def get_str_qualities(seq, out_format=None):
if out_format is None:
out_format = seq.file_format
out_format = remove_multiline(out_format)
if out_format in SANGER_FASTQ_FORMATS:
out_format = SANGER_QUALITY
elif out_format in ILLUMINA_FASTQ_FORMATS:
out_format = ILLUMINA_QUALITY
seq_class = seq.kind
if seq_class == SEQITEM:
in_format = remove_multiline(seq.file_format)
if 'fasta' in in_format:
raise ValueError('A fasta file has no qualities')
if in_format in SANGER_FASTQ_FORMATS:
in_format = SANGER_QUALITY
elif in_format in ILLUMINA_FASTQ_FORMATS:
in_format = ILLUMINA_QUALITY
else:
msg = 'Unknown or not supported quality format: '
msg += in_format
raise ValueError(msg)
if in_format == out_format:
quals = ''.join(line.rstrip() for line in _get_seqitem_qual_lines(seq))
else:
int_quals = get_int_qualities(seq)
quals = ''.join(_int_quals_to_str_quals(int_quals, out_format))
elif seq_class == SEQRECORD:
int_quals = get_int_qualities(seq)
quals = ''.join(_int_quals_to_str_quals(int_quals, out_format))
return quals
def _write_filter_trim_packets(passed_fhand, diverted_fhand, packets,
file_format='fastq', workers=None,
seqs_diverted=SEQS_FILTERED_OUT):
'It writes the filter stream into passed and filtered out sequence files'
file_format = remove_multiline(file_format)
if diverted_fhand is None:
seq_packets = (p[SEQS_PASSED] for p in packets)
seqs = (s for pair in chain.from_iterable(seq_packets) for s in pair)
try:
return write_seqs(seqs, passed_fhand, file_format=file_format)
except BaseException:
if workers is not None:
workers.terminate()
raise
flatten_pairs = lambda pairs: (seq for pair in pairs for seq in pair)
for packet in packets:
try:
write_seqs(flatten_pairs(packet[SEQS_PASSED]), fhand=passed_fhand,
file_format=file_format)
# if diverted seqs are filtered aout they are a list of list
# as not diverted seqs.
# if they are orphan, they are a list of seqs
if seqs_diverted == SEQS_FILTERED_OUT :
seqs = flatten_pairs(packet[seqs_diverted])
else:
seqs = packet[seqs_diverted]
write_seqs(seqs, fhand=diverted_fhand, file_format=file_format)
except BaseException:
if workers is not None:
workers.terminate()
raise
def _read_seqrecords(fhands, file_format=GUESS_FORMAT):
'It returns an iterator of seqrecords'
seq_iters = []
for fhand in fhands:
if file_format == GUESS_FORMAT or file_format is None:
fmt = guess_format(fhand)
else:
fmt = file_format
fmt = remove_multiline(fmt)
if fmt in ('fasta', 'qual') or 'fastq' in fmt:
title = title2ids
if fmt == 'fasta':
seq_iter = FastaIterator(fhand, title2ids=title)
elif fmt == 'qual':
seq_iter = QualPhredIterator(fhand, title2ids=title)
elif fmt == 'fastq' or fmt == 'fastq-sanger':
seq_iter = FastqPhredIterator(fhand, title2ids=title)
elif fmt == 'fastq-solexa':
seq_iter = FastqSolexaIterator(fhand, title2ids=title)
elif fmt == 'fastq-illumina':
seq_iter = FastqIlluminaIterator(fhand, title2ids=title)
else:
seq_iter = parse_into_seqrecs(fhand, fmt)
seq_iters.append(seq_iter)
return chain.from_iterable(seq_iters)
def write_filter_packets(passed_fhand, filtered_fhand, filter_packets,
file_format='fastq', workers=None):
'It writes the filter stream into passed and filtered out sequence files'
file_format = remove_multiline(file_format)
if filtered_fhand is None:
seq_packets = (p[SEQS_PASSED] for p in filter_packets)
seqs = (s for pair in chain.from_iterable(seq_packets) for s in pair)
try:
return write_seqs(seqs, passed_fhand, file_format=file_format)
except BaseException:
if workers is not None:
workers.terminate()
raise
flatten_pairs = lambda pairs: (seq for pair in pairs for seq in pair)
for packet in filter_packets:
try:
write_seqs(flatten_pairs(packet[SEQS_PASSED]), fhand=passed_fhand,
file_format=file_format)
write_seqs(flatten_pairs(packet[SEQS_FILTERED_OUT]),
fhand=filtered_fhand, file_format=file_format)
except BaseException:
if workers is not None:
workers.terminate()
raise
def _read_seqrecords(fhands, file_format=GUESS_FORMAT):
'It returns an iterator of seqrecords'
seq_iters = []
for fhand in fhands:
if file_format == GUESS_FORMAT or file_format is None:
fmt = guess_format(fhand)
else:
fmt = file_format
fmt = remove_multiline(fmt)
if fmt in ('fasta', 'qual') or 'fastq' in fmt:
title = title2ids
if fmt == 'fasta':
seq_iter = FastaIterator(fhand, title2ids=title)
elif fmt == 'qual':
seq_iter = QualPhredIterator(fhand, title2ids=title)
elif fmt == 'fastq' or fmt == 'fastq-sanger':
seq_iter = FastqPhredIterator(fhand, title2ids=title)
elif fmt == 'fastq-solexa':
seq_iter = FastqSolexaIterator(fhand, title2ids=title)
elif fmt == 'fastq-illumina':
seq_iter = FastqIlluminaIterator(fhand, title2ids=title)
else:
seq_iter = parse_into_seqrecs(fhand, fmt)
seq_iters.append(seq_iter)
return chain.from_iterable(seq_iters)
def _index_seq_file(fpath, file_format=None):
'''It indexes a seq file using Biopython index.
It uses the title line line as the key and not just the id.
'''
if file_format is None:
file_format = guess_format(open(fpath))
file_format = remove_multiline(file_format)
# pylint: disable W0212
# we monkey patch to be able to index using the whole tile line and not
# only the id. We need it because in a pair end file sequences with the
# same id could be found
accessor = _index._FormatToRandomAccess
old_accessor = accessor.copy()
accessor['fastq'] = FastqRandomAccess
accessor['astq-sanger'] = FastqRandomAccess
accessor['fastq-solexa'] = FastqRandomAccess
accessor['fastq-illumina'] = FastqRandomAccess
file_index = index(fpath, format=file_format)
_index._FormatToRandomAccess = old_accessor
return file_index
def _index_seq_file(fpath, file_format=None):
'''It indexes a seq file using Biopython index.
It uses the title line line as the key and not just the id.
'''
if file_format is None:
file_format = guess_format(open(fpath))
file_format = remove_multiline(file_format)
# pylint: disable W0212
# we monkey patch to be able to index using the whole tile line and not
# only the id. We need it because in a pair end file sequences with the
# same id could be found
accessor = _index._FormatToRandomAccess
old_accessor = accessor.copy()
accessor['fastq'] = FastqRandomAccess
accessor['astq-sanger'] = FastqRandomAccess
accessor['fastq-solexa'] = FastqRandomAccess
accessor['fastq-illumina'] = FastqRandomAccess
file_index = index(fpath, format=file_format)
_index._FormatToRandomAccess = old_accessor
return file_index
def write_filter_packets(passed_fhand,
filtered_fhand,
filter_packets,
file_format='fastq',
workers=None):
'It writes the filter stream into passed and filtered out sequence files'
file_format = remove_multiline(file_format)
if filtered_fhand is None:
seq_packets = (p[SEQS_PASSED] for p in filter_packets)
seqs = (s for pair in chain.from_iterable(seq_packets) for s in pair)
try:
return write_seqs(seqs, passed_fhand, file_format=file_format)
except BaseException:
if workers is not None:
workers.terminate()
raise
flatten_pairs = lambda pairs: (seq for pair in pairs for seq in pair)
for packet in filter_packets:
try:
write_seqs(flatten_pairs(packet[SEQS_PASSED]),
fhand=passed_fhand,
file_format=file_format)
write_seqs(flatten_pairs(packet[SEQS_FILTERED_OUT]),
fhand=filtered_fhand,
file_format=file_format)
except BaseException:
if workers is not None:
workers.terminate()
raise
def slice_seq(seq, start=None, stop=None):
seq_class = seq.kind
if seq_class == SEQITEM:
seq_obj = _slice_seqitem(seq, start, stop)
elif seq_class == SEQRECORD:
seq_obj = seq.object[start:stop]
return SeqWrapper(seq.kind, object=seq_obj,
file_format=remove_multiline(seq.file_format))
def write_seq_packets(fhand, seq_packets, file_format='fastq', workers=None):
'It writes to file a stream of seq lists'
file_format = remove_multiline(file_format)
try:
write_seqs(chain.from_iterable(seq_packets), fhand,
file_format=file_format)
except BaseException:
if workers is not None:
workers.terminate()
raise
def write_seq_packets(fhand, seq_packets, file_format='fastq', workers=None):
'It writes to file a stream of seq lists'
file_format = remove_multiline(file_format)
try:
write_seqs(chain.from_iterable(seq_packets),
fhand,
file_format=file_format)
except BaseException:
if workers is not None:
workers.terminate()
raise
def _write_seqrecords(seqs, fhand=None, file_format='fastq'):
'It writes a stream of sequences to a file'
file_format = remove_multiline(file_format)
if fhand is None:
fhand = NamedTemporaryFile(suffix='.' + file_format.replace('-', '_'))
seqs = _clean_seqrecord_stream(seqs)
try:
write_seqrecs(seqs, fhand, file_format)
except IOError, error:
# The pipe could be already closed
if not 'Broken pipe' in str(error):
raise
def _write_seqrecords(seqs, fhand=None, file_format='fastq'):
'It writes a stream of sequences to a file'
file_format = remove_multiline(file_format)
if fhand is None:
fhand = NamedTemporaryFile(suffix='.' + file_format.replace('-', '_'))
seqs = _clean_seqrecord_stream(seqs)
try:
write_seqrecs(seqs, fhand, file_format)
except IOError, error:
# The pipe could be already closed
if not 'Broken pipe' in str(error):
raise
def _write_seqitems(items, fhand, file_format):
'It writes one seq item (tuple of name and string)'
for seq in items:
seqitems_fmt = remove_multiline(seq.file_format)
if file_format and 'fastq' in seqitems_fmt and 'fasta' in file_format:
seq_lines = seq.object.lines
try:
fhand.write('>' + seq_lines[0][1:] + seq_lines[1])
except IOError, error:
# The pipe could be already closed
if not 'Broken pipe' in str(error):
raise
elif file_format and seqitems_fmt != file_format:
msg = 'Input and output file formats do not match, you should not '
msg += 'use SeqItems: ' + str(seq.file_format) + ' '
msg += str(file_format)
raise RuntimeError(msg)
def _write_seqitems(items, fhand, file_format):
'It writes one seq item (tuple of name and string)'
for seq in items:
seqitems_fmt = remove_multiline(seq.file_format)
if file_format and 'fastq' in seqitems_fmt and 'fasta' in file_format:
seq_lines = seq.object.lines
try:
fhand.write('>' + seq_lines[0][1:] + seq_lines[1])
except IOError, error:
# The pipe could be already closed
if not 'Broken pipe' in str(error):
raise
elif file_format and seqitems_fmt != file_format:
msg = 'Input and output file formats do not match, you should not '
msg += 'use SeqItems: ' + str(seq.file_format) + ' '
msg += str(file_format)
raise RuntimeError(msg)
def _copy_seqitem(seqwrapper, seq=None, name=None):
seq_item = seqwrapper.object
lines = seq_item.lines
fmt = seqwrapper.file_format
if seq is None:
lines = lines[:]
else:
if 'fasta' in fmt:
lines = [lines[0], seq + '\n']
elif 'multiline' in fmt and 'fastq' in fmt:
qline = ''.join([qline.strip() for qline in _get_seqitem_qual_lines(seqwrapper)])
lines = [lines[0], seq + '\n', '+\n', qline + '\n']
fmt = remove_multiline(fmt)
if len(lines[1]) != len(lines[3]):
msg = 'Sequence and quality line length do not match'
raise ValueError(msg)
elif 'fastq' in fmt:
lines = [lines[0], seq + '\n', lines[2], lines[3]]
if len(lines[1]) != len(lines[3]):
msg = 'Sequence and quality line length do not match'
raise ValueError(msg)
else:
raise RuntimeError('Unknown format for a SequenceItem')
if name:
lines[0] = lines[0][0] + name + '\n'
if 'fastq' in fmt:
if 'multiline' in fmt:
line_plus_index = _sitem_fastq_plus_line_index
else:
line_plus_index = 2
lines[line_plus_index] = '+\n'
name = seq_item.name if name is None else name
annotations = seq_item.annotations
if annotations is not None:
annotations = annotations.copy()
seq = SeqWrapper(kind=seqwrapper.kind,
object=SeqItem(name, lines, annotations),
file_format=fmt)
return seq
def write_seqs(seqs, fhand=None, file_format=None):
'It writes the given sequences'
if fhand is None:
fhand = NamedTemporaryFile(suffix='.' + file_format.replace('-', '_'))
file_format = remove_multiline(file_format)
seqs, seqs2 = tee(seqs)
try:
seq = seqs2.next()
except StopIteration:
# No sequences to write, so we're done
return fhand
del seqs2
seq_class = seq.kind
if seq_class == SEQITEM:
_write_seqitems(seqs, fhand, file_format)
elif seq_class == SEQRECORD:
seqs = (seq.object for seq in seqs)
_write_seqrecords(seqs, fhand, file_format)
else:
raise ValueError('Unknown class for seq: ' + seq_class)
return fhand
def write_seqs(seqs, fhand=None, file_format=None):
'It writes the given sequences'
if fhand is None:
fhand = NamedTemporaryFile(suffix='.' + file_format.replace('-', '_'))
file_format = remove_multiline(file_format)
seqs, seqs2 = tee(seqs)
try:
seq = seqs2.next()
except StopIteration:
# No sequences to write, so we're done
return fhand
del seqs2
seq_class = seq.kind
if seq_class == SEQITEM:
_write_seqitems(seqs, fhand, file_format)
elif seq_class == SEQRECORD:
seqs = (seq.object for seq in seqs)
_write_seqrecords(seqs, fhand, file_format)
else:
raise ValueError('Unknown class for seq: ' + seq_class)
return fhand
def seqio(in_fhands, out_fhand, out_format, copy_if_same_format=True):
'It converts sequence files between formats'
if out_format not in get_setting('SUPPORTED_OUTPUT_FORMATS'):
raise IncompatibleFormatError("This output format is not supported")
in_formats = [remove_multiline(guess_format(fhand)) for fhand in in_fhands]
if len(in_fhands) == 1 and in_formats[0] == out_format:
if copy_if_same_format:
copyfileobj(in_fhands[0], out_fhand)
else:
rel_symlink(in_fhands[0].name, out_fhand.name)
else:
seqs = _read_seqrecords(in_fhands)
try:
write_seqrecs(seqs, out_fhand, out_format)
except ValueError, error:
if error_quality_disagree(error):
raise MalformedFile(str(error))
if 'No suitable quality scores' in str(error):
msg = 'No qualities available to write output file'
raise IncompatibleFormatError(msg)
raise
def seqio(in_fhands, out_fhand, out_format, copy_if_same_format=True):
'It converts sequence files between formats'
if out_format not in get_setting('SUPPORTED_OUTPUT_FORMATS'):
raise IncompatibleFormatError("This output format is not supported")
in_formats = [remove_multiline(guess_format(fhand)) for fhand in in_fhands]
if len(in_fhands) == 1 and in_formats[0] == out_format:
if copy_if_same_format:
copyfileobj(in_fhands[0], out_fhand)
else:
rel_symlink(in_fhands[0].name, out_fhand.name)
else:
seqs = _read_seqrecords(in_fhands)
try:
write_seqrecs(seqs, out_fhand, out_format)
except ValueError, error:
if error_quality_disagree(error):
raise MalformedFile(str(error))
if 'No suitable quality scores' in str(error):
msg = 'No qualities available to write output file'
raise IncompatibleFormatError(msg)
raise
