pointer
.close()def reset(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids = []s
elf.file_pointer.seek(0)def visualize_sequence_length_distribution(self, title, 
dest = None, max_seq_len = None, xtickstep = None, ytickstep = None):import matp
lotlib.pyplot as plt
>temp_282|source:-*- coding: utf-8 -*-|start:4997|stop:5497
ickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:ytickste
p = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len +
 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, max(seq_len_dis
tribution) + 1, ytickstep),[y for y in range(0, max(seq_len_distribution) + 1, y
tickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin =
 0, ymax = max(seq_len_distribution) + (max(seq_len_distribution) / 20.0))plt.fi
gtext(0.5, 0.96, '%s
>temp_283|source:-*- coding: utf-8 -*-|start:5949|stop:6449
elf.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png
')try:plt.show()except:passreturnclass QualSource:def __init__(self, quals_file_
path, lazy_init = True):self.quals_file_path = quals_file_pathself.name = Nonese
lf.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_i
nt = Noneself.ids = []self.file_pointer = open(self.quals_file_path)self.file_po
inter.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = le
n([l for l in self.f
>temp_284|source:-*- coding: utf-8 -*-|start:6819|stop:7319
Falseif line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - l
en(line))breakqualscores += ' ' + line.strip()self.quals = qualscores.strip()sel
f.quals_int = [int(q) for q in self.quals.split()]self.pos += 1return Truedef cl
ose(self):self.file_pointer.close()def reset(self):self.pos = 0self.id  = Nonese
lf.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer.seek(0)if __n
ame__ == '__main__':fasta = SequenceSource(sys.argv[1])fasta.visualize_sequence_
length_distribution(
>temp_285|source:-*- coding: utf-8 -*-|start:6145|stop:6645
ile_path = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0
self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_poin
ter = open(self.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.
total_quals = Noneelse:self.total_quals = len([l for l in self.file_pointer.read
lines() if l.startswith('>')])self.reset()def next(self):self.id = self.file_poi
nter.readline()[1:].strip()self.quals = Noneself.quals_int = Nonequalscores = ''
while 1:line = self.
>temp_286|source:-*- coding: utf-8 -*-|start:5314|stop:5814
, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymi
n = 0, ymax = max(seq_len_distribution) + (max(seq_len_distribution) / 20.0))plt
.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-large', ha = 'c
enter')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.g
rid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2
f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence_
lengths), numpy.std(
>temp_287|source:-*- coding: utf-8 -*-|start:4065|stop:4565
= []self.reset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.
stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_
lengths.append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_seq_le
n:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or 1
0)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_le
n_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update({'
axes.linewidth' : 0.
>temp_288|source:-*- coding: utf-8 -*-|start:2704|stop:3204
ct)self.reset()def next(self):if self.unique:return self.next_unique()else:retur
n self.next_regular()def next_unique(self):if self.unique:if self.total_unique >
 0 and self.pos < self.total_unique:hash_entry = self.unique_hash_dict[self.uniq
ue_hash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.allow_m
ixed_case else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = ha
sh_entry['ids']return Trueelse:return Falseelse:return Falsedef next_regular(sel
f):self.seq = Nonese
>temp_289|source:-*- coding: utf-8 -*-|start:3202|stop:3702
self.id = self.file_pointer.readline()[1:].strip()sequence = ''while 1:line = se
lf.file_pointer.readline()if not line:if len(sequence):self.seq = sequenceself.p
os += 1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.see
k(self.file_pointer.tell() - len(line))breaksequence += line.strip()self.seq = s
equence if self.allow_mixed_case else sequence.upper()self.pos += 1return Truede
f get_seq_by_read_id(self, read_id):self.reset()while self.next():if self.id == 
read_id:return self.
>temp_290|source:-*- coding: utf-8 -*-|start:1327|stop:1827
 == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fast
a.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(sel
f.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close()class Sequen
ceSource():def __init__(self, fasta_file_path, lazy_init = True, unique = False,
 allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.name = Non
eself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0s
elf.id  = Noneself.s
>temp_291|source:-*- coding: utf-8 -*-|start:5461|stop:5961
) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-la
rge', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor
' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %
s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.m
ean(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_lengths),\max(s
equence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == None:
dest = self.fasta_fi
>temp_292|source:-*- coding: utf-8 -*-|start:517|stop:1017
(self, entry, split = True, store_frequencies = True):if entry.unique and store_
frequencies:self.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))
else:self.write_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, i
d):self.output_file_obj.write('>%s\n' % id)def write_seq(self, seq, split = True
):if split:seq = self.split(seq)self.output_file_obj.write('%s\n' % seq)def spli
t(self, sequence, piece_length = 80):ticks = range(0, len(sequence), piece_lengt
h) + [len(sequence)]
>temp_293|source:-*- coding: utf-8 -*-|start:516|stop:1016
e(self, entry, split = True, store_frequencies = True):if entry.unique and store
_frequencies:self.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids))
)else:self.write_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, 
id):self.output_file_obj.write('>%s\n' % id)def write_seq(self, seq, split = Tru
e):if split:seq = self.split(seq)self.output_file_obj.write('%s\n' % seq)def spl
it(self, sequence, piece_length = 80):ticks = range(0, len(sequence), piece_leng
th) + [len(sequence)
>temp_294|source:-*- coding: utf-8 -*-|start:690|stop:1190
te_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, id):self.outpu
t_file_obj.write('>%s\n' % id)def write_seq(self, seq, split = True):if split:se
q = self.split(seq)self.output_file_obj.write('%s\n' % seq)def split(self, seque
nce, piece_length = 80):ticks = range(0, len(sequence), piece_length) + [len(seq
uence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(t
icks) - 1)])def close(self):self.output_file_obj.close()class ReadFasta:def __in
it__(self, f_name):s
>temp_295|source:-*- coding: utf-8 -*-|start:2326|stop:2826
dict:self.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[has
h]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.
id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(s
elf.unique_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_dict], r
everse = True)]self.total_unique = len(self.unique_hash_dict)self.reset()def nex
t(self):if self.unique:return self.next_unique()else:return self.next_regular()d
ef next_unique(self)
>temp_296|source:-*- coding: utf-8 -*-|start:1047|stop:1547
ks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.outp
ut_file_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = []self.
sequences = []self.fasta = SequenceSource(f_name)while self.fasta.next():if self
.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Rea
ding FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append
(self.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def cl
ose(self):self.fasta
>temp_297|source:-*- coding: utf-8 -*-|start:6232|stop:6732
  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer = o
pen(self.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_q
uals = Noneelse:self.total_quals = len([l for l in self.file_pointer.readlines()
 if l.startswith('>')])self.reset()def next(self):self.id = self.file_pointer.re
adline()[1:].strip()self.quals = Noneself.quals_int = Nonequalscores = ''while 1
:line = self.file_pointer.readline()if not line:if len(qualscores):self.quals = 
qualscores.strip()se
>temp_298|source:-*- coding: utf-8 -*-|start:5035|stop:5535
en / 50) or 1if ytickstep == None:ytickstep = max(seq_len_distribution) / 20 or 
1plt.xticks(range(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-
small')plt.yticks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y in
 range(0, max(seq_len_distribution) + 1, ytickstep)],size='xx-small')plt.xlim(xm
in = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) +
 (max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weigh
t = 'black', size = 
>temp_299|source:-*- coding: utf-8 -*-|start:6768|stop:7268
.quals.split()]self.pos += 1return Trueelse:return Falseif line.startswith('>'):
self.file_pointer.seek(self.file_pointer.tell() - len(line))breakqualscores += '
 ' + line.strip()self.quals = qualscores.strip()self.quals_int = [int(q) for q i
n self.quals.split()]self.pos += 1return Truedef close(self):self.file_pointer.c
lose()def reset(self):self.pos = 0self.id  = Noneself.quals = Noneself.quals_int
 = Noneself.ids = []self.file_pointer.seek(0)if __name__ == '__main__':fasta = S
equenceSource(sys.ar
>temp_300|source:-*- coding: utf-8 -*-|start:5063|stop:5563
 None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep,
 max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, m
ax(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_distri
bution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)p
lt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_distribution) 
/ 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-larg
e', ha = 'center')ax
>temp_301|source:-*- coding: utf-8 -*-|start:2670|stop:3170
l_unique = len(self.unique_hash_dict)self.reset()def next(self):if self.unique:r
eturn self.next_unique()else:return self.next_regular()def next_unique(self):if 
self.unique:if self.total_unique > 0 and self.pos < self.total_unique:hash_entry
 = self.unique_hash_dict[self.unique_hash_list[self.pos]]self.pos += 1self.seq =
 hash_entry['seq'] if self.allow_mixed_case else hash_entry['seq'].upper()self.i
d  = hash_entry['id']self.ids = hash_entry['ids']return Trueelse:return Falseels
e:return Falsedef ne
>temp_302|source:-*- coding: utf-8 -*-|start:6176|stop:6676
name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Non
eself.quals_int = Noneself.ids = []self.file_pointer = open(self.quals_file_path
)self.file_pointer.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.tot
al_quals = len([l for l in self.file_pointer.readlines() if l.startswith('>')])s
elf.reset()def next(self):self.id = self.file_pointer.readline()[1:].strip()self
.quals = Noneself.quals_int = Nonequalscores = ''while 1:line = self.file_pointe
r.readline()if not l
>temp_303|source:-*- coding: utf-8 -*-|start:5219|stop:5719
0, max(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_di
stribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_l
en)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_distributi
on) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-
large', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecol
or' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total:
 %s / mean: %.2f / s
>temp_304|source:-*- coding: utf-8 -*-|start:3939|stop:4439
ne, xtickstep = None, ytickstep = None):import matplotlib.pyplot as pltimport ma
tplotlib.gridspec as gridspecsequence_lengths = []self.reset()while self.next():
if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fastalib] Reading
: %s' % (self.pos))sys.stderr.flush()sequence_lengths.append(len(self.seq))self.
reset()sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(sequence_lengt
hs) + (int(max(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (ma
x_seq_len + 1)for l 
>temp_305|source:-*- coding: utf-8 -*-|start:4551|stop:5051
inewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewidth=0.1)gs = 
gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_adj
ust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_distribu
tion, color = 'black', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), s
eq_len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of
 sequences')plt.xlabel('sequence length')if xtickstep == None:xtickstep = (max_s
eq_len / 50) or 1if 
>temp_306|source:-*- coding: utf-8 -*-|start:4839|stop:5339
een(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black', al
pha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence length')if xtic
kstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:ytickstep 
= max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1
, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, max(seq_len_distr
ibution) + 1, ytickstep),[y for y in range(0, max(seq_len_distribution) + 1, yti
ckstep)],size='xx-sm
>temp_307|source:-*- coding: utf-8 -*-|start:6457|stop:6957
ter.readlines() if l.startswith('>')])self.reset()def next(self):self.id = self.
file_pointer.readline()[1:].strip()self.quals = Noneself.quals_int = Nonequalsco
res = ''while 1:line = self.file_pointer.readline()if not line:if len(qualscores
):self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.quals.sp
lit()]self.pos += 1return Trueelse:return Falseif line.startswith('>'):self.file
_pointer.seek(self.file_pointer.tell() - len(line))breakqualscores += ' ' + line
.strip()self.quals =
>temp_308|source:-*- coding: utf-8 -*-|start:6575|stop:7075
f.quals = Noneself.quals_int = Nonequalscores = ''while 1:line = self.file_point
er.readline()if not line:if len(qualscores):self.quals = qualscores.strip()self.
quals_int = [int(q) for q in self.quals.split()]self.pos += 1return Trueelse:ret
urn Falseif line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell()
 - len(line))breakqualscores += ' ' + line.strip()self.quals = qualscores.strip(
)self.quals_int = [int(q) for q in self.quals.split()]self.pos += 1return Truede
f close(self):self.f
>temp_309|source:-*- coding: utf-8 -*-|start:258|stop:758
your option)# any later version.## Please read the docs/COPYING file.import sysi
mport numpyimport hashlibclass FastaOutput:def __init__(self, output_file_path):
self.output_file_path = output_file_pathself.output_file_obj = open(output_file_
path, 'w')def store(self, entry, split = True, store_frequencies = True):if entr
y.unique and store_frequencies:self.write_id('%s|%s' % (entry.id, 'frequency:%d'
 % len(entry.ids)))else:self.write_id(entry.id)self.write_seq(entry.seq, split)d
ef write_id(self, id
>temp_310|source:-*- coding: utf-8 -*-|start:1668|stop:2168
alse):self.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init = laz
y_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.se
q = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_h
ash_list = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_file_p
ath)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.to
tal_seq = len([l for l in self.file_pointer.readlines() if l.startswith('>')])se
lf.reset()if self.un
>temp_311|source:-*- coding: utf-8 -*-|start:4263|stop:4763
t()sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) 
+ (int(max(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_se
q_len + 1)for l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure
(figsize = (16, 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', 
color='0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = pl
t.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, to
p = 0.95, right = 0.
>temp_312|source:-*- coding: utf-8 -*-|start:1098|stop:1598
1)])def close(self):self.output_file_obj.close()class ReadFasta:def __init__(sel
f, f_name):self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)wh
ile self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.s
tderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))sys.
stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.seq
)sys.stderr.write('\n')def close(self):self.fasta.close()class SequenceSource():
def __init__(self, f
>temp_313|source:-*- coding: utf-8 -*-|start:2790|stop:3290
.next_regular()def next_unique(self):if self.unique:if self.total_unique > 0 and
 self.pos < self.total_unique:hash_entry = self.unique_hash_dict[self.unique_has
h_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_c
ase else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_ent
ry['ids']return Trueelse:return Falseelse:return Falsedef next_regular(self):sel
f.seq = Noneself.id = self.file_pointer.readline()[1:].strip()sequence = ''while
 1:line = self.file_
>temp_314|source:-*- coding: utf-8 -*-|start:5702|stop:6202
 / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.me
an(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_lengths),\max(se
quence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == None:d
est = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest 
+ '.png')try:plt.show()except:passreturnclass QualSource:def __init__(self, qual
s_file_path, lazy_init = True):self.quals_file_path = quals_file_pathself.name =
 Noneself.lazy_init 
>temp_315|source:-*- coding: utf-8 -*-|start:4387|stop:4887
seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_
distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update({'ax
es.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewidth=0.1)g
s = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots
_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_dist
ribution, color = 'black', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1
), seq_len_distribut
>temp_316|source:-*- coding: utf-8 -*-|start:5927|stop:6427
 dest == None:dest = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:pl
t.savefig(dest + '.png')try:plt.show()except:passreturnclass QualSource:def __in
it__(self, quals_file_path, lazy_init = True):self.quals_file_path = quals_file_
pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.qu
als = Noneself.quals_int = Noneself.ids = []self.file_pointer = open(self.quals_
file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_quals = Noneelse
:self.total_quals = 
>temp_317|source:-*- coding: utf-8 -*-|start:2128|stop:2628
.startswith('>')])self.reset()if self.unique:self.init_unique_hash()def init_uni
que_hash(self):while self.next_regular():hash = hashlib.sha1(self.seq.upper()).h
exdigest()if hash in self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].ap
pend(self.id)self.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_dict
[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'count': 1}self.unique
_hash_list = [i[1] for i in sorted([(self.unique_hash_dict[hash]['count'], hash)
\for hash in self.un
>temp_318|source:-*- coding: utf-8 -*-|start:2873|stop:3373
lf.pos < self.total_unique:hash_entry = self.unique_hash_dict[self.unique_hash_l
ist[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_case
 else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_entry[
'ids']return Trueelse:return Falseelse:return Falsedef next_regular(self):self.s
eq = Noneself.id = self.file_pointer.readline()[1:].strip()sequence = ''while 1:
line = self.file_pointer.readline()if not line:if len(sequence):self.seq = seque
nceself.pos += 1retu
>temp_319|source:-*- coding: utf-8 -*-|start:5750|stop:6250
 (len(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(sequence_lengths
),\min(sequence_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, si
ze = 'x-large')if dest == None:dest = self.fasta_file_pathtry:plt.savefig(dest +
 '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passreturnclass Qu
alSource:def __init__(self, quals_file_path, lazy_init = True):self.quals_file_p
ath = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.
id  = Noneself.quals
>temp_320|source:-*- coding: utf-8 -*-|start:6123|stop:6623
t = True):self.quals_file_path = quals_file_pathself.name = Noneself.lazy_init =
 lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.
ids = []self.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)i
f self.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l for l in 
self.file_pointer.readlines() if l.startswith('>')])self.reset()def next(self):s
elf.id = self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_int
 = Nonequalscores = 
>temp_321|source:-*- coding: utf-8 -*-|start:460|stop:960
lf.output_file_obj = open(output_file_path, 'w')def store(self, entry, split = T
rue, store_frequencies = True):if entry.unique and store_frequencies:self.write_
id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entr
y.id)self.write_seq(entry.seq, split)def write_id(self, id):self.output_file_obj
.write('>%s\n' % id)def write_seq(self, seq, split = True):if split:seq = self.s
plit(seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequence, piece
_length = 80):ticks 
>temp_322|source:-*- coding: utf-8 -*-|start:5890|stop:6390
er', alpha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_pathtr
y:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()excep
t:passreturnclass QualSource:def __init__(self, quals_file_path, lazy_init = Tru
e):self.quals_file_path = quals_file_pathself.name = Noneself.lazy_init = lazy_i
nitself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = [
]self.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if self.
lazy_init:self.total
>temp_323|source:-*- coding: utf-8 -*-|start:3447|stop:3947
f.file_pointer.tell() - len(line))breaksequence += line.strip()self.seq = sequen
ce if self.allow_mixed_case else sequence.upper()self.pos += 1return Truedef get
_seq_by_read_id(self, read_id):self.reset()while self.next():if self.id == read_
id:return self.seqreturn Falsedef close(self):self.file_pointer.close()def reset
(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.file_pointer.
seek(0)def visualize_sequence_length_distribution(self, title, dest = None, max_
seq_len = None, xtic
>temp_324|source:-*- coding: utf-8 -*-|start:4203|stop:4703
tderr.flush()sequence_lengths.append(len(self.seq))self.reset()sys.stderr.write(
'\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_
lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in s
equence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))
plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linest
yle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])pl
t.grid(True)plt.subp
>temp_325|source:-*- coding: utf-8 -*-|start:372|stop:872
taOutput:def __init__(self, output_file_path):self.output_file_path = output_fil
e_pathself.output_file_obj = open(output_file_path, 'w')def store(self, entry, s
plit = True, store_frequencies = True):if entry.unique and store_frequencies:sel
f.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write
_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, id):self.output_
file_obj.write('>%s\n' % id)def write_seq(self, seq, split = True):if split:seq 
= self.split(seq)sel
>temp_326|source:-*- coding: utf-8 -*-|start:5983|stop:6483
(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passreturnc
lass QualSource:def __init__(self, quals_file_path, lazy_init = True):self.quals
_file_path = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos =
 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_po
inter = open(self.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:sel
f.total_quals = Noneelse:self.total_quals = len([l for l in self.file_pointer.re
adlines() if l.start
>temp_327|source:-*- coding: utf-8 -*-|start:5146|stop:5646
x_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, max(
seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_distribut
ion) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.
ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_distribution) / 2
0.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-large',
 ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' : 2
0})plt.grid(False)pl
>temp_328|source:-*- coding: utf-8 -*-|start:5267|stop:5767
for y in range(0, max(seq_len_distribution) + 1, ytickstep)],size='xx-small')plt
.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distrib
ution) + (max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title
), weight = 'black', size = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])pl
t.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])plt.xtick
s([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s / max: %s
'\% (len(sequence_le
>temp_329|source:-*- coding: utf-8 -*-|start:2517|stop:3017
self.unique_hash_list = [i[1] for i in sorted([(self.unique_hash_dict[hash]['cou
nt'], hash)\for hash in self.unique_hash_dict], reverse = True)]self.total_uniqu
e = len(self.unique_hash_dict)self.reset()def next(self):if self.unique:return s
elf.next_unique()else:return self.next_regular()def next_unique(self):if self.un
ique:if self.total_unique > 0 and self.pos < self.total_unique:hash_entry = self
.unique_hash_dict[self.unique_hash_list[self.pos]]self.pos += 1self.seq = hash_e
ntry['seq'] if self.
>temp_330|source:-*- coding: utf-8 -*-|start:675|stop:1175
))else:self.write_id(entry.id)self.write_seq(entry.seq, split)def write_id(self,
 id):self.output_file_obj.write('>%s\n' % id)def write_seq(self, seq, split = Tr
ue):if split:seq = self.split(seq)self.output_file_obj.write('%s\n' % seq)def sp
lit(self, sequence, piece_length = 80):ticks = range(0, len(sequence), piece_len
gth) + [len(sequence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in
 range(0, len(ticks) - 1)])def close(self):self.output_file_obj.close()class Rea
dFasta:def __init__(
>temp_331|source:-*- coding: utf-8 -*-|start:2344|stop:2844
ash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] += 1el
se:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.se
q,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.unique_hash_di
ct[hash]['count'], hash)\for hash in self.unique_hash_dict], reverse = True)]sel
f.total_unique = len(self.unique_hash_dict)self.reset()def next(self):if self.un
ique:return self.next_unique()else:return self.next_regular()def next_unique(sel
f):if self.unique:if
>temp_332|source:-*- coding: utf-8 -*-|start:6746|stop:7246
 [int(q) for q in self.quals.split()]self.pos += 1return Trueelse:return Falseif
 line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line
))breakqualscores += ' ' + line.strip()self.quals = qualscores.strip()self.quals
_int = [int(q) for q in self.quals.split()]self.pos += 1return Truedef close(sel
f):self.file_pointer.close()def reset(self):self.pos = 0self.id  = Noneself.qual
s = Noneself.quals_int = Noneself.ids = []self.file_pointer.seek(0)if __name__ =
= '__main__':fasta =
>temp_333|source:-*- coding: utf-8 -*-|start:2078|stop:2578
len([l for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()
if self.unique:self.init_unique_hash()def init_unique_hash(self):while self.next
_regular():hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.uniq
ue_hash_dict:self.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_
dict[hash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids'
: [self.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in so
rted([(self.unique_h
>temp_334|source:-*- coding: utf-8 -*-|start:726|stop:1226
seq, split)def write_id(self, id):self.output_file_obj.write('>%s\n' % id)def wr
ite_seq(self, seq, split = True):if split:seq = self.split(seq)self.output_file_
obj.write('%s\n' % seq)def split(self, sequence, piece_length = 80):ticks = rang
e(0, len(sequence), piece_length) + [len(sequence)]return '\n'.join([sequence[ti
cks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.out
put_file_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = []self
.sequences = []self.
>temp_335|source:-*- coding: utf-8 -*-|start:3082|stop:3582
try['id']self.ids = hash_entry['ids']return Trueelse:return Falseelse:return Fal
sedef next_regular(self):self.seq = Noneself.id = self.file_pointer.readline()[1
:].strip()sequence = ''while 1:line = self.file_pointer.readline()if not line:if
 len(sequence):self.seq = sequenceself.pos += 1return Trueelse:return Falseif li
ne.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))b
reaksequence += line.strip()self.seq = sequence if self.allow_mixed_case else se
quence.upper()self.p
>temp_336|source:-*- coding: utf-8 -*-|start:5373|stop:5873
_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_dist
ribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size 
= 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.e
dgecolor' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, '
total: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths)
,numpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_lengths
),\max(sequence_leng
>temp_337|source:-*- coding: utf-8 -*-|start:3494|stop:3994
 += line.strip()self.seq = sequence if self.allow_mixed_case else sequence.upper
()self.pos += 1return Truedef get_seq_by_read_id(self, read_id):self.reset()whil
e self.next():if self.id == read_id:return self.seqreturn Falsedef close(self):s
elf.file_pointer.close()def reset(self):self.pos = 0self.id  = Noneself.seq = No
neself.ids = []self.file_pointer.seek(0)def visualize_sequence_length_distributi
on(self, title, dest = None, max_seq_len = None, xtickstep = None, ytickstep = N
one):import matplotl
>temp_338|source:-*- coding: utf-8 -*-|start:2720|stop:3220
ef next(self):if self.unique:return self.next_unique()else:return self.next_regu
lar()def next_unique(self):if self.unique:if self.total_unique > 0 and self.pos 
< self.total_unique:hash_entry = self.unique_hash_dict[self.unique_hash_list[sel
f.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_case else h
ash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_entry['ids']r
eturn Trueelse:return Falseelse:return Falsedef next_regular(self):self.seq = No
neself.id = self.fil
>temp_339|source:-*- coding: utf-8 -*-|start:4292|stop:4792
ot max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) 
/ 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_l
engths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcPar
ams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', 
linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(Tr
ue)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.pl
ot(seq_len_distribut
>temp_340|source:-*- coding: utf-8 -*-|start:1355|stop:1855
astalib] Reading FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()sel
f.ids.append(self.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write
('\n')def close(self):self.fasta.close()class SequenceSource():def __init__(self
, fasta_file_path, lazy_init = True, unique = False, allow_mixed_case = False):s
elf.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init = lazy_inits
elf.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Non
eself.ids = []self.u
>temp_341|source:-*- coding: utf-8 -*-|start:1232|stop:1732
= SequenceSource(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0 or
 self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %
s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.seque
nces.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.clos
e()class SequenceSource():def __init__(self, fasta_file_path, lazy_init = True, 
unique = False, allow_mixed_case = False):self.fasta_file_path = fasta_file_path
self.name = Noneself
>temp_342|source:-*- coding: utf-8 -*-|start:5654|stop:6154
([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min:
 %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(
sequence_lengths),\min(sequence_lengths),\max(sequence_lengths)),\va = 'center',
 alpha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_pathtry:pl
t.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:pa
ssreturnclass QualSource:def __init__(self, quals_file_path, lazy_init = True):s
elf.quals_file_path 
>temp_343|source:-*- coding: utf-8 -*-|start:4973|stop:5473
('sequence length')if xtickstep == None:xtickstep = (max_seq_len / 50) or 1if yt
ickstep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(
xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(
range(0, max(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq_
len_distribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max
_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_dist
ribution) / 20.0))pl
>temp_344|source:-*- coding: utf-8 -*-|start:2207|stop:2707
ique_hash(self):while self.next_regular():hash = hashlib.sha1(self.seq.upper()).
hexdigest()if hash in self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].a
ppend(self.id)self.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_dic
t[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'count': 1}self.uniqu
e_hash_list = [i[1] for i in sorted([(self.unique_hash_dict[hash]['count'], hash
)\for hash in self.unique_hash_dict], reverse = True)]self.total_unique = len(se
lf.unique_hash_dict)
>temp_345|source:-*- coding: utf-8 -*-|start:6777|stop:7277
lit()]self.pos += 1return Trueelse:return Falseif line.startswith('>'):self.file
_pointer.seek(self.file_pointer.tell() - len(line))breakqualscores += ' ' + line
.strip()self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.qu
als.split()]self.pos += 1return Truedef close(self):self.file_pointer.close()def
 reset(self):self.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Nonese
lf.ids = []self.file_pointer.seek(0)if __name__ == '__main__':fasta = SequenceSo
urce(sys.argv[1])fas
>temp_346|source:-*- coding: utf-8 -*-|start:4756|stop:5256
ht = 0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.fill_b
etween(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black',
 alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence length')if x
tickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:ytickst
ep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len 
+ 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, max(seq_len_di
stribution) + 1, yti
>temp_347|source:-*- coding: utf-8 -*-|start:1936|stop:2436
t_hash = 0self.file_pointer = open(self.fasta_file_path)self.file_pointer.seek(0
)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for l in se
lf.file_pointer.readlines() if l.startswith('>')])self.reset()if self.unique:sel
f.init_unique_hash()def init_unique_hash(self):while self.next_regular():hash = 
hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.
unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'
] += 1else:self.uniq
>temp_348|source:-*- coding: utf-8 -*-|start:3159|stop:3659
Falsedef next_regular(self):self.seq = Noneself.id = self.file_pointer.readline(
)[1:].strip()sequence = ''while 1:line = self.file_pointer.readline()if not line
:if len(sequence):self.seq = sequenceself.pos += 1return Trueelse:return Falseif
 line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line
))breaksequence += line.strip()self.seq = sequence if self.allow_mixed_case else
 sequence.upper()self.pos += 1return Truedef get_seq_by_read_id(self, read_id):s
elf.reset()while sel
>temp_349|source:-*- coding: utf-8 -*-|start:1325|stop:1825
os == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fa
sta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(s
elf.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close()class Sequ
enceSource():def __init__(self, fasta_file_path, lazy_init = True, unique = Fals
e, allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.name = N
oneself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 
0self.id  = Noneself
>temp_350|source:-*- coding: utf-8 -*-|start:2701|stop:3201
_dict)self.reset()def next(self):if self.unique:return self.next_unique()else:re
turn self.next_regular()def next_unique(self):if self.unique:if self.total_uniqu
e > 0 and self.pos < self.total_unique:hash_entry = self.unique_hash_dict[self.u
nique_hash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.allo
w_mixed_case else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids =
 hash_entry['ids']return Trueelse:return Falseelse:return Falsedef next_regular(
self):self.seq = Non
>temp_351|source:-*- coding: utf-8 -*-|start:4656|stop:5156
x1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0
.03, top = 0.95, right = 0.98)plt.plot(seq_len_distribution, color = 'black', al
pha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distribution, y2 =
 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('s
equence length')if xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytick
step == None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xti
ckstep, max_seq_len 
>temp_352|source:-*- coding: utf-8 -*-|start:880|stop:1380
_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length = 80):ticks 
= range(0, len(sequence), piece_length) + [len(sequence)]return '\n'.join([seque
nce[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):se
lf.output_file_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = 
[]self.sequences = []self.fasta = SequenceSource(f_name)while self.fasta.next():
if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastal
ib] Reading FASTA in
>temp_353|source:-*- coding: utf-8 -*-|start:3321|stop:3821
f len(sequence):self.seq = sequenceself.pos += 1return Trueelse:return Falseif l
ine.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))
breaksequence += line.strip()self.seq = sequence if self.allow_mixed_case else s
equence.upper()self.pos += 1return Truedef get_seq_by_read_id(self, read_id):sel
f.reset()while self.next():if self.id == read_id:return self.seqreturn Falsedef 
close(self):self.file_pointer.close()def reset(self):self.pos = 0self.id  = None
self.seq = Noneself.
>temp_354|source:-*- coding: utf-8 -*-|start:3864|stop:4364
ize_sequence_length_distribution(self, title, dest = None, max_seq_len = None, x
tickstep = None, ytickstep = None):import matplotlib.pyplot as pltimport matplot
lib.gridspec as gridspecsequence_lengths = []self.reset()while self.next():if se
lf.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s'
 % (self.pos))sys.stderr.flush()sequence_lengths.append(len(self.seq))self.reset
()sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) +
 (int(max(sequence_l
>temp_355|source:-*- coding: utf-8 -*-|start:3124|stop:3624
n Trueelse:return Falseelse:return Falsedef next_regular(self):self.seq = Nonese
lf.id = self.file_pointer.readline()[1:].strip()sequence = ''while 1:line = self
.file_pointer.readline()if not line:if len(sequence):self.seq = sequenceself.pos
 += 1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.seek(
self.file_pointer.tell() - len(line))breaksequence += line.strip()self.seq = seq
uence if self.allow_mixed_case else sequence.upper()self.pos += 1return Truedef 
get_seq_by_read_id(s
>temp_356|source:-*- coding: utf-8 -*-|start:5720|stop:6220
d: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence_length
s), numpy.std(sequence_lengths),\min(sequence_lengths),\max(sequence_lengths)),\
va = 'center', alpha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_f
ile_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.s
how()except:passreturnclass QualSource:def __init__(self, quals_file_path, lazy_
init = True):self.quals_file_path = quals_file_pathself.name = Noneself.lazy_ini
t = lazy_initself.po
>temp_357|source:-*- coding: utf-8 -*-|start:3790|stop:4290
.id  = Noneself.seq = Noneself.ids = []self.file_pointer.seek(0)def visualize_se
quence_length_distribution(self, title, dest = None, max_seq_len = None, xtickst
ep = None, ytickstep = None):import matplotlib.pyplot as pltimport matplotlib.gr
idspec as gridspecsequence_lengths = []self.reset()while self.next():if self.pos
 % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (se
lf.pos))sys.stderr.flush()sequence_lengths.append(len(self.seq))self.reset()sys.
stderr.write('\n')if
>temp_358|source:-*- coding: utf-8 -*-|start:3723|stop:4223
ose(self):self.file_pointer.close()def reset(self):self.pos = 0self.id  = Nonese
lf.seq = Noneself.ids = []self.file_pointer.seek(0)def visualize_sequence_length
_distribution(self, title, dest = None, max_seq_len = None, xtickstep = None, yt
ickstep = None):import matplotlib.pyplot as pltimport matplotlib.gridspec as gri
dspecsequence_lengths = []self.reset()while self.next():if self.pos % 10000 == 0
 or self.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.s
tderr.flush()sequenc
>temp_359|source:-*- coding: utf-8 -*-|start:6080|stop:6580
ef __init__(self, quals_file_path, lazy_init = True):self.quals_file_path = qual
s_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = None
self.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer = open(self
.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_quals = N
oneelse:self.total_quals = len([l for l in self.file_pointer.readlines() if l.st
artswith('>')])self.reset()def next(self):self.id = self.file_pointer.readline()
[1:].strip()self.qua
>temp_360|source:-*- coding: utf-8 -*-|start:1871|stop:2371
lf.unique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0self.
file_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy_i
nit:self.total_seq = Noneelse:self.total_seq = len([l for l in self.file_pointer
.readlines() if l.startswith('>')])self.reset()if self.unique:self.init_unique_h
ash()def init_unique_hash(self):while self.next_regular():hash = hashlib.sha1(se
lf.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_hash_dic
t[hash]['ids'].appen
>temp_361|source:-*- coding: utf-8 -*-|start:6720|stop:7220
es.strip()self.quals_int = [int(q) for q in self.quals.split()]self.pos += 1retu
rn Trueelse:return Falseif line.startswith('>'):self.file_pointer.seek(self.file
_pointer.tell() - len(line))breakqualscores += ' ' + line.strip()self.quals = qu
alscores.strip()self.quals_int = [int(q) for q in self.quals.split()]self.pos +=
 1return Truedef close(self):self.file_pointer.close()def reset(self):self.pos =
 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_po
inter.seek(0)if __na
>temp_362|source:-*- coding: utf-8 -*-|start:6678|stop:7178
e:if len(qualscores):self.quals = qualscores.strip()self.quals_int = [int(q) for
 q in self.quals.split()]self.pos += 1return Trueelse:return Falseif line.starts
with('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))breakquals
cores += ' ' + line.strip()self.quals = qualscores.strip()self.quals_int = [int(
q) for q in self.quals.split()]self.pos += 1return Truedef close(self):self.file
_pointer.close()def reset(self):self.pos = 0self.id  = Noneself.quals = Noneself
.quals_int = Nonesel
>temp_363|source:-*- coding: utf-8 -*-|start:1865|stop:2365
iqueself.unique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 
0self.file_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if self.
lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for l in self.file_p
ointer.readlines() if l.startswith('>')])self.reset()if self.unique:self.init_un
ique_hash()def init_unique_hash(self):while self.next_regular():hash = hashlib.s
ha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_ha
sh_dict[hash]['ids']
>temp_364|source:-*- coding: utf-8 -*-|start:6672|stop:7172
ot line:if len(qualscores):self.quals = qualscores.strip()self.quals_int = [int(
q) for q in self.quals.split()]self.pos += 1return Trueelse:return Falseif line.
startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))brea
kqualscores += ' ' + line.strip()self.quals = qualscores.strip()self.quals_int =
 [int(q) for q in self.quals.split()]self.pos += 1return Truedef close(self):sel
f.file_pointer.close()def reset(self):self.pos = 0self.id  = Noneself.quals = No
neself.quals_int = N
>temp_365|source:-*- coding: utf-8 -*-|start:3588|stop:4088
1return Truedef get_seq_by_read_id(self, read_id):self.reset()while self.next():
if self.id == read_id:return self.seqreturn Falsedef close(self):self.file_point
er.close()def reset(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids = [
]self.file_pointer.seek(0)def visualize_sequence_length_distribution(self, title
, dest = None, max_seq_len = None, xtickstep = None, ytickstep = None):import ma
tplotlib.pyplot as pltimport matplotlib.gridspec as gridspecsequence_lengths = [
]self.reset()while s
>temp_366|source:-*- coding: utf-8 -*-|start:3454|stop:3954
pointer.tell() - len(line))breaksequence += line.strip()self.seq = sequence if s
elf.allow_mixed_case else sequence.upper()self.pos += 1return Truedef get_seq_by
_read_id(self, read_id):self.reset()while self.next():if self.id == read_id:retu
rn self.seqreturn Falsedef close(self):self.file_pointer.close()def reset(self):
self.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.file_pointer.seek(0)
def visualize_sequence_length_distribution(self, title, dest = None, max_seq_len
 = None, xtickstep =
>temp_367|source:-*- coding: utf-8 -*-|start:1662|stop:2162
se = False):self.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init
 = lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Nones
elf.seq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.un
ique_hash_list = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_
file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:s
elf.total_seq = len([l for l in self.file_pointer.readlines() if l.startswith('>
')])self.reset()if s
>temp_368|source:-*- coding: utf-8 -*-|start:3952|stop:4452
 = None, ytickstep = None):import matplotlib.pyplot as pltimport matplotlib.grid
spec as gridspecsequence_lengths = []self.reset()while self.next():if self.pos %
 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self
.pos))sys.stderr.flush()sequence_lengths.append(len(self.seq))self.reset()sys.st
derr.write('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(ma
x(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1
)for l in sequence_l
>temp_369|source:-*- coding: utf-8 -*-|start:529|stop:1029
, split = True, store_frequencies = True):if entry.unique and store_frequencies:
self.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.wr
ite_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, id):self.outp
ut_file_obj.write('>%s\n' % id)def write_seq(self, seq, split = True):if split:s
eq = self.split(seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequ
ence, piece_length = 80):ticks = range(0, len(sequence), piece_length) + [len(se
quence)]return '\n'.
>temp_370|source:-*- coding: utf-8 -*-|start:1934|stop:2434
ext_hash = 0self.file_pointer = open(self.fasta_file_path)self.file_pointer.seek
(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for l in 
self.file_pointer.readlines() if l.startswith('>')])self.reset()if self.unique:s
elf.init_unique_hash()def init_unique_hash(self):while self.next_regular():hash 
= hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:sel
f.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['coun
t'] += 1else:self.un
>temp_371|source:-*- coding: utf-8 -*-|start:1430|stop:1930
()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.
write('\n')def close(self):self.fasta.close()class SequenceSource():def __init__
(self, fasta_file_path, lazy_init = True, unique = False, allow_mixed_case = Fal
se):self.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init = lazy_
initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq 
= Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_has
h_list = []self.uniq
>temp_372|source:-*- coding: utf-8 -*-|start:4275|stop:4775
r.write('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(s
equence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)fo
r l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (
16, 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50'
, linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs
[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, ri
ght = 0.98)plt.plot(
>temp_373|source:-*- coding: utf-8 -*-|start:6108|stop:6608
_path, lazy_init = True):self.quals_file_path = quals_file_pathself.name = Nones
elf.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_
int = Noneself.ids = []self.file_pointer = open(self.quals_file_path)self.file_p
ointer.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = l
en([l for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()d
ef next(self):self.id = self.file_pointer.readline()[1:].strip()self.quals = Non
eself.quals_int = No
>temp_374|source:-*- coding: utf-8 -*-|start:1551|stop:2051
se()class SequenceSource():def __init__(self, fasta_file_path, lazy_init = True,
 unique = False, allow_mixed_case = False):self.fasta_file_path = fasta_file_pat
hself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_c
aseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = uniquese
lf.unique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0self.
file_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy_i
nit:self.total_seq =
>temp_375|source:-*- coding: utf-8 -*-|start:4888|stop:5388
on, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequences')plt.x
label('sequence length')if xtickstep == None:xtickstep = (max_seq_len / 50) or 1
if ytickstep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(r
ange(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yt
icks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max
(seq_len_distribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax 
= max_seq_len)plt.yl
>temp_376|source:-*- coding: utf-8 -*-|start:465|stop:965
tput_file_obj = open(output_file_path, 'w')def store(self, entry, split = True, 
store_frequencies = True):if entry.unique and store_frequencies:self.write_id('%
s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)
self.write_seq(entry.seq, split)def write_id(self, id):self.output_file_obj.writ
e('>%s\n' % id)def write_seq(self, seq, split = True):if split:seq = self.split(
seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequence, piece_leng
th = 80):ticks = ran
>temp_377|source:-*- coding: utf-8 -*-|start:1131|stop:1631
ile_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = []self.sequ
ences = []self.fasta = SequenceSource(f_name)while self.fasta.next():if self.fas
ta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading
 FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(sel
f.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(
self):self.fasta.close()class SequenceSource():def __init__(self, fasta_file_pat
h, lazy_init = True,
>temp_378|source:-*- coding: utf-8 -*-|start:2749|stop:3249
return self.next_unique()else:return self.next_regular()def next_unique(self):if
 self.unique:if self.total_unique > 0 and self.pos < self.total_unique:hash_entr
y = self.unique_hash_dict[self.unique_hash_list[self.pos]]self.pos += 1self.seq 
= hash_entry['seq'] if self.allow_mixed_case else hash_entry['seq'].upper()self.
id  = hash_entry['id']self.ids = hash_entry['ids']return Trueelse:return Falseel
se:return Falsedef next_regular(self):self.seq = Noneself.id = self.file_pointer
.readline()[1:].stri
>temp_379|source:-*- coding: utf-8 -*-|start:5536|stop:6036
xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edge
color' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'tot
al: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths),nu
mpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_lengths),\
max(sequence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == 
None:dest = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig
(dest + '.png')try:p
>temp_380|source:-*- coding: utf-8 -*-|start:2822|stop:3322
elf):if self.unique:if self.total_unique > 0 and self.pos < self.total_unique:ha
sh_entry = self.unique_hash_dict[self.unique_hash_list[self.pos]]self.pos += 1se
lf.seq = hash_entry['seq'] if self.allow_mixed_case else hash_entry['seq'].upper
()self.id  = hash_entry['id']self.ids = hash_entry['ids']return Trueelse:return 
Falseelse:return Falsedef next_regular(self):self.seq = Noneself.id = self.file_
pointer.readline()[1:].strip()sequence = ''while 1:line = self.file_pointer.read
line()if not line:if
>temp_381|source:-*- coding: utf-8 -*-|start:3663|stop:4163
xt():if self.id == read_id:return self.seqreturn Falsedef close(self):self.file_
pointer.close()def reset(self):self.pos = 0self.id  = Noneself.seq = Noneself.id
s = []self.file_pointer.seek(0)def visualize_sequence_length_distribution(self, 
title, dest = None, max_seq_len = None, xtickstep = None, ytickstep = None):impo
rt matplotlib.pyplot as pltimport matplotlib.gridspec as gridspecsequence_length
s = []self.reset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:sy
s.stderr.write('\r[f
>temp_382|source:-*- coding: utf-8 -*-|start:2075|stop:2575
 = len([l for l in self.file_pointer.readlines() if l.startswith('>')])self.rese
t()if self.unique:self.init_unique_hash()def init_unique_hash(self):while self.n
ext_regular():hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.u
nique_hash_dict:self.unique_hash_dict[hash]['ids'].append(self.id)self.unique_ha
sh_dict[hash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'i
ds': [self.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in
 sorted([(self.uniqu
>temp_383|source:-*- coding: utf-8 -*-|start:3524|stop:4024
uence if self.allow_mixed_case else sequence.upper()self.pos += 1return Truedef 
get_seq_by_read_id(self, read_id):self.reset()while self.next():if self.id == re
ad_id:return self.seqreturn Falsedef close(self):self.file_pointer.close()def re
set(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.file_point
er.seek(0)def visualize_sequence_length_distribution(self, title, dest = None, m
ax_seq_len = None, xtickstep = None, ytickstep = None):import matplotlib.pyplot 
as pltimport matplot
>temp_384|source:-*- coding: utf-8 -*-|start:4832|stop:5332
ll_between(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'bla
ck', alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence length')
if xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:yti
ckstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_
len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, max(seq_le
n_distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_distribution) +
 1, ytickstep)],size
>temp_385|source:-*- coding: utf-8 -*-|start:6551|stop:7051
eadline()[1:].strip()self.quals = Noneself.quals_int = Nonequalscores = ''while 
1:line = self.file_pointer.readline()if not line:if len(qualscores):self.quals =
 qualscores.strip()self.quals_int = [int(q) for q in self.quals.split()]self.pos
 += 1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.seek(
self.file_pointer.tell() - len(line))breakqualscores += ' ' + line.strip()self.q
uals = qualscores.strip()self.quals_int = [int(q) for q in self.quals.split()]se
lf.pos += 1return Tr
>temp_386|source:-*- coding: utf-8 -*-|start:1868|stop:2368
eself.unique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0se
lf.file_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if self.laz
y_init:self.total_seq = Noneelse:self.total_seq = len([l for l in self.file_poin
ter.readlines() if l.startswith('>')])self.reset()if self.unique:self.init_uniqu
e_hash()def init_unique_hash(self):while self.next_regular():hash = hashlib.sha1
(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_hash_
dict[hash]['ids'].ap
>temp_387|source:-*- coding: utf-8 -*-|start:6475|stop:6975
 l.startswith('>')])self.reset()def next(self):self.id = self.file_pointer.readl
ine()[1:].strip()self.quals = Noneself.quals_int = Nonequalscores = ''while 1:li
ne = self.file_pointer.readline()if not line:if len(qualscores):self.quals = qua
lscores.strip()self.quals_int = [int(q) for q in self.quals.split()]self.pos += 
1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.seek(self
.file_pointer.tell() - len(line))breakqualscores += ' ' + line.strip()self.quals
 = qualscores.strip(
>temp_388|source:-*- coding: utf-8 -*-|start:5894|stop:6394
 alpha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_pathtry:pl
t.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:pa
ssreturnclass QualSource:def __init__(self, quals_file_path, lazy_init = True):s
elf.quals_file_path = quals_file_pathself.name = Noneself.lazy_init = lazy_inits
elf.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []sel
f.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if self.lazy
_init:self.total_qua
>temp_389|source:-*- coding: utf-8 -*-|start:1423|stop:1923
r.flush()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.seq)sys.
stderr.write('\n')def close(self):self.fasta.close()class SequenceSource():def _
_init__(self, fasta_file_path, lazy_init = True, unique = False, allow_mixed_cas
e = False):self.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init 
= lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Nonese
lf.seq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.uni
que_hash_list = []se
>temp_390|source:-*- coding: utf-8 -*-|start:3661|stop:4161
next():if self.id == read_id:return self.seqreturn Falsedef close(self):self.fil
e_pointer.close()def reset(self):self.pos = 0self.id  = Noneself.seq = Noneself.
ids = []self.file_pointer.seek(0)def visualize_sequence_length_distribution(self
, title, dest = None, max_seq_len = None, xtickstep = None, ytickstep = None):im
port matplotlib.pyplot as pltimport matplotlib.gridspec as gridspecsequence_leng
ths = []self.reset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:
sys.stderr.write('\r
>temp_391|source:-*- coding: utf-8 -*-|start:593|stop:1093
ore_frequencies:self.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.id
s)))else:self.write_id(entry.id)self.write_seq(entry.seq, split)def write_id(sel
f, id):self.output_file_obj.write('>%s\n' % id)def write_seq(self, seq, split = 
True):if split:seq = self.split(seq)self.output_file_obj.write('%s\n' % seq)def 
split(self, sequence, piece_length = 80):ticks = range(0, len(sequence), piece_l
ength) + [len(sequence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x 
in range(0, len(tick
>temp_392|source:-*- coding: utf-8 -*-|start:296|stop:796
ase read the docs/COPYING file.import sysimport numpyimport hashlibclass FastaOu
tput:def __init__(self, output_file_path):self.output_file_path = output_file_pa
thself.output_file_obj = open(output_file_path, 'w')def store(self, entry, split
 = True, store_frequencies = True):if entry.unique and store_frequencies:self.wr
ite_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(
entry.id)self.write_seq(entry.seq, split)def write_id(self, id):self.output_file
_obj.write('>%s\n' %
>temp_393|source:-*- coding: utf-8 -*-|start:751|stop:1251
elf, id):self.output_file_obj.write('>%s\n' % id)def write_seq(self, seq, split 
= True):if split:seq = self.split(seq)self.output_file_obj.write('%s\n' % seq)de
f split(self, sequence, piece_length = 80):ticks = range(0, len(sequence), piece
_length) + [len(sequence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for 
x in range(0, len(ticks) - 1)])def close(self):self.output_file_obj.close()class
 ReadFasta:def __init__(self, f_name):self.ids = []self.sequences = []self.fasta
 = SequenceSource(f_
>temp_394|source:-*- coding: utf-8 -*-|start:1477|stop:1977
append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close()cl
ass SequenceSource():def __init__(self, fasta_file_path, lazy_init = True, uniqu
e = False, allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.
name = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_casesel
f.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.uni
que_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0self.file_p
ointer = open(self.f
>temp_395|source:-*- coding: utf-8 -*-|start:4437|stop:4937
l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16
, 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', 
linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0
:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, righ
t = 0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.fill_be
tween(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black', 
alpha = 0.15)plt.yla
>temp_396|source:-*- coding: utf-8 -*-|start:140|stop:640
the GNU General Public License as published by the Free# Software Foundation; ei
ther version 2 of the License, or (at your option)# any later version.## Please 
read the docs/COPYING file.import sysimport numpyimport hashlibclass FastaOutput
:def __init__(self, output_file_path):self.output_file_path = output_file_pathse
lf.output_file_obj = open(output_file_path, 'w')def store(self, entry, split = T
rue, store_frequencies = True):if entry.unique and store_frequencies:self.write_
id('%s|%s' % (entry.
>temp_397|source:-*- coding: utf-8 -*-|start:4270|stop:4770
stderr.write('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(
max(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len +
 1)for l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsiz
e = (16, 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='
0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subpl
ot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.9
5, right = 0.98)plt.
>temp_398|source:-*- coding: utf-8 -*-|start:356|stop:856
hashlibclass FastaOutput:def __init__(self, output_file_path):self.output_file_p
ath = output_file_pathself.output_file_obj = open(output_file_path, 'w')def stor
e(self, entry, split = True, store_frequencies = True):if entry.unique and store
_frequencies:self.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids))
)else:self.write_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, 
id):self.output_file_obj.write('>%s\n' % id)def write_seq(self, seq, split = Tru
e):if split:seq = se
>temp_399|source:-*- coding: utf-8 -*-|start:1185|stop:1685
me):self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)while sel
f.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.w
rite('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))sys.stderr.
flush()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.seq)sys.st
derr.write('\n')def close(self):self.fasta.close()class SequenceSource():def __i
nit__(self, fasta_file_path, lazy_init = True, unique = False, allow_mixed_case 
= False):self.fasta_
>temp_400|source:-*- coding: utf-8 -*-|start:5783|stop:6283
n(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_lengths),\max(seq
uence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == None:de
st = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest +
 '.png')try:plt.show()except:passreturnclass QualSource:def __init__(self, quals
_file_path, lazy_init = True):self.quals_file_path = quals_file_pathself.name = 
Noneself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.q
uals_int = Noneself.
>temp_401|source:-*- coding: utf-8 -*-|start:2344|stop:2844
ash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] += 1el
se:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.se
q,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.unique_hash_di
ct[hash]['count'], hash)\for hash in self.unique_hash_dict], reverse = True)]sel
f.total_unique = len(self.unique_hash_dict)self.reset()def next(self):if self.un
ique:return self.next_unique()else:return self.next_regular()def next_unique(sel
f):if self.unique:if
>temp_402|source:-*- coding: utf-8 -*-|start:3528|stop:4028
e if self.allow_mixed_case else sequence.upper()self.pos += 1return Truedef get_
seq_by_read_id(self, read_id):self.reset()while self.next():if self.id == read_i
d:return self.seqreturn Falsedef close(self):self.file_pointer.close()def reset(
self):self.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.file_pointer.s
eek(0)def visualize_sequence_length_distribution(self, title, dest = None, max_s
eq_len = None, xtickstep = None, ytickstep = None):import matplotlib.pyplot as p
ltimport matplotlib.
>temp_403|source:-*- coding: utf-8 -*-|start:2754|stop:3254
n self.next_unique()else:return self.next_regular()def next_unique(self):if self
.unique:if self.total_unique > 0 and self.pos < self.total_unique:hash_entry = s
elf.unique_hash_dict[self.unique_hash_list[self.pos]]self.pos += 1self.seq = has
h_entry['seq'] if self.allow_mixed_case else hash_entry['seq'].upper()self.id  =
 hash_entry['id']self.ids = hash_entry['ids']return Trueelse:return Falseelse:re
turn Falsedef next_regular(self):self.seq = Noneself.id = self.file_pointer.read
line()[1:].strip()se
>temp_404|source:-*- coding: utf-8 -*-|start:3932|stop:4432
en = None, xtickstep = None, ytickstep = None):import matplotlib.pyplot as pltim
port matplotlib.gridspec as gridspecsequence_lengths = []self.reset()while self.
next():if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fastalib] 
Reading: %s' % (self.pos))sys.stderr.flush()sequence_lengths.append(len(self.seq
))self.reset()sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(sequenc
e_lengths) + (int(max(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0
] * (max_seq_len + 1
>temp_405|source:-*- coding: utf-8 -*-|start:1758|stop:2258
.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Nonese
lf.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list =
 []self.unique_next_hash = 0self.file_pointer = open(self.fasta_file_path)self.f
ile_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = 
len([l for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()
if self.unique:self.init_unique_hash()def init_unique_hash(self):while self.next
_regular():hash = ha
>temp_406|source:-*- coding: utf-8 -*-|start:1974|stop:2474
f.fasta_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = No
neelse:self.total_seq = len([l for l in self.file_pointer.readlines() if l.start
swith('>')])self.reset()if self.unique:self.init_unique_hash()def init_unique_ha
sh(self):while self.next_regular():hash = hashlib.sha1(self.seq.upper()).hexdige
st()if hash in self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].append(s
elf.id)self.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_dict[hash]
 = {'id' : self.id,'
>temp_407|source:-*- coding: utf-8 -*-|start:5259|stop:5759
tep),[y for y in range(0, max(seq_len_distribution) + 1, ytickstep)],size='xx-sm
all')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len
_distribution) + (max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' 
% (title), weight = 'black', size = 'xx-large', ha = 'center')ax1 = plt.subplot(
gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])p
lt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s /
 max: %s'\% (len(seq
>temp_408|source:-*- coding: utf-8 -*-|start:5487|stop:5987
 0.96, '%s' % (title), weight = 'black', size = 'xx-large', ha = 'center')ax1 = 
plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)plt
.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f
 / min: %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence_lengths), num
py.std(sequence_lengths),\min(sequence_lengths),\max(sequence_lengths)),\va = 'c
enter', alpha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_pat
htry:plt.savefig(des
>temp_409|source:-*- coding: utf-8 -*-|start:674|stop:1174
)))else:self.write_id(entry.id)self.write_seq(entry.seq, split)def write_id(self
, id):self.output_file_obj.write('>%s\n' % id)def write_seq(self, seq, split = T
rue):if split:seq = self.split(seq)self.output_file_obj.write('%s\n' % seq)def s
plit(self, sequence, piece_length = 80):ticks = range(0, len(sequence), piece_le
ngth) + [len(sequence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x i
n range(0, len(ticks) - 1)])def close(self):self.output_file_obj.close()class Re
adFasta:def __init__
>temp_410|source:-*- coding: utf-8 -*-|start:5827|stop:6327
ths),\min(sequence_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8,
 size = 'x-large')if dest == None:dest = self.fasta_file_pathtry:plt.savefig(des
t + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passreturnclass
 QualSource:def __init__(self, quals_file_path, lazy_init = True):self.quals_fil
e_path = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0se
lf.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_pointe
r = open(self.quals_
>temp_411|source:-*- coding: utf-8 -*-|start:560|stop:1060
es = True):if entry.unique and store_frequencies:self.write_id('%s|%s' % (entry.
id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_seq(
entry.seq, split)def write_id(self, id):self.output_file_obj.write('>%s\n' % id)
def write_seq(self, seq, split = True):if split:seq = self.split(seq)self.output
_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length = 80):ticks 
= range(0, len(sequence), piece_length) + [len(sequence)]return '\n'.join([seque
nce[ticks[x]:ticks[x
>temp_412|source:-*- coding: utf-8 -*-|start:1370|stop:1870
g FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(se
lf.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close
(self):self.fasta.close()class SequenceSource():def __init__(self, fasta_file_pa
th, lazy_init = True, unique = False, allow_mixed_case = False):self.fasta_file_
path = fasta_file_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed
_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []s
elf.unique = uniques
>temp_413|source:-*- coding: utf-8 -*-|start:6780|stop:7280
()]self.pos += 1return Trueelse:return Falseif line.startswith('>'):self.file_po
inter.seek(self.file_pointer.tell() - len(line))breakqualscores += ' ' + line.st
rip()self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.quals
.split()]self.pos += 1return Truedef close(self):self.file_pointer.close()def re
set(self):self.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.
ids = []self.file_pointer.seek(0)if __name__ == '__main__':fasta = SequenceSourc
e(sys.argv[1])fasta.
>temp_414|source:-*- coding: utf-8 -*-|start:1924|stop:2424
f.unique_next_hash = 0self.file_pointer = open(self.fasta_file_path)self.file_po
inter.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len([l
 for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()if sel
f.unique:self.init_unique_hash()def init_unique_hash(self):while self.next_regul
ar():hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_has
h_dict:self.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[h
ash]['count'] += 1el
>temp_415|source:-*- coding: utf-8 -*-|start:4302|stop:4802
_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) o
r 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq
_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update
({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewidth=
0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.sub
plots_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len
_distribution, color
>temp_416|source:-*- coding: utf-8 -*-|start:832|stop:1332
 True):if split:seq = self.split(seq)self.output_file_obj.write('%s\n' % seq)def
 split(self, sequence, piece_length = 80):ticks = range(0, len(sequence), piece_
length) + [len(sequence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x
 in range(0, len(ticks) - 1)])def close(self):self.output_file_obj.close()class 
ReadFasta:def __init__(self, f_name):self.ids = []self.sequences = []self.fasta 
= SequenceSource(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0 or
 self.fasta.pos == 1
>temp_417|source:-*- coding: utf-8 -*-|start:2791|stop:3291
next_regular()def next_unique(self):if self.unique:if self.total_unique > 0 and 
self.pos < self.total_unique:hash_entry = self.unique_hash_dict[self.unique_hash
_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_ca
se else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_entr
y['ids']return Trueelse:return Falseelse:return Falsedef next_regular(self):self
.seq = Noneself.id = self.file_pointer.readline()[1:].strip()sequence = ''while 
1:line = self.file_p
>temp_418|source:-*- coding: utf-8 -*-|start:4728|stop:5228
ttom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_distribution, color = 'bl
ack', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distributi
on, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequences')plt.x
label('sequence length')if xtickstep == None:xtickstep = (max_seq_len / 50) or 1
if ytickstep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(r
ange(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yt
icks(range(0, max(se
>temp_419|source:-*- coding: utf-8 -*-|start:2071|stop:2571
_seq = len([l for l in self.file_pointer.readlines() if l.startswith('>')])self.
reset()if self.unique:self.init_unique_hash()def init_unique_hash(self):while se
lf.next_regular():hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash in se
lf.unique_hash_dict:self.unique_hash_dict[hash]['ids'].append(self.id)self.uniqu
e_hash_dict[hash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.i
d,'ids': [self.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for 
i in sorted([(self.u
>temp_420|source:-*- coding: utf-8 -*-|start:5342|stop:5842
')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_di
stribution) + (max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (
title), weight = 'black', size = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[
9])plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])plt.
xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s / ma
x: %s'\% (len(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(sequence
_lengths),\min(seque
>temp_421|source:-*- coding: utf-8 -*-|start:3472|stop:3972
en(line))breaksequence += line.strip()self.seq = sequence if self.allow_mixed_ca
se else sequence.upper()self.pos += 1return Truedef get_seq_by_read_id(self, rea
d_id):self.reset()while self.next():if self.id == read_id:return self.seqreturn 
Falsedef close(self):self.file_pointer.close()def reset(self):self.pos = 0self.i
d  = Noneself.seq = Noneself.ids = []self.file_pointer.seek(0)def visualize_sequ
ence_length_distribution(self, title, dest = None, max_seq_len = None, xtickstep
 = None, ytickstep =
>temp_422|source:-*- coding: utf-8 -*-|start:638|stop:1138
y.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_se
q(entry.seq, split)def write_id(self, id):self.output_file_obj.write('>%s\n' % i
d)def write_seq(self, seq, split = True):if split:seq = self.split(seq)self.outp
ut_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length = 80):tick
s = range(0, len(sequence), piece_length) + [len(sequence)]return '\n'.join([seq
uence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):
self.output_file_obj
>temp_423|source:-*- coding: utf-8 -*-|start:2660|stop:3160
]self.total_unique = len(self.unique_hash_dict)self.reset()def next(self):if sel
f.unique:return self.next_unique()else:return self.next_regular()def next_unique
(self):if self.unique:if self.total_unique > 0 and self.pos < self.total_unique:
hash_entry = self.unique_hash_dict[self.unique_hash_list[self.pos]]self.pos += 1
self.seq = hash_entry['seq'] if self.allow_mixed_case else hash_entry['seq'].upp
er()self.id  = hash_entry['id']self.ids = hash_entry['ids']return Trueelse:retur
n Falseelse:return F
>temp_424|source:-*- coding: utf-8 -*-|start:309|stop:809
docs/COPYING file.import sysimport numpyimport hashlibclass FastaOutput:def __in
it__(self, output_file_path):self.output_file_path = output_file_pathself.output
_file_obj = open(output_file_path, 'w')def store(self, entry, split = True, stor
e_frequencies = True):if entry.unique and store_frequencies:self.write_id('%s|%s
' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)self
.write_seq(entry.seq, split)def write_id(self, id):self.output_file_obj.write('>
%s\n' % id)def write
>temp_425|source:-*- coding: utf-8 -*-|start:221|stop:721
her version 2 of the License, or (at your option)# any later version.## Please r
ead the docs/COPYING file.import sysimport numpyimport hashlibclass FastaOutput:
def __init__(self, output_file_path):self.output_file_path = output_file_pathsel
f.output_file_obj = open(output_file_path, 'w')def store(self, entry, split = Tr
ue, store_frequencies = True):if entry.unique and store_frequencies:self.write_i
d('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry
.id)self.write_seq(e
>temp_426|source:-*- coding: utf-8 -*-|start:3682|stop:4182
read_id:return self.seqreturn Falsedef close(self):self.file_pointer.close()def 
reset(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.file_poi
nter.seek(0)def visualize_sequence_length_distribution(self, title, dest = None,
 max_seq_len = None, xtickstep = None, ytickstep = None):import matplotlib.pyplo
t as pltimport matplotlib.gridspec as gridspecsequence_lengths = []self.reset()w
hile self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[
fastalib] Reading: %
>temp_427|source:-*- coding: utf-8 -*-|start:4128|stop:4628
elf.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr
.flush()sequence_lengths.append(len(self.seq))self.reset()sys.stderr.write('\n')
if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengt
hs) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequen
ce_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.r
cParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='
-', linewidth=0.1)gs
>temp_428|source:-*- coding: utf-8 -*-|start:6201|stop:6701
 = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Nonesel
f.ids = []self.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0
)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l for l i
n self.file_pointer.readlines() if l.startswith('>')])self.reset()def next(self)
:self.id = self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_i
nt = Nonequalscores = ''while 1:line = self.file_pointer.readline()if not line:i
f len(qualscores):se
>temp_429|source:-*- coding: utf-8 -*-|start:12|stop:512
Copyright (C) 2014, A. Murat Eren## This program is free software; you can redis
tribute it and/or modify it under# the terms of the GNU General Public License a
s published by the Free# Software Foundation; either version 2 of the License, o
r (at your option)# any later version.## Please read the docs/COPYING file.impor
t sysimport numpyimport hashlibclass FastaOutput:def __init__(self, output_file_
path):self.output_file_path = output_file_pathself.output_file_obj = open(output
_file_path, 'w')def 
>temp_430|source:-*- coding: utf-8 -*-|start:4683|stop:5183
t.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.
98)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.fill_between(
range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black', alpha 
= 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence length')if xtickste
p == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:ytickstep = ma
x(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1, xt
ickstep), rotation=9
>temp_431|source:-*- coding: utf-8 -*-|start:345|stop:845
umpyimport hashlibclass FastaOutput:def __init__(self, output_file_path):self.ou
tput_file_path = output_file_pathself.output_file_obj = open(output_file_path, '
w')def store(self, entry, split = True, store_frequencies = True):if entry.uniqu
e and store_frequencies:self.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(
entry.ids)))else:self.write_id(entry.id)self.write_seq(entry.seq, split)def writ
e_id(self, id):self.output_file_obj.write('>%s\n' % id)def write_seq(self, seq, 
split = True):if spl
>temp_432|source:-*- coding: utf-8 -*-|start:5874|stop:6374
hs)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == None:dest = self.f
asta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try
:plt.show()except:passreturnclass QualSource:def __init__(self, quals_file_path,
 lazy_init = True):self.quals_file_path = quals_file_pathself.name = Noneself.la
zy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = 
Noneself.ids = []self.file_pointer = open(self.quals_file_path)self.file_pointer
.seek(0)if self.lazy
>temp_433|source:-*- coding: utf-8 -*-|start:2274|stop:2774
seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_hash_dict[h
ash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] += 1else:self.un
ique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'count':
 1}self.unique_hash_list = [i[1] for i in sorted([(self.unique_hash_dict[hash]['
count'], hash)\for hash in self.unique_hash_dict], reverse = True)]self.total_un
ique = len(self.unique_hash_dict)self.reset()def next(self):if self.unique:retur
n self.next_unique()
>temp_434|source:-*- coding: utf-8 -*-|start:3290|stop:3790
pointer.readline()if not line:if len(sequence):self.seq = sequenceself.pos += 1r
eturn Trueelse:return Falseif line.startswith('>'):self.file_pointer.seek(self.f
ile_pointer.tell() - len(line))breaksequence += line.strip()self.seq = sequence 
if self.allow_mixed_case else sequence.upper()self.pos += 1return Truedef get_se
q_by_read_id(self, read_id):self.reset()while self.next():if self.id == read_id:
return self.seqreturn Falsedef close(self):self.file_pointer.close()def reset(se
lf):self.pos = 0self
>temp_435|source:-*- coding: utf-8 -*-|start:1507|stop:2007
rr.write('\n')def close(self):self.fasta.close()class SequenceSource():def __ini
t__(self, fasta_file_path, lazy_init = True, unique = False, allow_mixed_case = 
False):self.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init = la
zy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.s
eq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_
hash_list = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_file_
path)self.file_point
>temp_436|source:-*- coding: utf-8 -*-|start:4565|stop:5065
9})plt.rc('grid', color='0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridS
pec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05,
 bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_distribution, color = 
'black', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distrib
ution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequences')pl
t.xlabel('sequence length')if xtickstep == None:xtickstep = (max_seq_len / 50) o
r 1if ytickstep == N
>temp_437|source:-*- coding: utf-8 -*-|start:5704|stop:6204
 mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.mean
(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_lengths),\max(sequ
ence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == None:des
t = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + 
'.png')try:plt.show()except:passreturnclass QualSource:def __init__(self, quals_
file_path, lazy_init = True):self.quals_file_path = quals_file_pathself.name = N
oneself.lazy_init = 
>temp_438|source:-*- coding: utf-8 -*-|start:1104|stop:1604
f close(self):self.output_file_obj.close()class ReadFasta:def __init__(self, f_n
ame):self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)while se
lf.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.
write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))sys.stderr
.flush()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.seq)sys.s
tderr.write('\n')def close(self):self.fasta.close()class SequenceSource():def __
init__(self, fasta_f
>temp_439|source:-*- coding: utf-8 -*-|start:1234|stop:1734
SequenceSource(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0 or s
elf.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s'
 % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequenc
es.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close(
)class SequenceSource():def __init__(self, fasta_file_path, lazy_init = True, un
ique = False, allow_mixed_case = False):self.fasta_file_path = fasta_file_pathse
lf.name = Noneself.l
>temp_440|source:-*- coding: utf-8 -*-|start:5521|stop:6021
lack', size = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.upd
ate({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(
0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(seque
nce_lengths),numpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min(sequ
ence_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, size = 'x-lar
ge')if dest == None:dest = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')exc
ept:plt.savefig(dest
>temp_441|source:-*- coding: utf-8 -*-|start:1539|stop:2039
lf.fasta.close()class SequenceSource():def __init__(self, fasta_file_path, lazy_
init = True, unique = False, allow_mixed_case = False):self.fasta_file_path = fa
sta_file_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = a
llow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.uniqu
e = uniqueself.unique_hash_dict = {}self.unique_hash_list = []self.unique_next_h
ash = 0self.file_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if
 self.lazy_init:self
>temp_442|source:-*- coding: utf-8 -*-|start:5050|stop:5550
 ytickstep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(ran
ge(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.ytic
ks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(s
eq_len_distribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = 
max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_d
istribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', si
ze = 'xx-large', ha 
>temp_443|source:-*- coding: utf-8 -*-|start:1638|stop:2138
 = False, allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.n
ame = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself
.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.uniq
ue_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0self.file_po
inter = open(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy_init:sel
f.total_seq = Noneelse:self.total_seq = len([l for l in self.file_pointer.readli
nes() if l.startswit
>temp_444|source:-*- coding: utf-8 -*-|start:568|stop:1068
e):if entry.unique and store_frequencies:self.write_id('%s|%s' % (entry.id, 'fre
quency:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_seq(entry.se
q, split)def write_id(self, id):self.output_file_obj.write('>%s\n' % id)def writ
e_seq(self, seq, split = True):if split:seq = self.split(seq)self.output_file_ob
j.write('%s\n' % seq)def split(self, sequence, piece_length = 80):ticks = range(
0, len(sequence), piece_length) + [len(sequence)]return '\n'.join([sequence[tick
s[x]:ticks[x + 1]] f
>temp_445|source:-*- coding: utf-8 -*-|start:2951|stop:3451
_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_ca
se else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_entr
y['ids']return Trueelse:return Falseelse:return Falsedef next_regular(self):self
.seq = Noneself.id = self.file_pointer.readline()[1:].strip()sequence = ''while 
1:line = self.file_pointer.readline()if not line:if len(sequence):self.seq = seq
uenceself.pos += 1return Trueelse:return Falseif line.startswith('>'):self.file_
pointer.seek(self.fi
>temp_446|source:-*- coding: utf-8 -*-|start:4873|stop:5373
_len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of s
equences')plt.xlabel('sequence length')if xtickstep == None:xtickstep = (max_seq
_len / 50) or 1if ytickstep == None:ytickstep = max(seq_len_distribution) / 20 o
r 1plt.xticks(range(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='x
x-small')plt.yticks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y 
in range(0, max(seq_len_distribution) + 1, ytickstep)],size='xx-small')plt.xlim(
xmin = 0, xmax = max
>temp_447|source:-*- coding: utf-8 -*-|start:974|stop:1474
(sequence), piece_length) + [len(sequence)]return '\n'.join([sequence[ticks[x]:t
icks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.output_file
_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = []self.sequenc
es = []self.fasta = SequenceSource(f_name)while self.fasta.next():if self.fasta.
pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FA
STA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.f
asta.id)self.sequenc
>temp_448|source:-*- coding: utf-8 -*-|start:4624|stop:5124
1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subpl
ots_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_d
istribution, color = 'black', alpha = 0.3)plt.fill_between(range(0, max_seq_len 
+ 1), seq_len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('nu
mber of sequences')plt.xlabel('sequence length')if xtickstep == None:xtickstep =
 (max_seq_len / 50) or 1if ytickstep == None:ytickstep = max(seq_len_distributio
n) / 20 or 1plt.xtic
>temp_449|source:-*- coding: utf-8 -*-|start:2753|stop:3253
rn self.next_unique()else:return self.next_regular()def next_unique(self):if sel
f.unique:if self.total_unique > 0 and self.pos < self.total_unique:hash_entry = 
self.unique_hash_dict[self.unique_hash_list[self.pos]]self.pos += 1self.seq = ha
sh_entry['seq'] if self.allow_mixed_case else hash_entry['seq'].upper()self.id  
= hash_entry['id']self.ids = hash_entry['ids']return Trueelse:return Falseelse:r
eturn Falsedef next_regular(self):self.seq = Noneself.id = self.file_pointer.rea
dline()[1:].strip()s
>temp_450|source:-*- coding: utf-8 -*-|start:463|stop:963
output_file_obj = open(output_file_path, 'w')def store(self, entry, split = True
, store_frequencies = True):if entry.unique and store_frequencies:self.write_id(
'%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.i
d)self.write_seq(entry.seq, split)def write_id(self, id):self.output_file_obj.wr
ite('>%s\n' % id)def write_seq(self, seq, split = True):if split:seq = self.spli
t(seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequence, piece_le
ngth = 80):ticks = r
>temp_451|source:-*- coding: utf-8 -*-|start:874|stop:1374
output_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length = 80):
ticks = range(0, len(sequence), piece_length) + [len(sequence)]return '\n'.join(
[sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(se
lf):self.output_file_obj.close()class ReadFasta:def __init__(self, f_name):self.
ids = []self.sequences = []self.fasta = SequenceSource(f_name)while self.fasta.n
ext():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[
fastalib] Reading FA
>temp_452|source:-*- coding: utf-8 -*-|start:1744|stop:2244
 lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Nonesel
f.seq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.uniq
ue_hash_list = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_fi
le_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:sel
f.total_seq = len([l for l in self.file_pointer.readlines() if l.startswith('>')
])self.reset()if self.unique:self.init_unique_hash()def init_unique_hash(self):w
hile self.next_regul
>temp_453|source:-*- coding: utf-8 -*-|start:1849|stop:2349
self.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list = []self.uni
que_next_hash = 0self.file_pointer = open(self.fasta_file_path)self.file_pointer
.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for 
l in self.file_pointer.readlines() if l.startswith('>')])self.reset()if self.uni
que:self.init_unique_hash()def init_unique_hash(self):while self.next_regular():
hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dic
t:self.unique_hash_d
>temp_454|source:-*- coding: utf-8 -*-|start:4815|stop:5315
lpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distribution, y2 
= 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('
sequence length')if xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytic
kstep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xt
ickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(ra
nge(0, max(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq_le
n_distribution) + 1,
>temp_455|source:-*- coding: utf-8 -*-|start:3243|stop:3743
].strip()sequence = ''while 1:line = self.file_pointer.readline()if not line:if 
len(sequence):self.seq = sequenceself.pos += 1return Trueelse:return Falseif lin
e.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))br
eaksequence += line.strip()self.seq = sequence if self.allow_mixed_case else seq
uence.upper()self.pos += 1return Truedef get_seq_by_read_id(self, read_id):self.
reset()while self.next():if self.id == read_id:return self.seqreturn Falsedef cl
ose(self):self.file_
>temp_456|source:-*- coding: utf-8 -*-|start:2104|stop:2604
pointer.readlines() if l.startswith('>')])self.reset()if self.unique:self.init_u
nique_hash()def init_unique_hash(self):while self.next_regular():hash = hashlib.
sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_h
ash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] += 1el
se:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.se
q,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.unique_hash_di
ct[hash]['count'], h
>temp_457|source:-*- coding: utf-8 -*-|start:6770|stop:7270
uals.split()]self.pos += 1return Trueelse:return Falseif line.startswith('>'):se
lf.file_pointer.seek(self.file_pointer.tell() - len(line))breakqualscores += ' '
 + line.strip()self.quals = qualscores.strip()self.quals_int = [int(q) for q in 
self.quals.split()]self.pos += 1return Truedef close(self):self.file_pointer.clo
se()def reset(self):self.pos = 0self.id  = Noneself.quals = Noneself.quals_int =
 Noneself.ids = []self.file_pointer.seek(0)if __name__ == '__main__':fasta = Seq
uenceSource(sys.argv
>temp_458|source:-*- coding: utf-8 -*-|start:4808|stop:5308
ack', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distributi
on, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequences')plt.x
label('sequence length')if xtickstep == None:xtickstep = (max_seq_len / 50) or 1
if ytickstep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(r
ange(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yt
icks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max
(seq_len_distributio
>temp_459|source:-*- coding: utf-8 -*-|start:4356|stop:4856
quence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for
 l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (1
6, 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50',
 linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[
0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, rig
ht = 0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.fill_b
etween(range(0, max_
>temp_460|source:-*- coding: utf-8 -*-|start:5293|stop:5793
len_distribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max
_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_dist
ribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size 
= 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.e
dgecolor' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, '
total: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths)
,numpy.mean(sequence
>temp_461|source:-*- coding: utf-8 -*-|start:3543|stop:4043
_mixed_case else sequence.upper()self.pos += 1return Truedef get_seq_by_read_id(
self, read_id):self.reset()while self.next():if self.id == read_id:return self.s
eqreturn Falsedef close(self):self.file_pointer.close()def reset(self):self.pos 
= 0self.id  = Noneself.seq = Noneself.ids = []self.file_pointer.seek(0)def visua
lize_sequence_length_distribution(self, title, dest = None, max_seq_len = None, 
xtickstep = None, ytickstep = None):import matplotlib.pyplot as pltimport matplo
tlib.gridspec as gri
>temp_462|source:-*- coding: utf-8 -*-|start:6636|stop:7136
e = self.file_pointer.readline()if not line:if len(qualscores):self.quals = qual
scores.strip()self.quals_int = [int(q) for q in self.quals.split()]self.pos += 1
return Trueelse:return Falseif line.startswith('>'):self.file_pointer.seek(self.
file_pointer.tell() - len(line))breakqualscores += ' ' + line.strip()self.quals 
= qualscores.strip()self.quals_int = [int(q) for q in self.quals.split()]self.po
s += 1return Truedef close(self):self.file_pointer.close()def reset(self):self.p
os = 0self.id  = Non
>temp_463|source:-*- coding: utf-8 -*-|start:1472|stop:1972
nces.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.clos
e()class SequenceSource():def __init__(self, fasta_file_path, lazy_init = True, 
unique = False, allow_mixed_case = False):self.fasta_file_path = fasta_file_path
self.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_ca
seself.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = uniquesel
f.unique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0self.f
ile_pointer = open(s
>temp_464|source:-*- coding: utf-8 -*-|start:6093|stop:6593
elf, quals_file_path, lazy_init = True):self.quals_file_path = quals_file_pathse
lf.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = 
Noneself.quals_int = Noneself.ids = []self.file_pointer = open(self.quals_file_p
ath)self.file_pointer.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.
total_quals = len([l for l in self.file_pointer.readlines() if l.startswith('>')
])self.reset()def next(self):self.id = self.file_pointer.readline()[1:].strip()s
elf.quals = Noneself
>temp_465|source:-*- coding: utf-8 -*-|start:1979|stop:2479
ta_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneels
e:self.total_seq = len([l for l in self.file_pointer.readlines() if l.startswith
('>')])self.reset()if self.unique:self.init_unique_hash()def init_unique_hash(se
lf):while self.next_regular():hash = hashlib.sha1(self.seq.upper()).hexdigest()i
f hash in self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].append(self.i
d)self.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_dict[hash] = {'
id' : self.id,'ids':
>temp_466|source:-*- coding: utf-8 -*-|start:1996|stop:2496
.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq 
= len([l for l in self.file_pointer.readlines() if l.startswith('>')])self.reset
()if self.unique:self.init_unique_hash()def init_unique_hash(self):while self.ne
xt_regular():hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.un
ique_hash_dict:self.unique_hash_dict[hash]['ids'].append(self.id)self.unique_has
h_dict[hash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'id
s': [self.id],'seq':
>temp_467|source:-*- coding: utf-8 -*-|start:5992|stop:6492
.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passreturnclass Qual
Source:def __init__(self, quals_file_path, lazy_init = True):self.quals_file_pat
h = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id
  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer = o
pen(self.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_q
uals = Noneelse:self.total_quals = len([l for l in self.file_pointer.readlines()
 if l.startswith('>'
>temp_468|source:-*- coding: utf-8 -*-|start:1642|stop:2142
alse, allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.name 
= Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself.pos
 = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.unique_h
ash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0self.file_pointe
r = open(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy_init:self.to
tal_seq = Noneelse:self.total_seq = len([l for l in self.file_pointer.readlines(
) if l.startswith('>
>temp_469|source:-*- coding: utf-8 -*-|start:2029|stop:2529
_init:self.total_seq = Noneelse:self.total_seq = len([l for l in self.file_point
er.readlines() if l.startswith('>')])self.reset()if self.unique:self.init_unique
_hash()def init_unique_hash(self):while self.next_regular():hash = hashlib.sha1(
self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_hash_d
ict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] += 1else:se
lf.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'co
unt': 1}self.unique_
>temp_470|source:-*- coding: utf-8 -*-|start:5895|stop:6395
alpha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_pathtry:plt
.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:pas
sreturnclass QualSource:def __init__(self, quals_file_path, lazy_init = True):se
lf.quals_file_path = quals_file_pathself.name = Noneself.lazy_init = lazy_initse
lf.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self
.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if self.lazy_
init:self.total_qual
>temp_471|source:-*- coding: utf-8 -*-|start:1374|stop:1874
STA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.f
asta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(sel
f):self.fasta.close()class SequenceSource():def __init__(self, fasta_file_path, 
lazy_init = True, unique = False, allow_mixed_case = False):self.fasta_file_path
 = fasta_file_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_cas
e = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.
unique = uniqueself.
>temp_472|source:-*- coding: utf-8 -*-|start:4264|stop:4764
()sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) +
 (int(max(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq
_len + 1)for l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(
figsize = (16, 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', c
olor='0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt
.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top
 = 0.95, right = 0.9
>temp_473|source:-*- coding: utf-8 -*-|start:3588|stop:4088
1return Truedef get_seq_by_read_id(self, read_id):self.reset()while self.next():
if self.id == read_id:return self.seqreturn Falsedef close(self):self.file_point
er.close()def reset(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids = [
]self.file_pointer.seek(0)def visualize_sequence_length_distribution(self, title
, dest = None, max_seq_len = None, xtickstep = None, ytickstep = None):import ma
tplotlib.pyplot as pltimport matplotlib.gridspec as gridspecsequence_lengths = [
]self.reset()while s
>temp_474|source:-*- coding: utf-8 -*-|start:5668|stop:6168
([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s / max: %s'
\% (len(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(sequence_lengt
hs),\min(sequence_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, 
size = 'x-large')if dest == None:dest = self.fasta_file_pathtry:plt.savefig(dest
 + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passreturnclass 
QualSource:def __init__(self, quals_file_path, lazy_init = True):self.quals_file
_path = quals_file_p
>temp_475|source:-*- coding: utf-8 -*-|start:4856|stop:5356
seq_len + 1), seq_len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.yl
abel('number of sequences')plt.xlabel('sequence length')if xtickstep == None:xti
ckstep = (max_seq_len / 50) or 1if ytickstep == None:ytickstep = max(seq_len_dis
tribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1, xtickstep), rot
ation=90, size='xx-small')plt.yticks(range(0, max(seq_len_distribution) + 1, yti
ckstep),[y for y in range(0, max(seq_len_distribution) + 1, ytickstep)],size='xx
-small')plt.xlim(xmi
>temp_476|source:-*- coding: utf-8 -*-|start:5974|stop:6474
t.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:pa
ssreturnclass QualSource:def __init__(self, quals_file_path, lazy_init = True):s
elf.quals_file_path = quals_file_pathself.name = Noneself.lazy_init = lazy_inits
elf.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []sel
f.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if self.lazy
_init:self.total_quals = Noneelse:self.total_quals = len([l for l in self.file_p
ointer.readlines() i
>temp_477|source:-*- coding: utf-8 -*-|start:396|stop:896
lf, output_file_path):self.output_file_path = output_file_pathself.output_file_o
bj = open(output_file_path, 'w')def store(self, entry, split = True, store_frequ
encies = True):if entry.unique and store_frequencies:self.write_id('%s|%s' % (en
try.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_
seq(entry.seq, split)def write_id(self, id):self.output_file_obj.write('>%s\n' %
 id)def write_seq(self, seq, split = True):if split:seq = self.split(seq)self.ou
tput_file_obj.write(
>temp_478|source:-*- coding: utf-8 -*-|start:2247|stop:2747
):hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_d
ict:self.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash
]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.i
d],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(se
lf.unique_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_dict], re
verse = True)]self.total_unique = len(self.unique_hash_dict)self.reset()def next
(self):if self.uniqu
>temp_479|source:-*- coding: utf-8 -*-|start:6801|stop:7301
n Trueelse:return Falseif line.startswith('>'):self.file_pointer.seek(self.file_
pointer.tell() - len(line))breakqualscores += ' ' + line.strip()self.quals = qua
lscores.strip()self.quals_int = [int(q) for q in self.quals.split()]self.pos += 
1return Truedef close(self):self.file_pointer.close()def reset(self):self.pos = 
0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_poi
nter.seek(0)if __name__ == '__main__':fasta = SequenceSource(sys.argv[1])fasta.v
isualize_sequence_le
>temp_480|source:-*- coding: utf-8 -*-|start:4050|stop:4550
quence_lengths = []self.reset()while self.next():if self.pos % 10000 == 0 or sel
f.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.f
lush()sequence_lengths.append(len(self.seq))self.reset()sys.stderr.write('\n')if
 not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengths
) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence
_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcP
arams.update({'axes.
>temp_481|source:-*- coding: utf-8 -*-|start:4048|stop:4548
sequence_lengths = []self.reset()while self.next():if self.pos % 10000 == 0 or s
elf.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr
.flush()sequence_lengths.append(len(self.seq))self.reset()sys.stderr.write('\n')
if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengt
hs) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequen
ce_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.r
cParams.update({'axe
>temp_482|source:-*- coding: utf-8 -*-|start:4892|stop:5392
y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequences')plt.xlabe
l('sequence length')if xtickstep == None:xtickstep = (max_seq_len / 50) or 1if y
tickstep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(range
(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks
(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq
_len_distribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = ma
x_seq_len)plt.ylim(y
>temp_483|source:-*- coding: utf-8 -*-|start:1226|stop:1726
fasta = SequenceSource(f_name)while self.fasta.next():if self.fasta.pos % 1000 =
= 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into mem
ory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self
.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fast
a.close()class SequenceSource():def __init__(self, fasta_file_path, lazy_init = 
True, unique = False, allow_mixed_case = False):self.fasta_file_path = fasta_fil
e_pathself.name = No
>temp_484|source:-*- coding: utf-8 -*-|start:2579|stop:3079
sh_dict[hash]['count'], hash)\for hash in self.unique_hash_dict], reverse = True
)]self.total_unique = len(self.unique_hash_dict)self.reset()def next(self):if se
lf.unique:return self.next_unique()else:return self.next_regular()def next_uniqu
e(self):if self.unique:if self.total_unique > 0 and self.pos < self.total_unique
:hash_entry = self.unique_hash_dict[self.unique_hash_list[self.pos]]self.pos += 
1self.seq = hash_entry['seq'] if self.allow_mixed_case else hash_entry['seq'].up
per()self.id  = hash
>temp_485|source:-*- coding: utf-8 -*-|start:1568|stop:2068
eSource():def __init__(self, fasta_file_path, lazy_init = True, unique = False, 
allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.name = None
self.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0se
lf.id  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.unique_hash_di
ct = {}self.unique_hash_list = []self.unique_next_hash = 0self.file_pointer = op
en(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_se
q = Noneelse:self.to
>temp_486|source:-*- coding: utf-8 -*-|start:423|stop:923
output_file_path = output_file_pathself.output_file_obj = open(output_file_path,
 'w')def store(self, entry, split = True, store_frequencies = True):if entry.uni
que and store_frequencies:self.write_id('%s|%s' % (entry.id, 'frequency:%d' % le
n(entry.ids)))else:self.write_id(entry.id)self.write_seq(entry.seq, split)def wr
ite_id(self, id):self.output_file_obj.write('>%s\n' % id)def write_seq(self, seq
, split = True):if split:seq = self.split(seq)self.output_file_obj.write('%s\n' 
% seq)def split(self
>temp_487|source:-*- coding: utf-8 -*-|start:5693|stop:6193
total: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths)
,numpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_lengths
),\max(sequence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest 
== None:dest = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.save
fig(dest + '.png')try:plt.show()except:passreturnclass QualSource:def __init__(s
elf, quals_file_path, lazy_init = True):self.quals_file_path = quals_file_pathse
lf.name = Noneself.l
>temp_488|source:-*- coding: utf-8 -*-|start:3048|stop:3548
y['seq'].upper()self.id  = hash_entry['id']self.ids = hash_entry['ids']return Tr
ueelse:return Falseelse:return Falsedef next_regular(self):self.seq = Noneself.i
d = self.file_pointer.readline()[1:].strip()sequence = ''while 1:line = self.fil
e_pointer.readline()if not line:if len(sequence):self.seq = sequenceself.pos += 
1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.seek(self
.file_pointer.tell() - len(line))breaksequence += line.strip()self.seq = sequenc
e if self.allow_mixe
>temp_489|source:-*- coding: utf-8 -*-|start:4219|stop:4719
uence_lengths.append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_
seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0
) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:
seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.upd
ate({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewid
th=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.
subplots_adjust(left
>temp_490|source:-*- coding: utf-8 -*-|start:4630|stop:5130
 gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_ad
just(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_distrib
ution, color = 'black', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), 
seq_len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number o
f sequences')plt.xlabel('sequence length')if xtickstep == None:xtickstep = (max_
seq_len / 50) or 1if ytickstep == None:ytickstep = max(seq_len_distribution) / 2
0 or 1plt.xticks(ran
>temp_491|source:-*- coding: utf-8 -*-|start:2907|stop:3407
try = self.unique_hash_dict[self.unique_hash_list[self.pos]]self.pos += 1self.se
q = hash_entry['seq'] if self.allow_mixed_case else hash_entry['seq'].upper()sel
f.id  = hash_entry['id']self.ids = hash_entry['ids']return Trueelse:return False
else:return Falsedef next_regular(self):self.seq = Noneself.id = self.file_point
er.readline()[1:].strip()sequence = ''while 1:line = self.file_pointer.readline(
)if not line:if len(sequence):self.seq = sequenceself.pos += 1return Trueelse:re
turn Falseif line.st
>temp_492|source:-*- coding: utf-8 -*-|start:2478|stop:2978
: [self.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in so
rted([(self.unique_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_
dict], reverse = True)]self.total_unique = len(self.unique_hash_dict)self.reset(
)def next(self):if self.unique:return self.next_unique()else:return self.next_re
gular()def next_unique(self):if self.unique:if self.total_unique > 0 and self.po
s < self.total_unique:hash_entry = self.unique_hash_dict[self.unique_hash_list[s
elf.pos]]self.pos +=
>temp_493|source:-*- coding: utf-8 -*-|start:3616|stop:4116
ead_id(self, read_id):self.reset()while self.next():if self.id == read_id:return
 self.seqreturn Falsedef close(self):self.file_pointer.close()def reset(self):se
lf.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.file_pointer.seek(0)de
f visualize_sequence_length_distribution(self, title, dest = None, max_seq_len =
 None, xtickstep = None, ytickstep = None):import matplotlib.pyplot as pltimport
 matplotlib.gridspec as gridspecsequence_lengths = []self.reset()while self.next
():if self.pos % 100
>temp_494|source:-*- coding: utf-8 -*-|start:5837|stop:6337
sequence_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, size = 'x
-large')if dest == None:dest = self.fasta_file_pathtry:plt.savefig(dest + '.pdf'
)except:plt.savefig(dest + '.png')try:plt.show()except:passreturnclass QualSourc
e:def __init__(self, quals_file_path, lazy_init = True):self.quals_file_path = q
uals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = N
oneself.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer = open(s
elf.quals_file_path)
>temp_495|source:-*- coding: utf-8 -*-|start:4387|stop:4887
seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_
distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update({'ax
es.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewidth=0.1)g
s = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots
_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_dist
ribution, color = 'black', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1
), seq_len_distribut
>temp_496|source:-*- coding: utf-8 -*-|start:207|stop:707
oundation; either version 2 of the License, or (at your option)# any later versi
on.## Please read the docs/COPYING file.import sysimport numpyimport hashlibclas
s FastaOutput:def __init__(self, output_file_path):self.output_file_path = outpu
t_file_pathself.output_file_obj = open(output_file_path, 'w')def store(self, ent
ry, split = True, store_frequencies = True):if entry.unique and store_frequencie
s:self.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.
write_id(entry.id)se
>temp_497|source:-*- coding: utf-8 -*-|start:1242|stop:1742
Source(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fast
a.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self
.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.appen
d(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close()class S
equenceSource():def __init__(self, fasta_file_path, lazy_init = True, unique = F
alse, allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.name 
= Noneself.lazy_init
>temp_498|source:-*- coding: utf-8 -*-|start:4213|stop:4713
h()sequence_lengths.append(len(self.seq))self.reset()sys.stderr.write('\n')if no
t max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) /
 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_le
ngths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcPara
ms.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', l
inewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(Tru
e)plt.subplots_adjus
>temp_499|source:-*- coding: utf-8 -*-|start:4913|stop:5413
k', alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence length')i
f xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:ytic
kstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_l
en + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, max(seq_len
_distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_distribution) + 
1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ym
in = 0, ymax = max(s
>temp_500|source:-*- coding: utf-8 -*-|start:2127|stop:2627
l.startswith('>')])self.reset()if self.unique:self.init_unique_hash()def init_un
ique_hash(self):while self.next_regular():hash = hashlib.sha1(self.seq.upper()).
hexdigest()if hash in self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].a
ppend(self.id)self.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_dic
t[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'count': 1}self.uniqu
e_hash_list = [i[1] for i in sorted([(self.unique_hash_dict[hash]['count'], hash
)\for hash in self.u
>temp_501|source:-*- coding: utf-8 -*-|start:2255|stop:2755
 hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self
.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count
'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq'
: self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.uniqu
e_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_dict], reverse = 
True)]self.total_unique = len(self.unique_hash_dict)self.reset()def next(self):i
f self.unique:return
>temp_502|source:-*- coding: utf-8 -*-|start:5305|stop:5805
tion) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt
.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_distribution) / 
20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-large'
, ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' : 
20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / 
mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.mean(
sequence_lengths), n
>temp_503|source:-*- coding: utf-8 -*-|start:4662|stop:5162
lt.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, t
op = 0.95, right = 0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 
0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, co
lor = 'black', alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequenc
e length')if xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep =
= None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep
, max_seq_len + 1, x
>temp_504|source:-*- coding: utf-8 -*-|start:3017|stop:3517
allow_mixed_case else hash_entry['seq'].upper()self.id  = hash_entry['id']self.i
ds = hash_entry['ids']return Trueelse:return Falseelse:return Falsedef next_regu
lar(self):self.seq = Noneself.id = self.file_pointer.readline()[1:].strip()seque
nce = ''while 1:line = self.file_pointer.readline()if not line:if len(sequence):
self.seq = sequenceself.pos += 1return Trueelse:return Falseif line.startswith('
>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))breaksequence +=
 line.strip()self.se
>temp_505|source:-*- coding: utf-8 -*-|start:4914|stop:5414
', alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence length')if
 xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:ytick
step = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_le
n + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, max(seq_len_
distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_distribution) + 1
, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymi
n = 0, ymax = max(se
>temp_506|source:-*- coding: utf-8 -*-|start:3146|stop:3646
eelse:return Falsedef next_regular(self):self.seq = Noneself.id = self.file_poin
ter.readline()[1:].strip()sequence = ''while 1:line = self.file_pointer.readline
()if not line:if len(sequence):self.seq = sequenceself.pos += 1return Trueelse:r
eturn Falseif line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell
() - len(line))breaksequence += line.strip()self.seq = sequence if self.allow_mi
xed_case else sequence.upper()self.pos += 1return Truedef get_seq_by_read_id(sel
f, read_id):self.res
>temp_507|source:-*- coding: utf-8 -*-|start:614|stop:1114
write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write_i
d(entry.id)self.write_seq(entry.seq, split)def write_id(self, id):self.output_fi
le_obj.write('>%s\n' % id)def write_seq(self, seq, split = True):if split:seq = 
self.split(seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequence,
 piece_length = 80):ticks = range(0, len(sequence), piece_length) + [len(sequenc
e)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks
) - 1)])def close(se
>temp_508|source:-*- coding: utf-8 -*-|start:2836|stop:3336
nique:if self.total_unique > 0 and self.pos < self.total_unique:hash_entry = sel
f.unique_hash_dict[self.unique_hash_list[self.pos]]self.pos += 1self.seq = hash_
entry['seq'] if self.allow_mixed_case else hash_entry['seq'].upper()self.id  = h
ash_entry['id']self.ids = hash_entry['ids']return Trueelse:return Falseelse:retu
rn Falsedef next_regular(self):self.seq = Noneself.id = self.file_pointer.readli
ne()[1:].strip()sequence = ''while 1:line = self.file_pointer.readline()if not l
ine:if len(sequence)
>temp_509|source:-*- coding: utf-8 -*-|start:4432|stop:4932
)for l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize 
= (16, 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.
50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot
(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95,
 right = 0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.fi
ll_between(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'bla
ck', alpha = 0.15)pl
>temp_510|source:-*- coding: utf-8 -*-|start:6856|stop:7356
e_pointer.seek(self.file_pointer.tell() - len(line))breakqualscores += ' ' + lin
e.strip()self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.q
uals.split()]self.pos += 1return Truedef close(self):self.file_pointer.close()de
f reset(self):self.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Nones
elf.ids = []self.file_pointer.seek(0)if __name__ == '__main__':fasta = SequenceS
ource(sys.argv[1])fasta.visualize_sequence_length_distribution(title = sys.argv[
2] if len(sys.argv) 
>temp_511|source:-*- coding: utf-8 -*-|start:637|stop:1137
ry.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_s
eq(entry.seq, split)def write_id(self, id):self.output_file_obj.write('>%s\n' % 
id)def write_seq(self, seq, split = True):if split:seq = self.split(seq)self.out
put_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length = 80):tic
ks = range(0, len(sequence), piece_length) + [len(sequence)]return '\n'.join([se
quence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self)
:self.output_file_ob
>temp_512|source:-*- coding: utf-8 -*-|start:6854|stop:7354
ile_pointer.seek(self.file_pointer.tell() - len(line))breakqualscores += ' ' + l
ine.strip()self.quals = qualscores.strip()self.quals_int = [int(q) for q in self
.quals.split()]self.pos += 1return Truedef close(self):self.file_pointer.close()
def reset(self):self.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Non
eself.ids = []self.file_pointer.seek(0)if __name__ == '__main__':fasta = Sequenc
eSource(sys.argv[1])fasta.visualize_sequence_length_distribution(title = sys.arg
v[2] if len(sys.argv
>temp_513|source:-*- coding: utf-8 -*-|start:4983|stop:5483
 length')if xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep ==
 None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep,
 max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, m
ax(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_distri
bution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)p
lt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_distribution) 
/ 20.0))plt.figtext(
>temp_514|source:-*- coding: utf-8 -*-|start:5721|stop:6221
: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence_lengths
), numpy.std(sequence_lengths),\min(sequence_lengths),\max(sequence_lengths)),\v
a = 'center', alpha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_fi
le_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.sh
ow()except:passreturnclass QualSource:def __init__(self, quals_file_path, lazy_i
nit = True):self.quals_file_path = quals_file_pathself.name = Noneself.lazy_init
 = lazy_initself.pos
>temp_515|source:-*- coding: utf-8 -*-|start:4176|stop:4676
ing: %s' % (self.pos))sys.stderr.flush()sequence_lengths.append(len(self.seq))se
lf.reset()sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(sequence_le
ngths) + (int(max(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0] * 
(max_seq_len + 1)for l in sequence_lengths:seq_len_distribution[l] += 1fig = plt
.figure(figsize = (16, 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('
grid', color='0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)a
x1 = plt.subplot(gs[
>temp_516|source:-*- coding: utf-8 -*-|start:5962|stop:6462
e_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.sho
w()except:passreturnclass QualSource:def __init__(self, quals_file_path, lazy_in
it = True):self.quals_file_path = quals_file_pathself.name = Noneself.lazy_init 
= lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself
.ids = []self.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)
if self.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l for l in
 self.file_pointer.r
>temp_517|source:-*- coding: utf-8 -*-|start:231|stop:731
n 2 of the License, or (at your option)# any later version.## Please read the do
cs/COPYING file.import sysimport numpyimport hashlibclass FastaOutput:def __init
__(self, output_file_path):self.output_file_path = output_file_pathself.output_f
ile_obj = open(output_file_path, 'w')def store(self, entry, split = True, store_
frequencies = True):if entry.unique and store_frequencies:self.write_id('%s|%s' 
% (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)self.w
rite_seq(entry.seq, 
>temp_518|source:-*- coding: utf-8 -*-|start:4608|stop:5108
-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.gri
d(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)pl
t.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.fill_between(range
(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black', alpha = 0.1
5)plt.ylabel('number of sequences')plt.xlabel('sequence length')if xtickstep == 
None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:ytickstep = max(seq
_len_distribution) /
>temp_519|source:-*- coding: utf-8 -*-|start:4565|stop:5065
9})plt.rc('grid', color='0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridS
pec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05,
 bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_distribution, color = 
'black', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distrib
ution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequences')pl
t.xlabel('sequence length')if xtickstep == None:xtickstep = (max_seq_len / 50) o
r 1if ytickstep == N
>temp_520|source:-*- coding: utf-8 -*-|start:1984|stop:2484
le_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:sel
f.total_seq = len([l for l in self.file_pointer.readlines() if l.startswith('>')
])self.reset()if self.unique:self.init_unique_hash()def init_unique_hash(self):w
hile self.next_regular():hash = hashlib.sha1(self.seq.upper()).hexdigest()if has
h in self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].append(self.id)sel
f.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' :
 self.id,'ids': [sel
>temp_521|source:-*- coding: utf-8 -*-|start:6200|stop:6700
t = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Nonese
lf.ids = []self.file_pointer = open(self.quals_file_path)self.file_pointer.seek(
0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l for l 
in self.file_pointer.readlines() if l.startswith('>')])self.reset()def next(self
):self.id = self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_
int = Nonequalscores = ''while 1:line = self.file_pointer.readline()if not line:
if len(qualscores):s
>temp_522|source:-*- coding: utf-8 -*-|start:1398|stop:1898
self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.a
ppend(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close()cla
ss SequenceSource():def __init__(self, fasta_file_path, lazy_init = True, unique
 = False, allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.n
ame = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself
.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.uniq
ue_hash_dict = {}sel
>temp_523|source:-*- coding: utf-8 -*-|start:4615|stop:5115
ewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)
plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(
seq_len_distribution, color = 'black', alpha = 0.3)plt.fill_between(range(0, max
_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.y
label('number of sequences')plt.xlabel('sequence length')if xtickstep == None:xt
ickstep = (max_seq_len / 50) or 1if ytickstep == None:ytickstep = max(seq_len_di
stribution) / 20 or 
>temp_524|source:-*- coding: utf-8 -*-|start:2253|stop:2753
 = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:se
lf.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['cou
nt'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'se
q': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.uni
que_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_dict], reverse 
= True)]self.total_unique = len(self.unique_hash_dict)self.reset()def next(self)
:if self.unique:retu
>temp_525|source:-*- coding: utf-8 -*-|start:2700|stop:3200
h_dict)self.reset()def next(self):if self.unique:return self.next_unique()else:r
eturn self.next_regular()def next_unique(self):if self.unique:if self.total_uniq
ue > 0 and self.pos < self.total_unique:hash_entry = self.unique_hash_dict[self.
unique_hash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.all
ow_mixed_case else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids 
= hash_entry['ids']return Trueelse:return Falseelse:return Falsedef next_regular
(self):self.seq = No
>temp_526|source:-*- coding: utf-8 -*-|start:2147|stop:2647
elf.reset()if self.unique:self.init_unique_hash()def init_unique_hash(self):whil
e self.next_regular():hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash i
n self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].append(self.id)self.u
nique_hash_dict[hash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : se
lf.id,'ids': [self.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] 
for i in sorted([(self.unique_hash_dict[hash]['count'], hash)\for hash in self.u
nique_hash_dict], re
>temp_527|source:-*- coding: utf-8 -*-|start:1817|stop:2317
Noneself.seq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}se
lf.unique_hash_list = []self.unique_next_hash = 0self.file_pointer = open(self.f
asta_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Nonee
lse:self.total_seq = len([l for l in self.file_pointer.readlines() if l.startswi
th('>')])self.reset()if self.unique:self.init_unique_hash()def init_unique_hash(
self):while self.next_regular():hash = hashlib.sha1(self.seq.upper()).hexdigest(
)if hash in self.uni
>temp_528|source:-*- coding: utf-8 -*-|start:2942|stop:3442
ique_hash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.allow
_mixed_case else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = 
hash_entry['ids']return Trueelse:return Falseelse:return Falsedef next_regular(s
elf):self.seq = Noneself.id = self.file_pointer.readline()[1:].strip()sequence =
 ''while 1:line = self.file_pointer.readline()if not line:if len(sequence):self.
seq = sequenceself.pos += 1return Trueelse:return Falseif line.startswith('>'):s
elf.file_pointer.see
>temp_529|source:-*- coding: utf-8 -*-|start:4152|stop:4652
write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_lengths
.append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_seq_len:max_s
eq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or 10)seq_l
en_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_distr
ibution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update({'axes.li
newidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewidth=0.1)gs = g
ridspec.GridSpec(10,
>temp_530|source:-*- coding: utf-8 -*-|start:5734|stop:6234
 %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(
sequence_lengths),\min(sequence_lengths),\max(sequence_lengths)),\va = 'center',
 alpha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_pathtry:pl
t.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:pa
ssreturnclass QualSource:def __init__(self, quals_file_path, lazy_init = True):s
elf.quals_file_path = quals_file_pathself.name = Noneself.lazy_init = lazy_inits
elf.pos = 0self.id  
>temp_531|source:-*- coding: utf-8 -*-|start:3738|stop:4238
file_pointer.close()def reset(self):self.pos = 0self.id  = Noneself.seq = Nonese
lf.ids = []self.file_pointer.seek(0)def visualize_sequence_length_distribution(s
elf, title, dest = None, max_seq_len = None, xtickstep = None, ytickstep = None)
:import matplotlib.pyplot as pltimport matplotlib.gridspec as gridspecsequence_l
engths = []self.reset()while self.next():if self.pos % 10000 == 0 or self.pos ==
 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()se
quence_lengths.appen
>temp_532|source:-*- coding: utf-8 -*-|start:1591|stop:2091
self, fasta_file_path, lazy_init = True, unique = False, allow_mixed_case = Fals
e):self.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init = lazy_i
nitself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq =
 Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_hash
_list = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_file_path
)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.total
_seq = len([l for l 
>temp_533|source:-*- coding: utf-8 -*-|start:5361|stop:5861
, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(
seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'b
lack', size = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.upd
ate({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(
0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(seque
nce_lengths),numpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min(sequ
ence_lengths),\max(s
>temp_534|source:-*- coding: utf-8 -*-|start:4033|stop:4533
pec as gridspecsequence_lengths = []self.reset()while self.next():if self.pos % 
10000 == 0 or self.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.
pos))sys.stderr.flush()sequence_lengths.append(len(self.seq))self.reset()sys.std
err.write('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max
(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)
for l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize =
 (16, 12))plt.rcPara
>temp_535|source:-*- coding: utf-8 -*-|start:2051|stop:2551
 Noneelse:self.total_seq = len([l for l in self.file_pointer.readlines() if l.st
artswith('>')])self.reset()if self.unique:self.init_unique_hash()def init_unique
_hash(self):while self.next_regular():hash = hashlib.sha1(self.seq.upper()).hexd
igest()if hash in self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].appen
d(self.id)self.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_dict[ha
sh] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'count': 1}self.unique_ha
sh_list = [i[1] for 
>temp_536|source:-*- coding: utf-8 -*-|start:122|stop:622
der# the terms of the GNU General Public License as published by the Free# Softw
are Foundation; either version 2 of the License, or (at your option)# any later 
version.## Please read the docs/COPYING file.import sysimport numpyimport hashli
bclass FastaOutput:def __init__(self, output_file_path):self.output_file_path = 
output_file_pathself.output_file_obj = open(output_file_path, 'w')def store(self
, entry, split = True, store_frequencies = True):if entry.unique and store_frequ
encies:self.write_id
>temp_537|source:-*- coding: utf-8 -*-|start:717|stop:1217
eq(entry.seq, split)def write_id(self, id):self.output_file_obj.write('>%s\n' % 
id)def write_seq(self, seq, split = True):if split:seq = self.split(seq)self.out
put_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length = 80):tic
ks = range(0, len(sequence), piece_length) + [len(sequence)]return '\n'.join([se
quence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self)
:self.output_file_obj.close()class ReadFasta:def __init__(self, f_name):self.ids
 = []self.sequences 
>temp_538|source:-*- coding: utf-8 -*-|start:4595|stop:5095
, linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs
[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, ri
ght = 0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.fill_
between(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black'
, alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence length')if 
xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:yticks
tep = max(seq_len_di
>temp_539|source:-*- coding: utf-8 -*-|start:3722|stop:4222
lose(self):self.file_pointer.close()def reset(self):self.pos = 0self.id  = Nones
elf.seq = Noneself.ids = []self.file_pointer.seek(0)def visualize_sequence_lengt
h_distribution(self, title, dest = None, max_seq_len = None, xtickstep = None, y
tickstep = None):import matplotlib.pyplot as pltimport matplotlib.gridspec as gr
idspecsequence_lengths = []self.reset()while self.next():if self.pos % 10000 == 
0 or self.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.
stderr.flush()sequen
>temp_540|source:-*- coding: utf-8 -*-|start:939|stop:1439
e_length = 80):ticks = range(0, len(sequence), piece_length) + [len(sequence)]re
turn '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1
)])def close(self):self.output_file_obj.close()class ReadFasta:def __init__(self
, f_name):self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)whi
le self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.st
derr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))sys.s
tderr.flush()self.id
>temp_541|source:-*- coding: utf-8 -*-|start:1641|stop:2141
False, allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.name
 = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself.po
s = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.unique_
hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0self.file_point
er = open(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy_init:self.t
otal_seq = Noneelse:self.total_seq = len([l for l in self.file_pointer.readlines
() if l.startswith('
>temp_542|source:-*- coding: utf-8 -*-|start:1600|stop:2100
ta_file_path, lazy_init = True, unique = False, allow_mixed_case = False):self.f
asta_file_path = fasta_file_pathself.name = Noneself.lazy_init = lazy_initself.a
llow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself
.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list = [
]self.unique_next_hash = 0self.file_pointer = open(self.fasta_file_path)self.fil
e_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = le
n([l for l in self.f
>temp_543|source:-*- coding: utf-8 -*-|start:1168|stop:1668
init__(self, f_name):self.ids = []self.sequences = []self.fasta = SequenceSource
(f_name)while self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos 
== 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta
.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(self
.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close()class Sequenc
eSource():def __init__(self, fasta_file_path, lazy_init = True, unique = False, 
allow_mixed_case = F
>temp_544|source:-*- coding: utf-8 -*-|start:864|stop:1364
(seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequence, piece_len
gth = 80):ticks = range(0, len(sequence), piece_length) + [len(sequence)]return 
'\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])de
f close(self):self.output_file_obj.close()class ReadFasta:def __init__(self, f_n
ame):self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)while se
lf.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.
write('\r[fastalib] 
>temp_545|source:-*- coding: utf-8 -*-|start:4163|stop:4663
astalib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_lengths.append(len
(self.seq))self.reset()sys.stderr.write('\n')if not max_seq_len:max_seq_len = ma
x(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or 10)seq_len_distribu
tion = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_distribution[l] 
+= 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update({'axes.linewidth' : 
0.9})plt.rc('grid', color='0.50', linestyle='-', linewidth=0.1)gs = gridspec.Gri
dSpec(10, 1)ax1 = pl
>temp_546|source:-*- coding: utf-8 -*-|start:2730|stop:3230
lf):if self.unique:return self.next_unique()else:return self.next_regular()def n
ext_unique(self):if self.unique:if self.total_unique > 0 and self.pos < self.tot
al_unique:hash_entry = self.unique_hash_dict[self.unique_hash_list[self.pos]]sel
f.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_case else hash_entry[
'seq'].upper()self.id  = hash_entry['id']self.ids = hash_entry['ids']return True
else:return Falseelse:return Falsedef next_regular(self):self.seq = Noneself.id 
= self.file_pointer.
>temp_547|source:-*- coding: utf-8 -*-|start:6669|stop:7169
f not line:if len(qualscores):self.quals = qualscores.strip()self.quals_int = [i
nt(q) for q in self.quals.split()]self.pos += 1return Trueelse:return Falseif li
ne.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))b
reakqualscores += ' ' + line.strip()self.quals = qualscores.strip()self.quals_in
t = [int(q) for q in self.quals.split()]self.pos += 1return Truedef close(self):
self.file_pointer.close()def reset(self):self.pos = 0self.id  = Noneself.quals =
 Noneself.quals_int 
>temp_548|source:-*- coding: utf-8 -*-|start:6263|stop:6763
uals_int = Noneself.ids = []self.file_pointer = open(self.quals_file_path)self.f
ile_pointer.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_qual
s = len([l for l in self.file_pointer.readlines() if l.startswith('>')])self.res
et()def next(self):self.id = self.file_pointer.readline()[1:].strip()self.quals 
= Noneself.quals_int = Nonequalscores = ''while 1:line = self.file_pointer.readl
ine()if not line:if len(qualscores):self.quals = qualscores.strip()self.quals_in
t = [int(q) for q in
>temp_549|source:-*- coding: utf-8 -*-|start:1583|stop:2083
_init__(self, fasta_file_path, lazy_init = True, unique = False, allow_mixed_cas
e = False):self.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init 
= lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Nonese
lf.seq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.uni
que_hash_list = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_f
ile_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:se
lf.total_seq = len([
>temp_550|source:-*- coding: utf-8 -*-|start:3319|stop:3819
:if len(sequence):self.seq = sequenceself.pos += 1return Trueelse:return Falseif
 line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line
))breaksequence += line.strip()self.seq = sequence if self.allow_mixed_case else
 sequence.upper()self.pos += 1return Truedef get_seq_by_read_id(self, read_id):s
elf.reset()while self.next():if self.id == read_id:return self.seqreturn Falsede
f close(self):self.file_pointer.close()def reset(self):self.pos = 0self.id  = No
neself.seq = Nonesel
>temp_551|source:-*- coding: utf-8 -*-|start:4673|stop:5173
gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, 
right = 0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.fil
l_between(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'blac
k', alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence length')i
f xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:ytic
kstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_l
en + 1, xtickstep), 
>temp_552|source:-*- coding: utf-8 -*-|start:4983|stop:5483
 length')if xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep ==
 None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep,
 max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, m
ax(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_distri
bution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)p
lt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_distribution) 
/ 20.0))plt.figtext(
>temp_553|source:-*- coding: utf-8 -*-|start:1748|stop:2248
y_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.se
q = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_h
ash_list = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_file_p
ath)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.to
tal_seq = len([l for l in self.file_pointer.readlines() if l.startswith('>')])se
lf.reset()if self.unique:self.init_unique_hash()def init_unique_hash(self):while
 self.next_regular()
>temp_554|source:-*- coding: utf-8 -*-|start:3081|stop:3581
ntry['id']self.ids = hash_entry['ids']return Trueelse:return Falseelse:return Fa
lsedef next_regular(self):self.seq = Noneself.id = self.file_pointer.readline()[
1:].strip()sequence = ''while 1:line = self.file_pointer.readline()if not line:i
f len(sequence):self.seq = sequenceself.pos += 1return Trueelse:return Falseif l
ine.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))
breaksequence += line.strip()self.seq = sequence if self.allow_mixed_case else s
equence.upper()self.
>temp_555|source:-*- coding: utf-8 -*-|start:5973|stop:6473
lt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:p
assreturnclass QualSource:def __init__(self, quals_file_path, lazy_init = True):
self.quals_file_path = quals_file_pathself.name = Noneself.lazy_init = lazy_init
self.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []se
lf.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if self.laz
y_init:self.total_quals = Noneelse:self.total_quals = len([l for l in self.file_
pointer.readlines() 
>temp_556|source:-*- coding: utf-8 -*-|start:4886|stop:5386
tion, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequences')plt
.xlabel('sequence length')if xtickstep == None:xtickstep = (max_seq_len / 50) or
 1if ytickstep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks
(range(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.
yticks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y in range(0, m
ax(seq_len_distribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xma
x = max_seq_len)plt.
>temp_557|source:-*- coding: utf-8 -*-|start:2170|stop:2670
ue:self.init_unique_hash()def init_unique_hash(self):while self.next_regular():h
ash = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict
:self.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['
count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],
'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.
unique_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_dict], rever
se = True)]self.tota
>temp_558|source:-*- coding: utf-8 -*-|start:3081|stop:3581
ntry['id']self.ids = hash_entry['ids']return Trueelse:return Falseelse:return Fa
lsedef next_regular(self):self.seq = Noneself.id = self.file_pointer.readline()[
1:].strip()sequence = ''while 1:line = self.file_pointer.readline()if not line:i
f len(sequence):self.seq = sequenceself.pos += 1return Trueelse:return Falseif l
ine.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))
breaksequence += line.strip()self.seq = sequence if self.allow_mixed_case else s
equence.upper()self.
>temp_559|source:-*- coding: utf-8 -*-|start:2404|stop:2904
ash]['count'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [sel
f.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([
(self.unique_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_dict],
 reverse = True)]self.total_unique = len(self.unique_hash_dict)self.reset()def n
ext(self):if self.unique:return self.next_unique()else:return self.next_regular(
)def next_unique(self):if self.unique:if self.total_unique > 0 and self.pos < se
lf.total_unique:hash
>temp_560|source:-*- coding: utf-8 -*-|start:5765|stop:6265
lengths),numpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min(sequence
_lengths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')
if dest == None:dest = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:
plt.savefig(dest + '.png')try:plt.show()except:passreturnclass QualSource:def __
init__(self, quals_file_path, lazy_init = True):self.quals_file_path = quals_fil
e_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.
quals = Noneself.qua
>temp_561|source:-*- coding: utf-8 -*-|start:5770|stop:6270
hs),numpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_leng
ths),\max(sequence_lengths)),\va = 'center', alpha = 0.8, size = 'x-large')if de
st == None:dest = self.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.s
avefig(dest + '.png')try:plt.show()except:passreturnclass QualSource:def __init_
_(self, quals_file_path, lazy_init = True):self.quals_file_path = quals_file_pat
hself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals
 = Noneself.quals_in
>temp_562|source:-*- coding: utf-8 -*-|start:1365|stop:1865
eading FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.appe
nd(self.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def 
close(self):self.fasta.close()class SequenceSource():def __init__(self, fasta_fi
le_path, lazy_init = True, unique = False, allow_mixed_case = False):self.fasta_
file_path = fasta_file_pathself.name = Noneself.lazy_init = lazy_initself.allow_
mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids 
= []self.unique = un
>temp_563|source:-*- coding: utf-8 -*-|start:5910|stop:6410
ze = 'x-large')if dest == None:dest = self.fasta_file_pathtry:plt.savefig(dest +
 '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passreturnclass Qu
alSource:def __init__(self, quals_file_path, lazy_init = True):self.quals_file_p
ath = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.
id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer =
 open(self.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total
_quals = Noneelse:se
>temp_564|source:-*- coding: utf-8 -*-|start:873|stop:1373
.output_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length = 80)
:ticks = range(0, len(sequence), piece_length) + [len(sequence)]return '\n'.join
([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(s
elf):self.output_file_obj.close()class ReadFasta:def __init__(self, f_name):self
.ids = []self.sequences = []self.fasta = SequenceSource(f_name)while self.fasta.
next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r
[fastalib] Reading F
>temp_565|source:-*- coding: utf-8 -*-|start:3040|stop:3540
ash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_entry['ids']r
eturn Trueelse:return Falseelse:return Falsedef next_regular(self):self.seq = No
neself.id = self.file_pointer.readline()[1:].strip()sequence = ''while 1:line = 
self.file_pointer.readline()if not line:if len(sequence):self.seq = sequenceself
.pos += 1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.s
eek(self.file_pointer.tell() - len(line))breaksequence += line.strip()self.seq =
 sequence if self.al
>temp_566|source:-*- coding: utf-8 -*-|start:3067|stop:3567
f.id  = hash_entry['id']self.ids = hash_entry['ids']return Trueelse:return False
else:return Falsedef next_regular(self):self.seq = Noneself.id = self.file_point
er.readline()[1:].strip()sequence = ''while 1:line = self.file_pointer.readline(
)if not line:if len(sequence):self.seq = sequenceself.pos += 1return Trueelse:re
turn Falseif line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell(
) - len(line))breaksequence += line.strip()self.seq = sequence if self.allow_mix
ed_case else sequenc
>temp_567|source:-*- coding: utf-8 -*-|start:2045|stop:2545
_seq = Noneelse:self.total_seq = len([l for l in self.file_pointer.readlines() i
f l.startswith('>')])self.reset()if self.unique:self.init_unique_hash()def init_
unique_hash(self):while self.next_regular():hash = hashlib.sha1(self.seq.upper()
).hexdigest()if hash in self.unique_hash_dict:self.unique_hash_dict[hash]['ids']
.append(self.id)self.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_d
ict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'count': 1}self.uni
que_hash_list = [i[1
>temp_568|source:-*- coding: utf-8 -*-|start:5998|stop:6498
except:plt.savefig(dest + '.png')try:plt.show()except:passreturnclass QualSource
:def __init__(self, quals_file_path, lazy_init = True):self.quals_file_path = qu
als_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = No
neself.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer = open(se
lf.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_quals =
 Noneelse:self.total_quals = len([l for l in self.file_pointer.readlines() if l.
startswith('>')])sel
>temp_569|source:-*- coding: utf-8 -*-|start:3470|stop:3970
 len(line))breaksequence += line.strip()self.seq = sequence if self.allow_mixed_
case else sequence.upper()self.pos += 1return Truedef get_seq_by_read_id(self, r
ead_id):self.reset()while self.next():if self.id == read_id:return self.seqretur
n Falsedef close(self):self.file_pointer.close()def reset(self):self.pos = 0self
.id  = Noneself.seq = Noneself.ids = []self.file_pointer.seek(0)def visualize_se
quence_length_distribution(self, title, dest = None, max_seq_len = None, xtickst
ep = None, ytickstep
>temp_570|source:-*- coding: utf-8 -*-|start:4489|stop:4989
g = plt.figure(figsize = (16, 12))plt.rcParams.update({'axes.linewidth' : 0.9})p
lt.rc('grid', color='0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(
10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bot
tom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_distribution, color = 'bla
ck', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distributio
n, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequences')plt.xl
abel('sequence lengt
>temp_571|source:-*- coding: utf-8 -*-|start:242|stop:742
License, or (at your option)# any later version.## Please read the docs/COPYING 
file.import sysimport numpyimport hashlibclass FastaOutput:def __init__(self, ou
tput_file_path):self.output_file_path = output_file_pathself.output_file_obj = o
pen(output_file_path, 'w')def store(self, entry, split = True, store_frequencies
 = True):if entry.unique and store_frequencies:self.write_id('%s|%s' % (entry.id
, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_seq(en
try.seq, split)def w
>temp_572|source:-*- coding: utf-8 -*-|start:1865|stop:2365
iqueself.unique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 
0self.file_pointer = open(self.fasta_file_path)self.file_pointer.seek(0)if self.
lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for l in self.file_p
ointer.readlines() if l.startswith('>')])self.reset()if self.unique:self.init_un
ique_hash()def init_unique_hash(self):while self.next_regular():hash = hashlib.s
ha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.unique_ha
sh_dict[hash]['ids']
>temp_573|source:-*- coding: utf-8 -*-|start:6119|stop:6619
_init = True):self.quals_file_path = quals_file_pathself.name = Noneself.lazy_in
it = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Nones
elf.ids = []self.file_pointer = open(self.quals_file_path)self.file_pointer.seek
(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l for l
 in self.file_pointer.readlines() if l.startswith('>')])self.reset()def next(sel
f):self.id = self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals
_int = Nonequalscore
>temp_574|source:-*- coding: utf-8 -*-|start:5329|stop:5829
ize='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = m
ax(seq_len_distribution) + (max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0
.96, '%s' % (title), weight = 'black', size = 'xx-large', ha = 'center')ax1 = pl
t.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)plt.y
ticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f /
 min: %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence_lengths), numpy
.std(sequence_length
>temp_575|source:-*- coding: utf-8 -*-|start:1192|stop:1692
f.ids = []self.sequences = []self.fasta = SequenceSource(f_name)while self.fasta
.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\
r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()
self.ids.append(self.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.wr
ite('\n')def close(self):self.fasta.close()class SequenceSource():def __init__(s
elf, fasta_file_path, lazy_init = True, unique = False, allow_mixed_case = False
):self.fasta_file_pa
>temp_576|source:-*- coding: utf-8 -*-|start:4849|stop:5349
0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black', alpha = 0.15
)plt.ylabel('number of sequences')plt.xlabel('sequence length')if xtickstep == N
one:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:ytickstep = max(seq_
len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1, xtickste
p), rotation=90, size='xx-small')plt.yticks(range(0, max(seq_len_distribution) +
 1, ytickstep),[y for y in range(0, max(seq_len_distribution) + 1, ytickstep)],s
ize='xx-small')plt.x
>temp_577|source:-*- coding: utf-8 -*-|start:28|stop:528
14, A. Murat Eren## This program is free software; you can redistribute it and/o
r modify it under# the terms of the GNU General Public License as published by t
he Free# Software Foundation; either version 2 of the License, or (at your optio
n)# any later version.## Please read the docs/COPYING file.import sysimport nump
yimport hashlibclass FastaOutput:def __init__(self, output_file_path):self.outpu
t_file_path = output_file_pathself.output_file_obj = open(output_file_path, 'w')
def store(self, entr
>temp_578|source:-*- coding: utf-8 -*-|start:1744|stop:2244
 lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Nonesel
f.seq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.uniq
ue_hash_list = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_fi
le_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:sel
f.total_seq = len([l for l in self.file_pointer.readlines() if l.startswith('>')
])self.reset()if self.unique:self.init_unique_hash()def init_unique_hash(self):w
hile self.next_regul
>temp_579|source:-*- coding: utf-8 -*-|start:5345|stop:5845
lt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distr
ibution) + (max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (tit
le), weight = 'black', size = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])
plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])plt.xti
cks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s / max: 
%s'\% (len(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(sequence_le
ngths),\min(sequence
>temp_580|source:-*- coding: utf-8 -*-|start:679|stop:1179
se:self.write_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, id)
:self.output_file_obj.write('>%s\n' % id)def write_seq(self, seq, split = True):
if split:seq = self.split(seq)self.output_file_obj.write('%s\n' % seq)def split(
self, sequence, piece_length = 80):ticks = range(0, len(sequence), piece_length)
 + [len(sequence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in ran
ge(0, len(ticks) - 1)])def close(self):self.output_file_obj.close()class ReadFas
ta:def __init__(self
>temp_581|source:-*- coding: utf-8 -*-|start:973|stop:1473
n(sequence), piece_length) + [len(sequence)]return '\n'.join([sequence[ticks[x]:
ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.output_fil
e_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = []self.sequen
ces = []self.fasta = SequenceSource(f_name)while self.fasta.next():if self.fasta
.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading F
ASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.
fasta.id)self.sequen
>temp_582|source:-*- coding: utf-8 -*-|start:5902|stop:6402
 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_pathtry:plt.savefi
g(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passreturn
class QualSource:def __init__(self, quals_file_path, lazy_init = True):self.qual
s_file_path = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos 
= 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_p
ointer = open(self.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:se
lf.total_quals = Non
>temp_583|source:-*- coding: utf-8 -*-|start:5058|stop:5558
ep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtick
step, max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range
(0, max(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_d
istribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_
len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_distribut
ion) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx
-large', ha = 'cente
>temp_584|source:-*- coding: utf-8 -*-|start:22|stop:522
(C) 2014, A. Murat Eren## This program is free software; you can redistribute it
 and/or modify it under# the terms of the GNU General Public License as publishe
d by the Free# Software Foundation; either version 2 of the License, or (at your
 option)# any later version.## Please read the docs/COPYING file.import sysimpor
t numpyimport hashlibclass FastaOutput:def __init__(self, output_file_path):self
.output_file_path = output_file_pathself.output_file_obj = open(output_file_path
, 'w')def store(self
>temp_585|source:-*- coding: utf-8 -*-|start:6242|stop:6742
lf.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer = open(self.q
uals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_quals = Non
eelse:self.total_quals = len([l for l in self.file_pointer.readlines() if l.star
tswith('>')])self.reset()def next(self):self.id = self.file_pointer.readline()[1
:].strip()self.quals = Noneself.quals_int = Nonequalscores = ''while 1:line = se
lf.file_pointer.readline()if not line:if len(qualscores):self.quals = qualscores
.strip()self.quals_i
>temp_586|source:-*- coding: utf-8 -*-|start:171|stop:671
as published by the Free# Software Foundation; either version 2 of the License, 
or (at your option)# any later version.## Please read the docs/COPYING file.impo
rt sysimport numpyimport hashlibclass FastaOutput:def __init__(self, output_file
_path):self.output_file_path = output_file_pathself.output_file_obj = open(outpu
t_file_path, 'w')def store(self, entry, split = True, store_frequencies = True):
if entry.unique and store_frequencies:self.write_id('%s|%s' % (entry.id, 'freque
ncy:%d' % len(entry.
>temp_587|source:-*- coding: utf-8 -*-|start:3765|stop:4265
et(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.file_pointe
r.seek(0)def visualize_sequence_length_distribution(self, title, dest = None, ma
x_seq_len = None, xtickstep = None, ytickstep = None):import matplotlib.pyplot a
s pltimport matplotlib.gridspec as gridspecsequence_lengths = []self.reset()whil
e self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fas
talib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_lengths.append(len(s
elf.seq))self.reset(
>temp_588|source:-*- coding: utf-8 -*-|start:3241|stop:3741
1:].strip()sequence = ''while 1:line = self.file_pointer.readline()if not line:i
f len(sequence):self.seq = sequenceself.pos += 1return Trueelse:return Falseif l
ine.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))
breaksequence += line.strip()self.seq = sequence if self.allow_mixed_case else s
equence.upper()self.pos += 1return Truedef get_seq_by_read_id(self, read_id):sel
f.reset()while self.next():if self.id == read_id:return self.seqreturn Falsedef 
close(self):self.fil
>temp_589|source:-*- coding: utf-8 -*-|start:4246|stop:4746
lf.seq))self.reset()sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(s
equence_lengths) + (int(max(sequence_lengths) / 100.0) or 10)seq_len_distributio
n = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_distribution[l] += 
1fig = plt.figure(figsize = (16, 12))plt.rcParams.update({'axes.linewidth' : 0.9
})plt.rc('grid', color='0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSp
ec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, 
bottom = 0.03, top =
>temp_590|source:-*- coding: utf-8 -*-|start:1330|stop:1830
 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.p
os))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(self.f
asta.seq)sys.stderr.write('\n')def close(self):self.fasta.close()class SequenceS
ource():def __init__(self, fasta_file_path, lazy_init = True, unique = False, al
low_mixed_case = False):self.fasta_file_path = fasta_file_pathself.name = Nonese
lf.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self
.id  = Noneself.seq 
>temp_591|source:-*- coding: utf-8 -*-|start:3847|stop:4347
seek(0)def visualize_sequence_length_distribution(self, title, dest = None, max_
seq_len = None, xtickstep = None, ytickstep = None):import matplotlib.pyplot as 
pltimport matplotlib.gridspec as gridspecsequence_lengths = []self.reset()while 
self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fasta
lib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_lengths.append(len(sel
f.seq))self.reset()sys.stderr.write('\n')if not max_seq_len:max_seq_len = max(se
quence_lengths) + (i
>temp_592|source:-*- coding: utf-8 -*-|start:799|stop:1299
)def write_seq(self, seq, split = True):if split:seq = self.split(seq)self.outpu
t_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length = 80):ticks
 = range(0, len(sequence), piece_length) + [len(sequence)]return '\n'.join([sequ
ence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):s
elf.output_file_obj.close()class ReadFasta:def __init__(self, f_name):self.ids =
 []self.sequences = []self.fasta = SequenceSource(f_name)while self.fasta.next()
:if self.fasta.pos %
>temp_593|source:-*- coding: utf-8 -*-|start:1732|stop:2232
.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.i
d  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict =
 {}self.unique_hash_list = []self.unique_next_hash = 0self.file_pointer = open(s
elf.fasta_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = 
Noneelse:self.total_seq = len([l for l in self.file_pointer.readlines() if l.sta
rtswith('>')])self.reset()if self.unique:self.init_unique_hash()def init_unique_
hash(self):while sel
>temp_594|source:-*- coding: utf-8 -*-|start:4583|stop:5083
color='0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = pl
t.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, to
p = 0.95, right = 0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 0
.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, col
or = 'black', alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence
 length')if xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep ==
 None:ytickstep = ma
>temp_595|source:-*- coding: utf-8 -*-|start:5380|stop:5880
n)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_distributio
n) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-l
arge', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolo
r' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: 
%s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.
mean(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_lengths),\max(
sequence_lengths)),\
>temp_596|source:-*- coding: utf-8 -*-|start:5659|stop:6159
lt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min: %s /
 max: %s'\% (len(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(seque
nce_lengths),\min(sequence_lengths),\max(sequence_lengths)),\va = 'center', alph
a = 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_pathtry:plt.sav
efig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:passret
urnclass QualSource:def __init__(self, quals_file_path, lazy_init = True):self.q
uals_file_path = qua
>temp_597|source:-*- coding: utf-8 -*-|start:2261|stop:2761
ib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:self.uniqu
e_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['count'] += 
1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self
.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.unique_hash
_dict[hash]['count'], hash)\for hash in self.unique_hash_dict], reverse = True)]
self.total_unique = len(self.unique_hash_dict)self.reset()def next(self):if self
.unique:return self.
>temp_598|source:-*- coding: utf-8 -*-|start:3268|stop:3768
le 1:line = self.file_pointer.readline()if not line:if len(sequence):self.seq = 
sequenceself.pos += 1return Trueelse:return Falseif line.startswith('>'):self.fi
le_pointer.seek(self.file_pointer.tell() - len(line))breaksequence += line.strip
()self.seq = sequence if self.allow_mixed_case else sequence.upper()self.pos += 
1return Truedef get_seq_by_read_id(self, read_id):self.reset()while self.next():
if self.id == read_id:return self.seqreturn Falsedef close(self):self.file_point
er.close()def reset(
>temp_599|source:-*- coding: utf-8 -*-|start:4034|stop:4534
ec as gridspecsequence_lengths = []self.reset()while self.next():if self.pos % 1
0000 == 0 or self.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.p
os))sys.stderr.flush()sequence_lengths.append(len(self.seq))self.reset()sys.stde
rr.write('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(
sequence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)f
or l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = 
(16, 12))plt.rcParam
>temp_600|source:-*- coding: utf-8 -*-|start:3350|stop:3850
quenceself.pos += 1return Trueelse:return Falseif line.startswith('>'):self.file
_pointer.seek(self.file_pointer.tell() - len(line))breaksequence += line.strip()
self.seq = sequence if self.allow_mixed_case else sequence.upper()self.pos += 1r
eturn Truedef get_seq_by_read_id(self, read_id):self.reset()while self.next():if
 self.id == read_id:return self.seqreturn Falsedef close(self):self.file_pointer
.close()def reset(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids = []s
elf.file_pointer.see
>temp_601|source:-*- coding: utf-8 -*-|start:92|stop:592
tribute it and/or modify it under# the terms of the GNU General Public License a
s published by the Free# Software Foundation; either version 2 of the License, o
r (at your option)# any later version.## Please read the docs/COPYING file.impor
t sysimport numpyimport hashlibclass FastaOutput:def __init__(self, output_file_
path):self.output_file_path = output_file_pathself.output_file_obj = open(output
_file_path, 'w')def store(self, entry, split = True, store_frequencies = True):i
f entry.unique and s
>temp_602|source:-*- coding: utf-8 -*-|start:1851|stop:2351
lf.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list = []self.uniqu
e_next_hash = 0self.file_pointer = open(self.fasta_file_path)self.file_pointer.s
eek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for l 
in self.file_pointer.readlines() if l.startswith('>')])self.reset()if self.uniqu
e:self.init_unique_hash()def init_unique_hash(self):while self.next_regular():ha
sh = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:
self.unique_hash_dic
>temp_603|source:-*- coding: utf-8 -*-|start:3332|stop:3832
nce):self.seq = sequenceself.pos += 1return Trueelse:return Falseif line.startsw
ith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))breaksequen
ce += line.strip()self.seq = sequence if self.allow_mixed_case else sequence.upp
er()self.pos += 1return Truedef get_seq_by_read_id(self, read_id):self.reset()wh
ile self.next():if self.id == read_id:return self.seqreturn Falsedef close(self)
:self.file_pointer.close()def reset(self):self.pos = 0self.id  = Noneself.seq = 
Noneself.ids = []sel
>temp_604|source:-*- coding: utf-8 -*-|start:452|stop:952
e_pathself.output_file_obj = open(output_file_path, 'w')def store(self, entry, s
plit = True, store_frequencies = True):if entry.unique and store_frequencies:sel
f.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.write
_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, id):self.output_
file_obj.write('>%s\n' % id)def write_seq(self, seq, split = True):if split:seq 
= self.split(seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequenc
e, piece_length = 80
>temp_605|source:-*- coding: utf-8 -*-|start:1132|stop:1632
le_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = []self.seque
nces = []self.fasta = SequenceSource(f_name)while self.fasta.next():if self.fast
a.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading 
FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self
.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(s
elf):self.fasta.close()class SequenceSource():def __init__(self, fasta_file_path
, lazy_init = True, 
>temp_606|source:-*- coding: utf-8 -*-|start:3619|stop:4119
_id(self, read_id):self.reset()while self.next():if self.id == read_id:return se
lf.seqreturn Falsedef close(self):self.file_pointer.close()def reset(self):self.
pos = 0self.id  = Noneself.seq = Noneself.ids = []self.file_pointer.seek(0)def v
isualize_sequence_length_distribution(self, title, dest = None, max_seq_len = No
ne, xtickstep = None, ytickstep = None):import matplotlib.pyplot as pltimport ma
tplotlib.gridspec as gridspecsequence_lengths = []self.reset()while self.next():
if self.pos % 10000 
>temp_607|source:-*- coding: utf-8 -*-|start:478|stop:978
 = open(output_file_path, 'w')def store(self, entry, split = True, store_frequen
cies = True):if entry.unique and store_frequencies:self.write_id('%s|%s' % (entr
y.id, 'frequency:%d' % len(entry.ids)))else:self.write_id(entry.id)self.write_se
q(entry.seq, split)def write_id(self, id):self.output_file_obj.write('>%s\n' % i
d)def write_seq(self, seq, split = True):if split:seq = self.split(seq)self.outp
ut_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length = 80):tick
s = range(0, len(seq
>temp_608|source:-*- coding: utf-8 -*-|start:439|stop:939
 = output_file_pathself.output_file_obj = open(output_file_path, 'w')def store(s
elf, entry, split = True, store_frequencies = True):if entry.unique and store_fr
equencies:self.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))el
se:self.write_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, id)
:self.output_file_obj.write('>%s\n' % id)def write_seq(self, seq, split = True):
if split:seq = self.split(seq)self.output_file_obj.write('%s\n' % seq)def split(
self, sequence, piec
>temp_609|source:-*- coding: utf-8 -*-|start:5953|stop:6453
fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')tr
y:plt.show()except:passreturnclass QualSource:def __init__(self, quals_file_path
, lazy_init = True):self.quals_file_path = quals_file_pathself.name = Noneself.l
azy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_int =
 Noneself.ids = []self.file_pointer = open(self.quals_file_path)self.file_pointe
r.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l
 for l in self.file_
>temp_610|source:-*- coding: utf-8 -*-|start:4535|stop:5035
.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', lin
ewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)
plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(
seq_len_distribution, color = 'black', alpha = 0.3)plt.fill_between(range(0, max
_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.y
label('number of sequences')plt.xlabel('sequence length')if xtickstep == None:xt
ickstep = (max_seq_l
>temp_611|source:-*- coding: utf-8 -*-|start:6038|stop:6538
.show()except:passreturnclass QualSource:def __init__(self, quals_file_path, laz
y_init = True):self.quals_file_path = quals_file_pathself.name = Noneself.lazy_i
nit = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = None
self.ids = []self.file_pointer = open(self.quals_file_path)self.file_pointer.see
k(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l for 
l in self.file_pointer.readlines() if l.startswith('>')])self.reset()def next(se
lf):self.id = self.f
>temp_612|source:-*- coding: utf-8 -*-|start:5205|stop:5705
.yticks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y in range(0, 
max(seq_len_distribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xm
ax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_
len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black
', size = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update(
{'axes.edgecolor' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(0.02
, 0.5, 'total: %s / 
>temp_613|source:-*- coding: utf-8 -*-|start:6287|stop:6787
= []self.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if se
lf.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l for l in self
.file_pointer.readlines() if l.startswith('>')])self.reset()def next(self):self.
id = self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_int = N
onequalscores = ''while 1:line = self.file_pointer.readline()if not line:if len(
qualscores):self.quals = qualscores.strip()self.quals_int = [int(q) for q in sel
f.quals.split()]self
>temp_614|source:-*- coding: utf-8 -*-|start:5098|stop:5598
ibution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1, xtickstep), rotat
ion=90, size='xx-small')plt.yticks(range(0, max(seq_len_distribution) + 1, ytick
step),[y for y in range(0, max(seq_len_distribution) + 1, ytickstep)],size='xx-s
mall')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_le
n_distribution) + (max(seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s'
 % (title), weight = 'black', size = 'xx-large', ha = 'center')ax1 = plt.subplot
(gs[9])plt.rcParams.
>temp_615|source:-*- coding: utf-8 -*-|start:6321|stop:6821
quals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_quals = No
neelse:self.total_quals = len([l for l in self.file_pointer.readlines() if l.sta
rtswith('>')])self.reset()def next(self):self.id = self.file_pointer.readline()[
1:].strip()self.quals = Noneself.quals_int = Nonequalscores = ''while 1:line = s
elf.file_pointer.readline()if not line:if len(qualscores):self.quals = qualscore
s.strip()self.quals_int = [int(q) for q in self.quals.split()]self.pos += 1retur
n Trueelse:return Fa
>temp_616|source:-*- coding: utf-8 -*-|start:2874|stop:3374
f.pos < self.total_unique:hash_entry = self.unique_hash_dict[self.unique_hash_li
st[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_case 
else hash_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_entry['
ids']return Trueelse:return Falseelse:return Falsedef next_regular(self):self.se
q = Noneself.id = self.file_pointer.readline()[1:].strip()sequence = ''while 1:l
ine = self.file_pointer.readline()if not line:if len(sequence):self.seq = sequen
ceself.pos += 1retur
>temp_617|source:-*- coding: utf-8 -*-|start:1430|stop:1930
()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.
write('\n')def close(self):self.fasta.close()class SequenceSource():def __init__
(self, fasta_file_path, lazy_init = True, unique = False, allow_mixed_case = Fal
se):self.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init = lazy_
initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq 
= Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_has
h_list = []self.uniq
>temp_618|source:-*- coding: utf-8 -*-|start:3041|stop:3541
sh_entry['seq'].upper()self.id  = hash_entry['id']self.ids = hash_entry['ids']re
turn Trueelse:return Falseelse:return Falsedef next_regular(self):self.seq = Non
eself.id = self.file_pointer.readline()[1:].strip()sequence = ''while 1:line = s
elf.file_pointer.readline()if not line:if len(sequence):self.seq = sequenceself.
pos += 1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.se
ek(self.file_pointer.tell() - len(line))breaksequence += line.strip()self.seq = 
sequence if self.all
>temp_619|source:-*- coding: utf-8 -*-|start:1302|stop:1802
00 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into
 memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)
self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.
fasta.close()class SequenceSource():def __init__(self, fasta_file_path, lazy_ini
t = True, unique = False, allow_mixed_case = False):self.fasta_file_path = fasta
_file_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = allo
w_mixed_caseself.pos
>temp_620|source:-*- coding: utf-8 -*-|start:1457|stop:1957
a.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):
self.fasta.close()class SequenceSource():def __init__(self, fasta_file_path, laz
y_init = True, unique = False, allow_mixed_case = False):self.fasta_file_path = 
fasta_file_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case =
 allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.uni
que = uniqueself.unique_hash_dict = {}self.unique_hash_list = []self.unique_next
_hash = 0self.file_p
>temp_621|source:-*- coding: utf-8 -*-|start:6014|stop:6514
ig(dest + '.png')try:plt.show()except:passreturnclass QualSource:def __init__(se
lf, quals_file_path, lazy_init = True):self.quals_file_path = quals_file_pathsel
f.name = Noneself.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = N
oneself.quals_int = Noneself.ids = []self.file_pointer = open(self.quals_file_pa
th)self.file_pointer.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.t
otal_quals = len([l for l in self.file_pointer.readlines() if l.startswith('>')]
)self.reset()def nex
>temp_622|source:-*- coding: utf-8 -*-|start:2094|stop:2594
self.file_pointer.readlines() if l.startswith('>')])self.reset()if self.unique:s
elf.init_unique_hash()def init_unique_hash(self):while self.next_regular():hash 
= hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dict:sel
f.unique_hash_dict[hash]['ids'].append(self.id)self.unique_hash_dict[hash]['coun
t'] += 1else:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq
': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.uniq
ue_hash_dict[hash]['
>temp_623|source:-*- coding: utf-8 -*-|start:1692|stop:2192
th = fasta_file_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_c
ase = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []sel
f.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list = []self.unique
_next_hash = 0self.file_pointer = open(self.fasta_file_path)self.file_pointer.se
ek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for l i
n self.file_pointer.readlines() if l.startswith('>')])self.reset()if self.unique
:self.init_unique_ha
>temp_624|source:-*- coding: utf-8 -*-|start:4879|stop:5379
istribution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequenc
es')plt.xlabel('sequence length')if xtickstep == None:xtickstep = (max_seq_len /
 50) or 1if ytickstep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt
.xticks(range(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-smal
l')plt.yticks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y in ran
ge(0, max(seq_len_distribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin =
 0, xmax = max_seq_l
>temp_625|source:-*- coding: utf-8 -*-|start:4069|stop:4569
self.reset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.stde
rr.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_leng
ths.append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_seq_len:ma
x_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or 10)se
q_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_di
stribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update({'axes
.linewidth' : 0.9})p
>temp_626|source:-*- coding: utf-8 -*-|start:866|stop:1366
eq)self.output_file_obj.write('%s\n' % seq)def split(self, sequence, piece_lengt
h = 80):ticks = range(0, len(sequence), piece_length) + [len(sequence)]return '\
n'.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def 
close(self):self.output_file_obj.close()class ReadFasta:def __init__(self, f_nam
e):self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)while self
.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.wr
ite('\r[fastalib] Re
>temp_627|source:-*- coding: utf-8 -*-|start:1668|stop:2168
alse):self.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init = laz
y_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.se
q = Noneself.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_h
ash_list = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_file_p
ath)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.to
tal_seq = len([l for l in self.file_pointer.readlines() if l.startswith('>')])se
lf.reset()if self.un
>temp_628|source:-*- coding: utf-8 -*-|start:2128|stop:2628
.startswith('>')])self.reset()if self.unique:self.init_unique_hash()def init_uni
que_hash(self):while self.next_regular():hash = hashlib.sha1(self.seq.upper()).h
exdigest()if hash in self.unique_hash_dict:self.unique_hash_dict[hash]['ids'].ap
pend(self.id)self.unique_hash_dict[hash]['count'] += 1else:self.unique_hash_dict
[hash] = {'id' : self.id,'ids': [self.id],'seq': self.seq,'count': 1}self.unique
_hash_list = [i[1] for i in sorted([(self.unique_hash_dict[hash]['count'], hash)
\for hash in self.un
>temp_629|source:-*- coding: utf-8 -*-|start:1326|stop:1826
s == 1:sys.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fas
ta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(se
lf.fasta.seq)sys.stderr.write('\n')def close(self):self.fasta.close()class Seque
nceSource():def __init__(self, fasta_file_path, lazy_init = True, unique = False
, allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.name = No
neself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0
self.id  = Noneself.
>temp_630|source:-*- coding: utf-8 -*-|start:4571|stop:5071
.rc('grid', color='0.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10
, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, botto
m = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_distribution, color = 'black
', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distribution,
 y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of sequences')plt.xlab
el('sequence length')if xtickstep == None:xtickstep = (max_seq_len / 50) or 1if 
ytickstep == None:yt
>temp_631|source:-*- coding: utf-8 -*-|start:1440|stop:1940
.append(self.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n'
)def close(self):self.fasta.close()class SequenceSource():def __init__(self, fas
ta_file_path, lazy_init = True, unique = False, allow_mixed_case = False):self.f
asta_file_path = fasta_file_pathself.name = Noneself.lazy_init = lazy_initself.a
llow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself
.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list = [
]self.unique_next_ha
>temp_632|source:-*- coding: utf-8 -*-|start:6447|stop:6947
.file_pointer.readlines() if l.startswith('>')])self.reset()def next(self):self.
id = self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_int = N
onequalscores = ''while 1:line = self.file_pointer.readline()if not line:if len(
qualscores):self.quals = qualscores.strip()self.quals_int = [int(q) for q in sel
f.quals.split()]self.pos += 1return Trueelse:return Falseif line.startswith('>')
:self.file_pointer.seek(self.file_pointer.tell() - len(line))breakqualscores += 
' ' + line.strip()se
>temp_633|source:-*- coding: utf-8 -*-|start:5954|stop:6454
asta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try
:plt.show()except:passreturnclass QualSource:def __init__(self, quals_file_path,
 lazy_init = True):self.quals_file_path = quals_file_pathself.name = Noneself.la
zy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = 
Noneself.ids = []self.file_pointer = open(self.quals_file_path)self.file_pointer
.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l 
for l in self.file_p
>temp_634|source:-*- coding: utf-8 -*-|start:1306|stop:1806
= 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Reading FASTA into mem
ory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta.id)self
.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.fast
a.close()class SequenceSource():def __init__(self, fasta_file_path, lazy_init = 
True, unique = False, allow_mixed_case = False):self.fasta_file_path = fasta_fil
e_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mi
xed_caseself.pos = 0
>temp_635|source:-*- coding: utf-8 -*-|start:4764|stop:5264
8)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.fill_between(r
ange(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'black', alpha =
 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence length')if xtickstep
 == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:ytickstep = max
(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1, xti
ckstep), rotation=90, size='xx-small')plt.yticks(range(0, max(seq_len_distributi
on) + 1, ytickstep),
>temp_636|source:-*- coding: utf-8 -*-|start:4431|stop:4931
1)for l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize
 = (16, 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0
.50', linestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplo
t(gs[0:8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95
, right = 0.98)plt.plot(seq_len_distribution, color = 'black', alpha = 0.3)plt.f
ill_between(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, color = 'bl
ack', alpha = 0.15)p
>temp_637|source:-*- coding: utf-8 -*-|start:2773|stop:3273
)else:return self.next_regular()def next_unique(self):if self.unique:if self.tot
al_unique > 0 and self.pos < self.total_unique:hash_entry = self.unique_hash_dic
t[self.unique_hash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if s
elf.allow_mixed_case else hash_entry['seq'].upper()self.id  = hash_entry['id']se
lf.ids = hash_entry['ids']return Trueelse:return Falseelse:return Falsedef next_
regular(self):self.seq = Noneself.id = self.file_pointer.readline()[1:].strip()s
equence = ''while 1:
>temp_638|source:-*- coding: utf-8 -*-|start:4278|stop:4778
rite('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequ
ence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l
 in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16,
 12))plt.rcParams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', l
inestyle='-', linewidth=0.1)gs = gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:
8])plt.grid(True)plt.subplots_adjust(left=0.05, bottom = 0.03, top = 0.95, right
 = 0.98)plt.plot(seq
>temp_639|source:-*- coding: utf-8 -*-|start:1732|stop:2232
.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself.pos = 0self.i
d  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.unique_hash_dict =
 {}self.unique_hash_list = []self.unique_next_hash = 0self.file_pointer = open(s
elf.fasta_file_path)self.file_pointer.seek(0)if self.lazy_init:self.total_seq = 
Noneelse:self.total_seq = len([l for l in self.file_pointer.readlines() if l.sta
rtswith('>')])self.reset()if self.unique:self.init_unique_hash()def init_unique_
hash(self):while sel
>temp_640|source:-*- coding: utf-8 -*-|start:2424|stop:2924
se:self.unique_hash_dict[hash] = {'id' : self.id,'ids': [self.id],'seq': self.se
q,'count': 1}self.unique_hash_list = [i[1] for i in sorted([(self.unique_hash_di
ct[hash]['count'], hash)\for hash in self.unique_hash_dict], reverse = True)]sel
f.total_unique = len(self.unique_hash_dict)self.reset()def next(self):if self.un
ique:return self.next_unique()else:return self.next_regular()def next_unique(sel
f):if self.unique:if self.total_unique > 0 and self.pos < self.total_unique:hash
_entry = self.unique
>temp_641|source:-*- coding: utf-8 -*-|start:3890|stop:4390
bution(self, title, dest = None, max_seq_len = None, xtickstep = None, ytickstep
 = None):import matplotlib.pyplot as pltimport matplotlib.gridspec as gridspecse
quence_lengths = []self.reset()while self.next():if self.pos % 10000 == 0 or sel
f.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.f
lush()sequence_lengths.append(len(self.seq))self.reset()sys.stderr.write('\n')if
 not max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengths
) / 100.0) or 10)seq
>temp_642|source:-*- coding: utf-8 -*-|start:1378|stop:1878
into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(self.fasta
.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):s
elf.fasta.close()class SequenceSource():def __init__(self, fasta_file_path, lazy
_init = True, unique = False, allow_mixed_case = False):self.fasta_file_path = f
asta_file_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = 
allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.uniq
ue = uniqueself.uniq
>temp_643|source:-*- coding: utf-8 -*-|start:670|stop:1170
.ids)))else:self.write_id(entry.id)self.write_seq(entry.seq, split)def write_id(
self, id):self.output_file_obj.write('>%s\n' % id)def write_seq(self, seq, split
 = True):if split:seq = self.split(seq)self.output_file_obj.write('%s\n' % seq)d
ef split(self, sequence, piece_length = 80):ticks = range(0, len(sequence), piec
e_length) + [len(sequence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for
 x in range(0, len(ticks) - 1)])def close(self):self.output_file_obj.close()clas
s ReadFasta:def __in
>temp_644|source:-*- coding: utf-8 -*-|start:5144|stop:5644
max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, ma
x(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_distrib
ution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)pl
t.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_distribution) /
 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-large
', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' :
 20})plt.grid(False)
>temp_645|source:-*- coding: utf-8 -*-|start:2892|stop:3392
_unique:hash_entry = self.unique_hash_dict[self.unique_hash_list[self.pos]]self.
pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_case else hash_entry['s
eq'].upper()self.id  = hash_entry['id']self.ids = hash_entry['ids']return Trueel
se:return Falseelse:return Falsedef next_regular(self):self.seq = Noneself.id = 
self.file_pointer.readline()[1:].strip()sequence = ''while 1:line = self.file_po
inter.readline()if not line:if len(sequence):self.seq = sequenceself.pos += 1ret
urn Trueelse:return 
>temp_646|source:-*- coding: utf-8 -*-|start:1463|stop:1963
elf.sequences.append(self.fasta.seq)sys.stderr.write('\n')def close(self):self.f
asta.close()class SequenceSource():def __init__(self, fasta_file_path, lazy_init
 = True, unique = False, allow_mixed_case = False):self.fasta_file_path = fasta_
file_pathself.name = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow
_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = 
uniqueself.unique_hash_dict = {}self.unique_hash_list = []self.unique_next_hash 
= 0self.file_pointer
>temp_647|source:-*- coding: utf-8 -*-|start:6801|stop:7301
n Trueelse:return Falseif line.startswith('>'):self.file_pointer.seek(self.file_
pointer.tell() - len(line))breakqualscores += ' ' + line.strip()self.quals = qua
lscores.strip()self.quals_int = [int(q) for q in self.quals.split()]self.pos += 
1return Truedef close(self):self.file_pointer.close()def reset(self):self.pos = 
0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_poi
nter.seek(0)if __name__ == '__main__':fasta = SequenceSource(sys.argv[1])fasta.v
isualize_sequence_le
>temp_648|source:-*- coding: utf-8 -*-|start:5081|stop:5581
max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1, 
xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, max(seq_len_distrib
ution) + 1, ytickstep),[y for y in range(0, max(seq_len_distribution) + 1, ytick
step)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, 
ymax = max(seq_len_distribution) + (max(seq_len_distribution) / 20.0))plt.figtex
t(0.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-large', ha = 'center')
ax1 = plt.subplot(gs
>temp_649|source:-*- coding: utf-8 -*-|start:4310|stop:4810
_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or 10)seq
_len_distribution = [0] * (max_seq_len + 1)for l in sequence_lengths:seq_len_dis
tribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcParams.update({'axes.
linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', linewidth=0.1)gs =
 gridspec.GridSpec(10, 1)ax1 = plt.subplot(gs[0:8])plt.grid(True)plt.subplots_ad
just(left=0.05, bottom = 0.03, top = 0.95, right = 0.98)plt.plot(seq_len_distrib
ution, color = 'blac
>temp_650|source:-*- coding: utf-8 -*-|start:1596|stop:2096
 fasta_file_path, lazy_init = True, unique = False, allow_mixed_case = False):se
lf.fasta_file_path = fasta_file_pathself.name = Noneself.lazy_init = lazy_initse
lf.allow_mixed_case = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = None
self.ids = []self.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list
 = []self.unique_next_hash = 0self.file_pointer = open(self.fasta_file_path)self
.file_pointer.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq 
= len([l for l in se
>temp_651|source:-*- coding: utf-8 -*-|start:6065|stop:6565
ss QualSource:def __init__(self, quals_file_path, lazy_init = True):self.quals_f
ile_path = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0
self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_poin
ter = open(self.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.
total_quals = Noneelse:self.total_quals = len([l for l in self.file_pointer.read
lines() if l.startswith('>')])self.reset()def next(self):self.id = self.file_poi
nter.readline()[1:].
>temp_652|source:-*- coding: utf-8 -*-|start:2540|stop:3040
 [i[1] for i in sorted([(self.unique_hash_dict[hash]['count'], hash)\for hash in
 self.unique_hash_dict], reverse = True)]self.total_unique = len(self.unique_has
h_dict)self.reset()def next(self):if self.unique:return self.next_unique()else:r
eturn self.next_regular()def next_unique(self):if self.unique:if self.total_uniq
ue > 0 and self.pos < self.total_unique:hash_entry = self.unique_hash_dict[self.
unique_hash_list[self.pos]]self.pos += 1self.seq = hash_entry['seq'] if self.all
ow_mixed_case else h
>temp_653|source:-*- coding: utf-8 -*-|start:6697|stop:7197
):self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.quals.sp
lit()]self.pos += 1return Trueelse:return Falseif line.startswith('>'):self.file
_pointer.seek(self.file_pointer.tell() - len(line))breakqualscores += ' ' + line
.strip()self.quals = qualscores.strip()self.quals_int = [int(q) for q in self.qu
als.split()]self.pos += 1return Truedef close(self):self.file_pointer.close()def
 reset(self):self.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Nonese
lf.ids = []self.file
>temp_654|source:-*- coding: utf-8 -*-|start:3004|stop:3504
eq'] if self.allow_mixed_case else hash_entry['seq'].upper()self.id  = hash_entr
y['id']self.ids = hash_entry['ids']return Trueelse:return Falseelse:return False
def next_regular(self):self.seq = Noneself.id = self.file_pointer.readline()[1:]
.strip()sequence = ''while 1:line = self.file_pointer.readline()if not line:if l
en(sequence):self.seq = sequenceself.pos += 1return Trueelse:return Falseif line
.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))bre
aksequence += line.s
>temp_655|source:-*- coding: utf-8 -*-|start:2569|stop:3069
.unique_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_dict], reve
rse = True)]self.total_unique = len(self.unique_hash_dict)self.reset()def next(s
elf):if self.unique:return self.next_unique()else:return self.next_regular()def 
next_unique(self):if self.unique:if self.total_unique > 0 and self.pos < self.to
tal_unique:hash_entry = self.unique_hash_dict[self.unique_hash_list[self.pos]]se
lf.pos += 1self.seq = hash_entry['seq'] if self.allow_mixed_case else hash_entry
['seq'].upper()self.
>temp_656|source:-*- coding: utf-8 -*-|start:5372|stop:5872
x_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(seq_len_dis
tribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'black', size
 = 'xx-large', ha = 'center')ax1 = plt.subplot(gs[9])plt.rcParams.update({'axes.
edgecolor' : 20})plt.grid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 
'total: %s / mean: %.2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths
),numpy.mean(sequence_lengths), numpy.std(sequence_lengths),\min(sequence_length
s),\max(sequence_len
>temp_657|source:-*- coding: utf-8 -*-|start:1015|stop:1515
)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks)
 - 1)])def close(self):self.output_file_obj.close()class ReadFasta:def __init__(
self, f_name):self.ids = []self.sequences = []self.fasta = SequenceSource(f_name
)while self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sy
s.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))s
ys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.
seq)sys.stderr.write
>temp_658|source:-*- coding: utf-8 -*-|start:969|stop:1469
, len(sequence), piece_length) + [len(sequence)]return '\n'.join([sequence[ticks
[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(self):self.output
_file_obj.close()class ReadFasta:def __init__(self, f_name):self.ids = []self.se
quences = []self.fasta = SequenceSource(f_name)while self.fasta.next():if self.f
asta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] Readi
ng FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.append(s
elf.fasta.id)self.se
>temp_659|source:-*- coding: utf-8 -*-|start:6627|stop:7127
ile 1:line = self.file_pointer.readline()if not line:if len(qualscores):self.qua
ls = qualscores.strip()self.quals_int = [int(q) for q in self.quals.split()]self
.pos += 1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.s
eek(self.file_pointer.tell() - len(line))breakqualscores += ' ' + line.strip()se
lf.quals = qualscores.strip()self.quals_int = [int(q) for q in self.quals.split(
)]self.pos += 1return Truedef close(self):self.file_pointer.close()def reset(sel
f):self.pos = 0self.
>temp_660|source:-*- coding: utf-8 -*-|start:6149|stop:6649
path = quals_file_pathself.name = Noneself.lazy_init = lazy_initself.pos = 0self
.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []self.file_pointer 
= open(self.quals_file_path)self.file_pointer.seek(0)if self.lazy_init:self.tota
l_quals = Noneelse:self.total_quals = len([l for l in self.file_pointer.readline
s() if l.startswith('>')])self.reset()def next(self):self.id = self.file_pointer
.readline()[1:].strip()self.quals = Noneself.quals_int = Nonequalscores = ''whil
e 1:line = self.file
>temp_661|source:-*- coding: utf-8 -*-|start:4032|stop:4532
spec as gridspecsequence_lengths = []self.reset()while self.next():if self.pos %
 10000 == 0 or self.pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self
.pos))sys.stderr.flush()sequence_lengths.append(len(self.seq))self.reset()sys.st
derr.write('\n')if not max_seq_len:max_seq_len = max(sequence_lengths) + (int(ma
x(sequence_lengths) / 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1
)for l in sequence_lengths:seq_len_distribution[l] += 1fig = plt.figure(figsize 
= (16, 12))plt.rcPar
>temp_662|source:-*- coding: utf-8 -*-|start:4862|stop:5362
n + 1), seq_len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('
number of sequences')plt.xlabel('sequence length')if xtickstep == None:xtickstep
 = (max_seq_len / 50) or 1if ytickstep == None:ytickstep = max(seq_len_distribut
ion) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1, xtickstep), rotation=
90, size='xx-small')plt.yticks(range(0, max(seq_len_distribution) + 1, ytickstep
),[y for y in range(0, max(seq_len_distribution) + 1, ytickstep)],size='xx-small
')plt.xlim(xmin = 0,
>temp_663|source:-*- coding: utf-8 -*-|start:4824|stop:5324
3)plt.fill_between(range(0, max_seq_len + 1), seq_len_distribution, y2 = 0, colo
r = 'black', alpha = 0.15)plt.ylabel('number of sequences')plt.xlabel('sequence 
length')if xtickstep == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == 
None:ytickstep = max(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, 
max_seq_len + 1, xtickstep), rotation=90, size='xx-small')plt.yticks(range(0, ma
x(seq_len_distribution) + 1, ytickstep),[y for y in range(0, max(seq_len_distrib
ution) + 1, ytickste
>temp_664|source:-*- coding: utf-8 -*-|start:3217|stop:3717
file_pointer.readline()[1:].strip()sequence = ''while 1:line = self.file_pointer
.readline()if not line:if len(sequence):self.seq = sequenceself.pos += 1return T
rueelse:return Falseif line.startswith('>'):self.file_pointer.seek(self.file_poi
nter.tell() - len(line))breaksequence += line.strip()self.seq = sequence if self
.allow_mixed_case else sequence.upper()self.pos += 1return Truedef get_seq_by_re
ad_id(self, read_id):self.reset()while self.next():if self.id == read_id:return 
self.seqreturn False
>temp_665|source:-*- coding: utf-8 -*-|start:1094|stop:1594
) - 1)])def close(self):self.output_file_obj.close()class ReadFasta:def __init__
(self, f_name):self.ids = []self.sequences = []self.fasta = SequenceSource(f_nam
e)while self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:s
ys.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))
sys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(self.fasta
.seq)sys.stderr.write('\n')def close(self):self.fasta.close()class SequenceSourc
e():def __init__(sel
>temp_666|source:-*- coding: utf-8 -*-|start:3349|stop:3849
equenceself.pos += 1return Trueelse:return Falseif line.startswith('>'):self.fil
e_pointer.seek(self.file_pointer.tell() - len(line))breaksequence += line.strip(
)self.seq = sequence if self.allow_mixed_case else sequence.upper()self.pos += 1
return Truedef get_seq_by_read_id(self, read_id):self.reset()while self.next():i
f self.id == read_id:return self.seqreturn Falsedef close(self):self.file_pointe
r.close()def reset(self):self.pos = 0self.id  = Noneself.seq = Noneself.ids = []
self.file_pointer.se
>temp_667|source:-*- coding: utf-8 -*-|start:5003|stop:5503
p == None:xtickstep = (max_seq_len / 50) or 1if ytickstep == None:ytickstep = ma
x(seq_len_distribution) / 20 or 1plt.xticks(range(xtickstep, max_seq_len + 1, xt
ickstep), rotation=90, size='xx-small')plt.yticks(range(0, max(seq_len_distribut
ion) + 1, ytickstep),[y for y in range(0, max(seq_len_distribution) + 1, ytickst
ep)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, ym
ax = max(seq_len_distribution) + (max(seq_len_distribution) / 20.0))plt.figtext(
0.5, 0.96, '%s' % (t
>temp_668|source:-*- coding: utf-8 -*-|start:3908|stop:4408
, dest = None, max_seq_len = None, xtickstep = None, ytickstep = None):import ma
tplotlib.pyplot as pltimport matplotlib.gridspec as gridspecsequence_lengths = [
]self.reset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.std
err.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_len
gths.append(len(self.seq))self.reset()sys.stderr.write('\n')if not max_seq_len:m
ax_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or 10)s
eq_len_distribution 
>temp_669|source:-*- coding: utf-8 -*-|start:1029|stop:1529
join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def clo
se(self):self.output_file_obj.close()class ReadFasta:def __init__(self, f_name):
self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)while self.fa
sta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write
('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flus
h()self.ids.append(self.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr
.write('\n')def clos
>temp_670|source:-*- coding: utf-8 -*-|start:3914|stop:4414
 = None, max_seq_len = None, xtickstep = None, ytickstep = None):import matplotl
ib.pyplot as pltimport matplotlib.gridspec as gridspecsequence_lengths = []self.
reset()while self.next():if self.pos % 10000 == 0 or self.pos == 1:sys.stderr.wr
ite('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flush()sequence_lengths.a
ppend(len(self.seq))self.reset()sys.stderr.write('\n')if not max_seq_len:max_seq
_len = max(sequence_lengths) + (int(max(sequence_lengths) / 100.0) or 10)seq_len
_distribution = [0] 
>temp_671|source:-*- coding: utf-8 -*-|start:529|stop:1029
, split = True, store_frequencies = True):if entry.unique and store_frequencies:
self.write_id('%s|%s' % (entry.id, 'frequency:%d' % len(entry.ids)))else:self.wr
ite_id(entry.id)self.write_seq(entry.seq, split)def write_id(self, id):self.outp
ut_file_obj.write('>%s\n' % id)def write_seq(self, seq, split = True):if split:s
eq = self.split(seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequ
ence, piece_length = 80):ticks = range(0, len(sequence), piece_length) + [len(se
quence)]return '\n'.
>temp_672|source:-*- coding: utf-8 -*-|start:5244|stop:5744
on) + 1, ytickstep),[y for y in range(0, max(seq_len_distribution) + 1, ytickste
p)],size='xx-small')plt.xlim(xmin = 0, xmax = max_seq_len)plt.ylim(ymin = 0, yma
x = max(seq_len_distribution) + (max(seq_len_distribution) / 20.0))plt.figtext(0
.5, 0.96, '%s' % (title), weight = 'black', size = 'xx-large', ha = 'center')ax1
 = plt.subplot(gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)
plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %
.2f / min: %s / max:
>temp_673|source:-*- coding: utf-8 -*-|start:953|stop:1453
:ticks = range(0, len(sequence), piece_length) + [len(sequence)]return '\n'.join
([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def close(s
elf):self.output_file_obj.close()class ReadFasta:def __init__(self, f_name):self
.ids = []self.sequences = []self.fasta = SequenceSource(f_name)while self.fasta.
next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r
[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()s
elf.ids.append(self.
>temp_674|source:-*- coding: utf-8 -*-|start:5041|stop:5541
0) or 1if ytickstep == None:ytickstep = max(seq_len_distribution) / 20 or 1plt.x
ticks(range(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx-small'
)plt.yticks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y in range
(0, max(seq_len_distribution) + 1, ytickstep)],size='xx-small')plt.xlim(xmin = 0
, xmax = max_seq_len)plt.ylim(ymin = 0, ymax = max(seq_len_distribution) + (max(
seq_len_distribution) / 20.0))plt.figtext(0.5, 0.96, '%s' % (title), weight = 'b
lack', size = 'xx-la
>temp_675|source:-*- coding: utf-8 -*-|start:710|stop:1210
write_seq(entry.seq, split)def write_id(self, id):self.output_file_obj.write('>%
s\n' % id)def write_seq(self, seq, split = True):if split:seq = self.split(seq)s
elf.output_file_obj.write('%s\n' % seq)def split(self, sequence, piece_length = 
80):ticks = range(0, len(sequence), piece_length) + [len(sequence)]return '\n'.j
oin([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1)])def clos
e(self):self.output_file_obj.close()class ReadFasta:def __init__(self, f_name):s
elf.ids = []self.seq
>temp_676|source:-*- coding: utf-8 -*-|start:1772|stop:2272
ase = allow_mixed_caseself.pos = 0self.id  = Noneself.seq = Noneself.ids = []sel
f.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list = []self.unique
_next_hash = 0self.file_pointer = open(self.fasta_file_path)self.file_pointer.se
ek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for l i
n self.file_pointer.readlines() if l.startswith('>')])self.reset()if self.unique
:self.init_unique_hash()def init_unique_hash(self):while self.next_regular():has
h = hashlib.sha1(sel
>temp_677|source:-*- coding: utf-8 -*-|start:4132|stop:4632
pos == 1:sys.stderr.write('\r[fastalib] Reading: %s' % (self.pos))sys.stderr.flu
sh()sequence_lengths.append(len(self.seq))self.reset()sys.stderr.write('\n')if n
ot max_seq_len:max_seq_len = max(sequence_lengths) + (int(max(sequence_lengths) 
/ 100.0) or 10)seq_len_distribution = [0] * (max_seq_len + 1)for l in sequence_l
engths:seq_len_distribution[l] += 1fig = plt.figure(figsize = (16, 12))plt.rcPar
ams.update({'axes.linewidth' : 0.9})plt.rc('grid', color='0.50', linestyle='-', 
linewidth=0.1)gs = g
>temp_678|source:-*- coding: utf-8 -*-|start:6249|stop:6749
s = Noneself.quals_int = Noneself.ids = []self.file_pointer = open(self.quals_fi
le_path)self.file_pointer.seek(0)if self.lazy_init:self.total_quals = Noneelse:s
elf.total_quals = len([l for l in self.file_pointer.readlines() if l.startswith(
'>')])self.reset()def next(self):self.id = self.file_pointer.readline()[1:].stri
p()self.quals = Noneself.quals_int = Nonequalscores = ''while 1:line = self.file
_pointer.readline()if not line:if len(qualscores):self.quals = qualscores.strip(
)self.quals_int = [i
>temp_679|source:-*- coding: utf-8 -*-|start:19|stop:519
ht (C) 2014, A. Murat Eren## This program is free software; you can redistribute
 it and/or modify it under# the terms of the GNU General Public License as publi
shed by the Free# Software Foundation; either version 2 of the License, or (at y
our option)# any later version.## Please read the docs/COPYING file.import sysim
port numpyimport hashlibclass FastaOutput:def __init__(self, output_file_path):s
elf.output_file_path = output_file_pathself.output_file_obj = open(output_file_p
ath, 'w')def store(s
>temp_680|source:-*- coding: utf-8 -*-|start:1849|stop:2349
self.unique = uniqueself.unique_hash_dict = {}self.unique_hash_list = []self.uni
que_next_hash = 0self.file_pointer = open(self.fasta_file_path)self.file_pointer
.seek(0)if self.lazy_init:self.total_seq = Noneelse:self.total_seq = len([l for 
l in self.file_pointer.readlines() if l.startswith('>')])self.reset()if self.uni
que:self.init_unique_hash()def init_unique_hash(self):while self.next_regular():
hash = hashlib.sha1(self.seq.upper()).hexdigest()if hash in self.unique_hash_dic
t:self.unique_hash_d
>temp_681|source:-*- coding: utf-8 -*-|start:6391|stop:6891
quals = Noneelse:self.total_quals = len([l for l in self.file_pointer.readlines(
) if l.startswith('>')])self.reset()def next(self):self.id = self.file_pointer.r
eadline()[1:].strip()self.quals = Noneself.quals_int = Nonequalscores = ''while 
1:line = self.file_pointer.readline()if not line:if len(qualscores):self.quals =
 qualscores.strip()self.quals_int = [int(q) for q in self.quals.split()]self.pos
 += 1return Trueelse:return Falseif line.startswith('>'):self.file_pointer.seek(
self.file_pointer.te
>temp_682|source:-*- coding: utf-8 -*-|start:4794|stop:5294
n, color = 'black', alpha = 0.3)plt.fill_between(range(0, max_seq_len + 1), seq_
len_distribution, y2 = 0, color = 'black', alpha = 0.15)plt.ylabel('number of se
quences')plt.xlabel('sequence length')if xtickstep == None:xtickstep = (max_seq_
len / 50) or 1if ytickstep == None:ytickstep = max(seq_len_distribution) / 20 or
 1plt.xticks(range(xtickstep, max_seq_len + 1, xtickstep), rotation=90, size='xx
-small')plt.yticks(range(0, max(seq_len_distribution) + 1, ytickstep),[y for y i
n range(0, max(seq_l
>temp_683|source:-*- coding: utf-8 -*-|start:3250|stop:3750
()sequence = ''while 1:line = self.file_pointer.readline()if not line:if len(seq
uence):self.seq = sequenceself.pos += 1return Trueelse:return Falseif line.start
swith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))breaksequ
ence += line.strip()self.seq = sequence if self.allow_mixed_case else sequence.u
pper()self.pos += 1return Truedef get_seq_by_read_id(self, read_id):self.reset()
while self.next():if self.id == read_id:return self.seqreturn Falsedef close(sel
f):self.file_pointer
>temp_684|source:-*- coding: utf-8 -*-|start:2479|stop:2979
 [self.id],'seq': self.seq,'count': 1}self.unique_hash_list = [i[1] for i in sor
ted([(self.unique_hash_dict[hash]['count'], hash)\for hash in self.unique_hash_d
ict], reverse = True)]self.total_unique = len(self.unique_hash_dict)self.reset()
def next(self):if self.unique:return self.next_unique()else:return self.next_reg
ular()def next_unique(self):if self.unique:if self.total_unique > 0 and self.pos
 < self.total_unique:hash_entry = self.unique_hash_dict[self.unique_hash_list[se
lf.pos]]self.pos += 
>temp_685|source:-*- coding: utf-8 -*-|start:3090|stop:3590
]self.ids = hash_entry['ids']return Trueelse:return Falseelse:return Falsedef ne
xt_regular(self):self.seq = Noneself.id = self.file_pointer.readline()[1:].strip
()sequence = ''while 1:line = self.file_pointer.readline()if not line:if len(seq
uence):self.seq = sequenceself.pos += 1return Trueelse:return Falseif line.start
swith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(line))breaksequ
ence += line.strip()self.seq = sequence if self.allow_mixed_case else sequence.u
pper()self.pos += 1r
>temp_686|source:-*- coding: utf-8 -*-|start:6644|stop:7144
.file_pointer.readline()if not line:if len(qualscores):self.quals = qualscores.s
trip()self.quals_int = [int(q) for q in self.quals.split()]self.pos += 1return T
rueelse:return Falseif line.startswith('>'):self.file_pointer.seek(self.file_poi
nter.tell() - len(line))breakqualscores += ' ' + line.strip()self.quals = qualsc
ores.strip()self.quals_int = [int(q) for q in self.quals.split()]self.pos += 1re
turn Truedef close(self):self.file_pointer.close()def reset(self):self.pos = 0se
lf.id  = Noneself.qu
>temp_687|source:-*- coding: utf-8 -*-|start:6353|stop:6853
r.seek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l
 for l in self.file_pointer.readlines() if l.startswith('>')])self.reset()def ne
xt(self):self.id = self.file_pointer.readline()[1:].strip()self.quals = Noneself
.quals_int = Nonequalscores = ''while 1:line = self.file_pointer.readline()if no
t line:if len(qualscores):self.quals = qualscores.strip()self.quals_int = [int(q
) for q in self.quals.split()]self.pos += 1return Trueelse:return Falseif line.s
tartswith('>'):self.
>temp_688|source:-*- coding: utf-8 -*-|start:664|stop:1164
(entry.ids)))else:self.write_id(entry.id)self.write_seq(entry.seq, split)def wri
te_id(self, id):self.output_file_obj.write('>%s\n' % id)def write_seq(self, seq,
 split = True):if split:seq = self.split(seq)self.output_file_obj.write('%s\n' %
 seq)def split(self, sequence, piece_length = 80):ticks = range(0, len(sequence)
, piece_length) + [len(sequence)]return '\n'.join([sequence[ticks[x]:ticks[x + 1
]] for x in range(0, len(ticks) - 1)])def close(self):self.output_file_obj.close
()class ReadFasta:de
>temp_689|source:-*- coding: utf-8 -*-|start:6037|stop:6537
t.show()except:passreturnclass QualSource:def __init__(self, quals_file_path, la
zy_init = True):self.quals_file_path = quals_file_pathself.name = Noneself.lazy_
init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Non
eself.ids = []self.file_pointer = open(self.quals_file_path)self.file_pointer.se
ek(0)if self.lazy_init:self.total_quals = Noneelse:self.total_quals = len([l for
 l in self.file_pointer.readlines() if l.startswith('>')])self.reset()def next(s
elf):self.id = self.
>temp_690|source:-*- coding: utf-8 -*-|start:6505|stop:7005
()def next(self):self.id = self.file_pointer.readline()[1:].strip()self.quals = 
Noneself.quals_int = Nonequalscores = ''while 1:line = self.file_pointer.readlin
e()if not line:if len(qualscores):self.quals = qualscores.strip()self.quals_int 
= [int(q) for q in self.quals.split()]self.pos += 1return Trueelse:return Falsei
f line.startswith('>'):self.file_pointer.seek(self.file_pointer.tell() - len(lin
e))breakqualscores += ' ' + line.strip()self.quals = qualscores.strip()self.qual
s_int = [int(q) for 
>temp_691|source:-*- coding: utf-8 -*-|start:3482|stop:3982
reaksequence += line.strip()self.seq = sequence if self.allow_mixed_case else se
quence.upper()self.pos += 1return Truedef get_seq_by_read_id(self, read_id):self
.reset()while self.next():if self.id == read_id:return self.seqreturn Falsedef c
lose(self):self.file_pointer.close()def reset(self):self.pos = 0self.id  = Nones
elf.seq = Noneself.ids = []self.file_pointer.seek(0)def visualize_sequence_lengt
h_distribution(self, title, dest = None, max_seq_len = None, xtickstep = None, y
tickstep = None):imp
>temp_692|source:-*- coding: utf-8 -*-|start:5633|stop:6133
grid(False)plt.yticks([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.
2f / std: %.2f / min: %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence
_lengths), numpy.std(sequence_lengths),\min(sequence_lengths),\max(sequence_leng
ths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == None:dest = self.
fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')tr
y:plt.show()except:passreturnclass QualSource:def __init__(self, quals_file_path
, lazy_init = True):
>temp_693|source:-*- coding: utf-8 -*-|start:6131|stop:6631
):self.quals_file_path = quals_file_pathself.name = Noneself.lazy_init = lazy_in
itself.pos = 0self.id  = Noneself.quals = Noneself.quals_int = Noneself.ids = []
self.file_pointer = open(self.quals_file_path)self.file_pointer.seek(0)if self.l
azy_init:self.total_quals = Noneelse:self.total_quals = len([l for l in self.fil
e_pointer.readlines() if l.startswith('>')])self.reset()def next(self):self.id =
 self.file_pointer.readline()[1:].strip()self.quals = Noneself.quals_int = Noneq
ualscores = ''while 
>temp_694|source:-*- coding: utf-8 -*-|start:779|stop:1279
j.write('>%s\n' % id)def write_seq(self, seq, split = True):if split:seq = self.
split(seq)self.output_file_obj.write('%s\n' % seq)def split(self, sequence, piec
e_length = 80):ticks = range(0, len(sequence), piece_length) + [len(sequence)]re
turn '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks) - 1
)])def close(self):self.output_file_obj.close()class ReadFasta:def __init__(self
, f_name):self.ids = []self.sequences = []self.fasta = SequenceSource(f_name)whi
le self.fasta.next()
>temp_695|source:-*- coding: utf-8 -*-|start:5870|stop:6370
engths)),\va = 'center', alpha = 0.8, size = 'x-large')if dest == None:dest = se
lf.fasta_file_pathtry:plt.savefig(dest + '.pdf')except:plt.savefig(dest + '.png'
)try:plt.show()except:passreturnclass QualSource:def __init__(self, quals_file_p
ath, lazy_init = True):self.quals_file_path = quals_file_pathself.name = Nonesel
f.lazy_init = lazy_initself.pos = 0self.id  = Noneself.quals = Noneself.quals_in
t = Noneself.ids = []self.file_pointer = open(self.quals_file_path)self.file_poi
nter.seek(0)if self.
>temp_696|source:-*- coding: utf-8 -*-|start:1284|stop:1784
elf.fasta.pos % 1000 == 0 or self.fasta.pos == 1:sys.stderr.write('\r[fastalib] 
Reading FASTA into memory: %s' % (self.fasta.pos))sys.stderr.flush()self.ids.app
end(self.fasta.id)self.sequences.append(self.fasta.seq)sys.stderr.write('\n')def
 close(self):self.fasta.close()class SequenceSource():def __init__(self, fasta_f
ile_path, lazy_init = True, unique = False, allow_mixed_case = False):self.fasta
_file_path = fasta_file_pathself.name = Noneself.lazy_init = lazy_initself.allow
_mixed_case = allow_
>temp_697|source:-*- coding: utf-8 -*-|start:5574|stop:6074
plot(gs[9])plt.rcParams.update({'axes.edgecolor' : 20})plt.grid(False)plt.yticks
([])plt.xticks([])plt.text(0.02, 0.5, 'total: %s / mean: %.2f / std: %.2f / min:
 %s / max: %s'\% (len(sequence_lengths),numpy.mean(sequence_lengths), numpy.std(
sequence_lengths),\min(sequence_lengths),\max(sequence_lengths)),\va = 'center',
 alpha = 0.8, size = 'x-large')if dest == None:dest = self.fasta_file_pathtry:pl
t.savefig(dest + '.pdf')except:plt.savefig(dest + '.png')try:plt.show()except:pa
ssreturnclass QualSo
>temp_698|source:-*- coding: utf-8 -*-|start:1560|stop:2060
 SequenceSource():def __init__(self, fasta_file_path, lazy_init = True, unique =
 False, allow_mixed_case = False):self.fasta_file_path = fasta_file_pathself.nam
e = Noneself.lazy_init = lazy_initself.allow_mixed_case = allow_mixed_caseself.p
os = 0self.id  = Noneself.seq = Noneself.ids = []self.unique = uniqueself.unique
_hash_dict = {}self.unique_hash_list = []self.unique_next_hash = 0self.file_poin
ter = open(self.fasta_file_path)self.file_pointer.seek(0)if self.lazy_init:self.
total_seq = Noneelse
>temp_699|source:-*- coding: utf-8 -*-|start:1014|stop:1514
e)]return '\n'.join([sequence[ticks[x]:ticks[x + 1]] for x in range(0, len(ticks
) - 1)])def close(self):self.output_file_obj.close()class ReadFasta:def __init__
(self, f_name):self.ids = []self.sequences = []self.fasta = SequenceSource(f_nam
e)while self.fasta.next():if self.fasta.pos % 1000 == 0 or self.fasta.pos == 1:s
ys.stderr.write('\r[fastalib] Reading FASTA into memory: %s' % (self.fasta.pos))
sys.stderr.flush()self.ids.append(self.fasta.id)self.sequences.append(self.fasta
.seq)sys.stderr.writ