def split_paired_files(fp_obj, no_gzip=False,
barcodes=None, linker='',
min_length=4,
max_length=-1,
strip_before_barcode=0,
strip_after_barcode=0,
logger=None,
no_clipping=False,
**kwargs):
filename = fp_obj.input_file
filename2 = fp_obj.second_file
open_func, format1 = discover_file_format(filename)
open_func2, format2 = discover_file_format(filename2)
if not (format1 == 'FASTQ' and format2 == 'FASTQ'):
logger.error('Only FASTQ files are supported at this time')
return
f = open_func(filename, "rU")
f2 = open_func2(filename2, "rU")
records = FasterFastqIterator(f)
records2 = FasterFastqIterator(f2)
barcoded_file_pairs = {}
filenames = []
if no_gzip:
open_func = open
elif PATH_TO_GZIP is not None:
open_func = gzip_class_factory(PATH_TO_GZIP)
else:
open_func = GzipFile
output_filename = partial(fp_obj.output_filename, no_gzip=no_gzip)
output_filename2 = partial(fp_obj.output_filename2, no_gzip=no_gzip)
if barcodes is None:
barcodes = []
if barcodes is not None and len(barcodes) > 0:
processed_files = None
orphaned_read_files = None
for barcode in barcodes:
fname = output_filename(barcode)
fname2 = output_filename2(barcode)
filenames.extend((fname, fname2))
barcoded_file_pairs[barcode] = (open_func(fname, 'w'),
open_func(fname2, 'w'))
# and make a unmatched file
unmatched_filename = output_filename("unmatched")
unmatched_filename2 = output_filename2("unmatched")
unmatched_files = (open_func(unmatched_filename, 'w'),
open_func(unmatched_filename2, 'w'))
mismatched_filename = output_filename("mismatched")
mismatched_filename2 = output_filename2("mismatched")
mismatched_files = (open_func(mismatched_filename, 'w'),
open_func(mismatched_filename2, 'w'))
filenames.extend((unmatched_filename, unmatched_filename2,
mismatched_filename, mismatched_filename2))
else:
barcoded_file_pairs = None
unmatched_files = None
mismatched_files = None
orphaned_read_filename = output_filename("orphaned",
is_barcoded=False)
orphaned_read_filename2 = output_filename2("mismatched",
is_barcoded=False)
orphaned_read_files = (open_func(orphaned_read_filename, 'w'),
open_func(orphaned_read_filename2, 'w'))
processed_filename = output_filename("processed", is_barcoded=False)
processed_filename2 = output_filename2("processed", is_barcoded=False)
processed_files = (open_func(processed_filename, 'w'),
open_func(processed_filename2, 'w'))
filenames.extend((orphaned_read_filename, orphaned_read_filename2),
(processed_filename, processed_filename2))
writer_args = {'barcoded_file_pairs': barcoded_file_pairs,
'unmatched_files': unmatched_files,
'mismatched_files': mismatched_files,
'processed_files': processed_files,
'orphaned_read_files': orphaned_read_files,
'linker': linker,
'min_length': min_length
}
results = apply_plan_pe(records, records2, writer_args,
barcodes=barcodes, linker=linker,
min_length=min_length,
max_length=max_length,
strip_after_barcode=strip_after_barcode,
strip_before_barcode=strip_before_barcode,
no_clipping=no_clipping,
logger=logger)
linker_only = results['linker']
too_short = results['short']
record_count = results['all']
# close and exit #
f.close()
for f_, f2_ in barcoded_file_pairs.values():
f_.close()
f2_.close()
unmatched_files[0].close()
unmatched_files[1].close()
mismatched_files[0].close()
mismatched_files[1].close()
logger.info('Split %s, %s as %s', fp_obj.input_file, fp_obj.second_file,
', '.join(filenames))
logger.info('Processed %s records', record_count)
logger.info('%s linker only dimers', linker_only)
logger.info('%s sequences too short (1-3 bp)', too_short)
def split_file(fp_obj, no_gzip=False,
barcodes=[], linker='', min_length=4,
max_length=-1, logger=None,
strip_before_barcode=0,
strip_after_barcode=0,
no_clipping=False,
**kwargs):
if logger is None:
logger = get_logger()
filename = fp_obj.input_file
open_func, format_ = discover_file_format(filename)
if not format_ == 'FASTQ':
logger.error('Only FASTQ files are supported at this time')
return
f = open_func(filename, "rU")
records = FasterFastqIterator(f)
barcoded_files = {}
filenames = []
output_filename = partial(fp_obj.output_filename, no_gzip=no_gzip)
if no_gzip:
open_func = open
elif PATH_TO_GZIP is not None:
open_func = gzip_class_factory(PATH_TO_GZIP)
else:
open_func = GzipFile
if barcodes is None:
barcodes = []
if barcodes is not None and len(barcodes) > 0:
processed_file = None
for barcode in barcodes:
fname = output_filename(barcode)
filenames.append(fname)
barcoded_files[barcode] = open_func(fname, 'w')
# and make a unmatched file
unmatched_filename = output_filename("unmatched")
filenames.append(unmatched_filename)
unmatched_file = open_func(unmatched_filename, 'w')
else:
barcoded_files = None
unmatched_file = None
processed_filename = output_filename("processed", is_barcode=False)
filenames.append(processed_filename)
processed_file = open_func(processed_filename, 'w')
writer_args = {'barcoded_files': barcoded_files,
'unmatched_file': unmatched_file,
'processed_file': processed_file}
results = apply_plan(records, writer_args, barcodes=barcodes,
linker=linker,
min_length=min_length, max_length=max_length,
strip_after_barcode=strip_after_barcode,
strip_before_barcode=strip_before_barcode,
no_clipping=no_clipping,
logger=logger)
linker_only = results['linker']
too_short = results['short']
record_count = results['all']
# close and exit #
f.close()
if barcoded_files is not None:
logger.debug('closing barcoded files')
for f_ in barcoded_files.values():
f_.close()
if unmatched_file is not None:
logger.debug('closing unmatched file')
unmatched_file.close()
if processed_file is not None:
logger.debug('closing output file')
processed_file.close()
logger.info('Split %s as %s ', fp_obj.input_file, ', '.join(filenames))
logger.info('Processed %s records', record_count)
logger.info('%s linker only dimers', linker_only)
logger.info('%s sequences too short (1-3 bp)', too_short)
