def _get_some_qual_and_lengths(fhand, force_file_as_non_seek):
'It returns the quality characters and the lengths'
seqs_to_peek = get_setting('SEQS_TO_GUESS_FASTQ_VERSION')
chunk_size = get_setting('CHUNK_TO_GUESS_FASTQ_VERSION')
lengths = array('I')
seqs_analyzed = 0
if fhand_is_seekable(fhand) and not force_file_as_non_seek:
fmt_fhand = fhand
chunk = fmt_fhand.read(chunk_size)
fhand.seek(0)
else:
chunk = peek_chunk_from_file(fhand, chunk_size)
fmt_fhand = cStringIO.StringIO(chunk)
try:
for seq in FastqGeneralIterator(fmt_fhand):
qual = [ord(char) for char in seq[2]]
sanger_chars = [q for q in qual if q < 64]
if sanger_chars:
fhand.seek(0)
return None, True, chunk # no quals, no lengths, is_sanger
lengths.append(len(qual))
seqs_analyzed += 1
if seqs_analyzed > seqs_to_peek:
break
except ValueError:
msg = 'The file is Fastq, but the version is difficult to guess'
raise UndecidedFastqVersionError(msg)
finally:
fhand.seek(0)
return lengths, None, chunk # don't know if it's sanger
def _guess_format(fhand, force_file_as_non_seek):
'''It guesses the format of the sequence file.
This function is just for testing forcing the fhand as non-seekable.
It does ignore the solexa fastq version.
'''
chunk_size = 2048
chunk = peek_chunk_from_file(fhand, chunk_size)
if not chunk:
raise FileIsEmptyError('The file is empty')
lines = chunk.splitlines()
if chunk.startswith('>'):
if lines[1].startswith('>'):
raise UnknownFormatError('Malformed fasta')
else:
first_item = lines[1].strip().split()[0]
if first_item.isdigit():
return 'qual'
else:
return 'fasta'
elif chunk.startswith('@'):
return _guess_fastq_version(fhand, force_file_as_non_seek)
elif chunk.startswith('LOCUS'):
return 'genbank'
elif chunk.startswith('ID'):
return 'embl'
raise UnknownFormatError('Sequence file of unknown format.')
def _guess_format(fhand, force_file_as_non_seek):
"""It guesses the format of the sequence file.
This function is just for testing forcing the fhand as non-seekable.
It does ignore the solexa fastq version.
"""
chunk_size = 1024
chunk = peek_chunk_from_file(fhand, chunk_size)
if not chunk:
raise UnknownFormatError("The file is empty")
lines = chunk.splitlines()
if chunk.startswith(">"):
if lines[1].startswith(">"):
raise UnknownFormatError("Malformed fasta")
else:
first_item = lines[1].strip().split()[0]
if first_item.isdigit():
return "qual"
else:
return "fasta"
elif chunk.startswith("@"):
return _guess_fastq_version(fhand, force_file_as_non_seek)
elif chunk.startswith("LOCUS"):
return "genbank"
elif chunk.startswith("ID"):
return "embl"
raise UnknownFormatError("Sequence file of unknown format.")
def _get_some_qual_and_lengths(fhand, force_file_as_non_seek):
"It returns the quality characters and the lengths"
seqs_to_peek = SEQS_TO_GUESS_FASTQ_VERSION
chunk_size = CHUNK_TO_GUESS_FASTQ_VERSION
lengths = array("I")
seqs_analyzed = 0
if fhand_is_seekable(fhand) and not force_file_as_non_seek:
fmt_fhand = fhand
else:
chunk = peek_chunk_from_file(fhand, chunk_size)
fmt_fhand = cStringIO.StringIO(chunk)
try:
for seq in FastqGeneralIterator(fmt_fhand):
qual = [ord(char) for char in seq[2]]
sanger_chars = [q for q in qual if q < 64]
if sanger_chars:
fhand.seek(0)
return None, True # no quals, no lengths, is_sanger
lengths.append(len(qual))
seqs_analyzed += 1
if seqs_analyzed > seqs_to_peek:
break
except ValueError:
raise UnknownFormatError("Malformed fastq")
finally:
fhand.seek(0)
return lengths, None # quals, lengths, don't know if it's sanger
def _get_some_qual_and_lengths(fhand, force_file_as_non_seek):
'It returns the quality characters and the lengths'
seqs_to_peek = get_setting('SEQS_TO_GUESS_FASTQ_VERSION')
chunk_size = get_setting('CHUNK_TO_GUESS_FASTQ_VERSION')
lengths = array('I')
seqs_analyzed = 0
if fhand_is_seekable(fhand) and not force_file_as_non_seek:
fmt_fhand = fhand
chunk = fmt_fhand.read(chunk_size)
fhand.seek(0)
else:
chunk = peek_chunk_from_file(fhand, chunk_size)
fmt_fhand = cStringIO.StringIO(chunk)
try:
for seq in FastqGeneralIterator(fmt_fhand):
qual = [ord(char) for char in seq[2]]
sanger_chars = [q for q in qual if q < 64]
if sanger_chars:
fhand.seek(0)
return None, True, chunk # no quals, no lengths, is_sanger
lengths.append(len(qual))
seqs_analyzed += 1
if seqs_analyzed > seqs_to_peek:
break
except ValueError:
msg = 'The file is Fastq, but the version is difficult to guess'
raise UndecidedFastqVersionError(msg)
finally:
fhand.seek(0)
return lengths, None, chunk # don't know if it's sanger
def _guess_format(fhand, force_file_as_non_seek):
'''It guesses the format of the sequence file.
This function is just for testing forcing the fhand as non-seekable.
It does ignore the solexa fastq version.
'''
chunk_size = 2048
chunk = peek_chunk_from_file(fhand, chunk_size)
if not chunk:
raise FileIsEmptyError('The file is empty')
lines = chunk.splitlines()
if chunk.startswith('>'):
if lines[1].startswith('>'):
raise UnknownFormatError('Malformed fasta')
else:
first_item = lines[1].strip().split()[0]
if first_item.isdigit():
return 'qual'
else:
return 'fasta'
elif chunk.startswith('@'):
return _guess_fastq_version(fhand, force_file_as_non_seek)
elif chunk.startswith('LOCUS'):
return 'genbank'
elif chunk.startswith('ID'):
return 'embl'
raise UnknownFormatError('Sequence file of unknown format.')
