def main(argv):
parser = make_argparser()
args = parser.parse_args(argv[1:])
if args.help:
parser.print_help()
return 0
logging.basicConfig(stream=args.log,
level=args.volume,
format='%(message)s')
tone_down_logger()
start_time = time.time()
# If the user requested, report back some data about the start of the run.
if args.phone_home:
call = phone.Call(__file__,
version.get_version(),
platform=args.platform,
test=args.test,
fail='warn')
call.send_data('start')
data = {
'stdin': args.infile is sys.stdin,
'processes': args.processes,
'queue_size': args.queue_size,
}
if data['stdin']:
data['input_size'] = None
else:
data['input_size'] = os.path.getsize(args.infile.name)
call.send_data('prelim', run_data=data)
else:
call = None
# Execute as much of the script as possible in a try/except to catch any exception that occurs
# and report it via ET.phone.
try:
# Process and validate arguments.
if args.queue_size is not None and args.queue_size <= 0:
fail('Error: --queue-size must be greater than zero.')
qual_start = QUAL_OFFSETS[args.qual_format]
qual_thres = chr(args.qual + qual_start)
if args.fastq_out is None:
# Output FASTA.
output_qual = None
else:
# Output FASTQ.
if qual_start + args.fastq_out > 126:
fail(
'Error: --fastq-out PHRED score ({}) is too large.'.format(
args.fastq_out))
output_qual = chr(qual_start + args.fastq_out)
if args.min_cons_reads > args.min_reads:
fail(
'Error: --min-reads must be greater than --min-cons-reads (or you\'ll have a lot of '
'consensus sequences with only N\'s!). If you want to exclude families with fewer than X '
'reads, give --min-reads X instead of --min-cons-reads X.')
if not any((args.dcs1, args.dcs2, args.sscs1, args.sscs2)):
fail('Error: must specify an output file!')
# A dict of output filehandles.
# Indexed so we can do filehandles['dcs'][mate].
filehandles = {
'dcs': (args.dcs1, args.dcs2),
'sscs': (args.sscs1, args.sscs2),
}
# Open a pool of worker processes.
stats = {'time': 0, 'reads': 0, 'runs': 0, 'duplexes': 0}
static_kwargs = {
'min_reads': args.min_reads,
'cons_thres': args.cons_thres,
'min_cons_reads': args.min_cons_reads,
'qual_thres': qual_thres,
'output_qual': output_qual,
}
pool = parallel_tools.SyncAsyncPool(
process_duplex,
processes=args.processes,
static_kwargs=static_kwargs,
queue_size=args.queue_size,
callback=process_result,
callback_args=[filehandles, stats],
)
try:
process_families(args.infile, pool, stats)
finally:
# If the root process encounters an exception and doesn't tell the workers to stop, it will
# hang forever.
pool.close()
pool.join()
# Close all open filehandles.
if args.infile is not sys.stdin:
args.infile.close()
for fh_group in filehandles.values():
for fh in fh_group:
if fh:
fh.close()
# Final stats on the run.
run_time = int(time.time() - start_time)
max_mem = get_max_mem()
logging.info(
'Processed {} reads and {} duplexes in {} seconds.'.format(
stats['total_reads'], stats['runs'], run_time))
if stats['reads'] > 0 and stats['runs'] > 0:
per_read = stats['time'] / stats['reads']
per_run = stats['time'] / stats['runs']
logging.info('{:0.3f}s per read, {:0.3f}s per run.'.format(
per_read, per_run))
logging.info('in {}s total time and {:0.2f}MB RAM.'.format(
run_time, max_mem))
except (Exception, KeyboardInterrupt) as exception:
if args.phone_home and call:
try:
exception_data = getattr(exception, 'child_context',
parallel_tools.get_exception_data())
logging.critical(
parallel_tools.format_traceback(exception_data))
exception_data = parallel_tools.scrub_tb_paths(
exception_data, script_path=__file__)
except Exception:
exception_data = {}
run_time = int(time.time() - start_time)
try:
run_data = get_run_data(stats, pool)
except (Exception, UnboundLocalError):
run_data = {}
try:
run_data['mem'] = get_max_mem()
except Exception:
pass
run_data['failed'] = True
if exception_data:
run_data['exception'] = exception_data
call.send_data('end', run_time=run_time, run_data=run_data)
raise exception
else:
raise
if args.phone_home and call:
run_data = get_run_data(stats, pool, max_mem)
call.send_data('end', run_time=run_time, run_data=run_data)
def main(argv):
parser = make_argparser()
args = parser.parse_args(argv[1:])
logging.basicConfig(stream=args.log_file, level=args.volume, format='%(message)s')
tone_down_logger()
start_time = time.time()
# If the user requested, report back some data about the start of the run.
if args.phone_home:
call = phone.Call(__file__, version.get_version(), platform=args.platform, test=args.test,
fail='warn')
call.send_data('start')
data = {
'stdin': args.infile is sys.stdin,
'aligner': args.aligner,
'processes': args.processes,
'queue_size': args.queue_size,
}
if data['stdin']:
data['input_size'] = None
else:
data['input_size'] = os.path.getsize(args.infile.name)
call.send_data('prelim', run_data=data)
# Execute as much of the script as possible in a try/except to catch any exception that occurs
# and report it via ET.phone.
try:
if args.queue_size is not None and args.queue_size <= 0:
fail('Error: --queue-size must be greater than zero.')
# If we're using mafft, check that we can execute it.
if args.aligner == 'mafft' and not distutils.spawn.find_executable('mafft'):
fail('Error: Could not find "mafft" command on $PATH.')
# Open a pool of worker processes.
stats = {'duplexes':0, 'time':0, 'pairs':0, 'runs':0, 'failures':0, 'aligned_pairs':0}
pool = parallel_tools.SyncAsyncPool(
process_duplex, processes=args.processes, static_kwargs={'aligner':args.aligner},
queue_size=args.queue_size, callback=process_result, callback_args=[stats]
)
try:
# The main loop.
align_families(args.infile, pool, stats, check_ids=args.check_ids)
finally:
# If an exception occurs in the parent without stopping the child processes, this will hang.
# Make sure to kill the children in all cases.
pool.close()
pool.join()
# Close input filehandle if it's open.
if args.infile is not sys.stdin:
args.infile.close()
# Final stats on the run.
run_time = int(time.time() - start_time)
max_mem = get_max_mem()
logging.error(
'Processed {pairs} read pairs in {duplexes} duplexes, with {failures} alignment failures.'
.format(**stats)
)
if stats['aligned_pairs'] > 0 and stats['runs'] > 0:
per_pair = stats['time'] / stats['aligned_pairs']
per_run = stats['time'] / stats['runs']
logging.error(f'{per_pair:0.3f}s per pair, {per_run:0.3f}s per run.')
logging.error(f'in {run_time}s total time and {max_mem:0.2f}MB RAM.')
except (Exception, KeyboardInterrupt) as exception:
if args.phone_home and call:
try:
exception_data = getattr(exception, 'child_context', parallel_tools.get_exception_data())
logging.critical(parallel_tools.format_traceback(exception_data))
exception_data = parallel_tools.scrub_tb_paths(exception_data, script_path=__file__)
except Exception:
exception_data = {}
run_time = int(time.time() - start_time)
try:
run_data = get_run_data(stats, pool, args.aligner)
except (Exception, UnboundLocalError):
run_data = {}
try:
run_data['mem'] = get_max_mem()
except Exception:
pass
run_data['failed'] = True
if exception_data:
run_data['exception'] = exception_data
call.send_data('end', run_time=run_time, run_data=run_data)
raise exception
else:
raise
if args.phone_home and call:
run_data = get_run_data(stats, pool, args.aligner, max_mem)
call.send_data('end', run_time=run_time, run_data=run_data)
def main(argv):
parser = make_argparser()
args = parser.parse_args(argv[1:])
if args.help:
parser.print_help()
return 0
logging.basicConfig(stream=args.log, level=args.volume, format='%(message)s')
tone_down_logger()
start_time = time.time()
# If the user requested, report back some data about the start of the run.
if args.phone_home:
call = phone.Call(__file__, version.get_version(), platform=args.platform, test=args.test,
fail='warn')
call.send_data('start')
data = {
'stdin': args.infile is sys.stdin,
'processes': args.processes,
'queue_size': args.queue_size,
}
if data['stdin']:
data['input_size'] = None
else:
data['input_size'] = os.path.getsize(args.infile.name)
call.send_data('prelim', run_data=data)
else:
call = None
# Execute as much of the script as possible in a try/except to catch any exception that occurs
# and report it via ET.phone.
try:
# Process and validate arguments.
if args.queue_size is not None and args.queue_size <= 0:
fail('Error: --queue-size must be greater than zero.')
qual_start = QUAL_OFFSETS[args.qual_format]
qual_thres = chr(args.qual + qual_start)
if args.fastq_out is None:
# Output FASTA.
output_qual = None
else:
# Output FASTQ.
if qual_start+args.fastq_out > 126:
fail('Error: --fastq-out PHRED score ({}) is too large.'.format(args.fastq_out))
output_qual = chr(qual_start+args.fastq_out)
if args.min_cons_reads > args.min_reads:
fail('Error: --min-reads must be greater than --min-cons-reads (or you\'ll have a lot of '
'consensus sequences with only N\'s!). If you want to exclude families with fewer than X '
'reads, give --min-reads X instead of --min-cons-reads X.')
if not any((args.dcs1, args.dcs2, args.sscs1, args.sscs2)):
fail('Error: must specify an output file!')
# A dict of output filehandles.
# Indexed so we can do filehandles['dcs'][mate].
filehandles = {
'dcs': (args.dcs1, args.dcs2),
'sscs': (args.sscs1, args.sscs2),
}
# Open a pool of worker processes.
stats = {'time':0, 'reads':0, 'runs':0, 'duplexes':0}
static_kwargs = {
'min_reads': args.min_reads,
'cons_thres': args.cons_thres,
'min_cons_reads': args.min_cons_reads,
'qual_thres': qual_thres,
'output_qual': output_qual,
}
pool = parallel_tools.SyncAsyncPool(process_duplex,
processes=args.processes,
static_kwargs=static_kwargs,
queue_size=args.queue_size,
callback=process_result,
callback_args=[filehandles, stats],
)
try:
total_reads = 0
duplex = collections.OrderedDict()
family = []
barcode = None
order = None
# Note: mate is a 0-indexed integer ("mate 1" from the input file is mate 0 here).
mate = None
for line in args.infile:
# Allow comments (e.g. for test input files).
if line.startswith('#'):
continue
fields = line.rstrip('\r\n').split('\t')
if len(fields) != 6:
continue
this_barcode, this_order, this_mate, name, seq, qual = fields
this_mate = int(this_mate)-1
# If the barcode, order, and mate are the same, we're just continuing the add reads to the
# current family. Otherwise, store the current family, start a new one, and process the
# duplex if we're at the end of one.
new_barcode = this_barcode != barcode
new_order = this_order != order
new_mate = this_mate != mate
if new_barcode or new_order or new_mate:
if order is not None and mate is not None:
duplex[(order, mate)] = family
# If the barcode changed, process the last duplex and start a new one.
if new_barcode and barcode is not None:
assert len(duplex) <= 4, duplex.keys()
pool.compute(duplex, barcode)
stats['duplexes'] += 1
duplex = collections.OrderedDict()
barcode = this_barcode
order = this_order
mate = this_mate
family = []
read = {'name': name, 'seq':seq, 'qual':qual}
family.append(read)
total_reads += 1
# Process the last family.
if order is not None and mate is not None:
duplex[(order, mate)] = family
assert len(duplex) <= 4, duplex.keys()
pool.compute(duplex, barcode)
stats['duplexes'] += 1
# Retrieve the remaining results.
logging.info('Flushing remaining results from worker processes..')
pool.flush()
finally:
# If the root process encounters an exception and doesn't tell the workers to stop, it will
# hang forever.
pool.close()
pool.join()
# Close all open filehandles.
if args.infile is not sys.stdin:
args.infile.close()
for fh_group in filehandles.values():
for fh in fh_group:
if fh:
fh.close()
# Final stats on the run.
run_time = int(time.time() - start_time)
max_mem = get_max_mem()
logging.info('Processed {} reads and {} duplexes in {} seconds.'
.format(total_reads, stats['runs'], run_time))
if stats['reads'] > 0 and stats['runs'] > 0:
per_read = stats['time'] / stats['reads']
per_run = stats['time'] / stats['runs']
logging.info('{:0.3f}s per read, {:0.3f}s per run.'.format(per_read, per_run))
logging.info('in {}s total time and {:0.2f}MB RAM.'.format(run_time, max_mem))
except (Exception, KeyboardInterrupt) as exception:
if args.phone_home and call:
try:
exception_data = getattr(exception, 'child_context', parallel_tools.get_exception_data())
logging.critical(parallel_tools.format_traceback(exception_data))
exception_data = parallel_tools.scrub_tb_paths(exception_data, script_path=__file__)
except Exception:
exception_data = {}
run_time = int(time.time() - start_time)
try:
run_data = get_run_data(stats, pool)
except (Exception, UnboundLocalError):
run_data = {}
try:
run_data['mem'] = get_max_mem()
except Exception:
pass
run_data['failed'] = True
if exception_data:
run_data['exception'] = exception_data
call.send_data('end', run_time=run_time, run_data=run_data)
raise exception
else:
raise
if args.phone_home and call:
run_data = get_run_data(stats, pool, max_mem)
call.send_data('end', run_time=run_time, run_data=run_data)
def main(argv):
parser = make_argparser()
args = parser.parse_args(argv[1:])
logging.basicConfig(stream=args.log_file, level=args.volume, format='%(message)s')
tone_down_logger()
start_time = time.time()
# If the user requested, report back some data about the start of the run.
if args.phone_home:
call = phone.Call(__file__, version.get_version(), platform=args.platform, test=args.test,
fail='warn')
call.send_data('start')
data = {
'stdin': args.infile is sys.stdin,
'aligner': args.aligner,
'processes': args.processes,
'queue_size': args.queue_size,
}
if data['stdin']:
data['input_size'] = None
else:
data['input_size'] = os.path.getsize(args.infile.name)
call.send_data('prelim', run_data=data)
# Execute as much of the script as possible in a try/except to catch any exception that occurs
# and report it via ET.phone.
try:
if args.queue_size is not None and args.queue_size <= 0:
fail('Error: --queue-size must be greater than zero.')
# If we're using mafft, check that we can execute it.
if args.aligner == 'mafft' and not distutils.spawn.find_executable('mafft'):
fail('Error: Could not find "mafft" command on $PATH.')
# Open a pool of worker processes.
stats = {'duplexes':0, 'time':0, 'pairs':0, 'runs':0, 'failures':0, 'aligned_pairs':0}
pool = parallel_tools.SyncAsyncPool(process_duplex,
processes=args.processes,
static_kwargs={'aligner':args.aligner},
queue_size=args.queue_size,
callback=process_result,
callback_args=[stats],
)
"""Now the main loop.
This processes whole duplexes (pairs of strands) at a time for a future option to align the
whole duplex at a time.
duplex data structure:
duplex = {
'ab': [
{'name1': 'read_name1a',
'seq1': 'GATT-ACA',
'qual1': 'sc!0 /J*',
'name2': 'read_name1b',
'seq2': 'ACTGACTA',
'qual2': '34I&SDF)'
},
{'name1': 'read_name2a',
...
},
...
],
'ba': [
...
]
}
e.g.:
seq = duplex[order][pair_num]['seq1']"""
try:
duplex = collections.OrderedDict()
family = []
barcode = None
order = None
for line in args.infile:
fields = line.rstrip('\r\n').split('\t')
if len(fields) != 8:
continue
(this_barcode, this_order, name1, seq1, qual1, name2, seq2, qual2) = fields
# If the barcode or order has changed, we're in a new family.
# Process the reads we've previously gathered as one family and start a new family.
if this_barcode != barcode or this_order != order:
duplex[order] = family
# If the barcode is different, we're at the end of the whole duplex. Process the it and start
# a new one. If the barcode is the same, we're in the same duplex, but we've switched strands.
if this_barcode != barcode:
# logging.debug('processing {}: {} orders ({})'.format(barcode, len(duplex),
# '/'.join([str(len(duplex[o])) for o in duplex])))
if barcode is not None:
pool.compute(duplex, barcode)
stats['duplexes'] += 1
duplex = collections.OrderedDict()
barcode = this_barcode
order = this_order
family = []
pair = {'name1': name1, 'seq1':seq1, 'qual1':qual1, 'name2':name2, 'seq2':seq2, 'qual2':qual2}
family.append(pair)
stats['pairs'] += 1
# Process the last family.
duplex[order] = family
# logging.debug('processing {}: {} orders ({}) [last]'.format(barcode, len(duplex),
# '/'.join([str(len(duplex[o])) for o in duplex])))
pool.compute(duplex, barcode)
stats['duplexes'] += 1
# Retrieve the remaining results.
logging.info('Flushing remaining results from worker processes..')
pool.flush()
finally:
# If an exception occurs in the parent without stopping the child processes, this will hang.
# Make sure to kill the children in all cases.
pool.close()
pool.join()
# Close input filehandle if it's open.
if args.infile is not sys.stdin:
args.infile.close()
# Final stats on the run.
run_time = int(time.time() - start_time)
max_mem = get_max_mem()
logging.error('Processed {pairs} read pairs in {duplexes} duplexes, with {failures} alignment '
'failures.'.format(**stats))
if stats['aligned_pairs'] > 0 and stats['runs'] > 0:
per_pair = stats['time'] / stats['aligned_pairs']
per_run = stats['time'] / stats['runs']
logging.error('{:0.3f}s per pair, {:0.3f}s per run.'.format(per_pair, per_run))
logging.error('in {}s total time and {:0.2f}MB RAM.'.format(run_time, max_mem))
except (Exception, KeyboardInterrupt) as exception:
if args.phone_home and call:
try:
exception_data = getattr(exception, 'child_context', parallel_tools.get_exception_data())
logging.critical(parallel_tools.format_traceback(exception_data))
exception_data = parallel_tools.scrub_tb_paths(exception_data, script_path=__file__)
except Exception:
exception_data = {}
run_time = int(time.time() - start_time)
try:
run_data = get_run_data(stats, pool, args.aligner)
except (Exception, UnboundLocalError):
run_data = {}
try:
run_data['mem'] = get_max_mem()
except Exception:
pass
run_data['failed'] = True
if exception_data:
run_data['exception'] = exception_data
call.send_data('end', run_time=run_time, run_data=run_data)
raise exception
else:
raise
if args.phone_home and call:
run_data = get_run_data(stats, pool, args.aligner, max_mem)
call.send_data('end', run_time=run_time, run_data=run_data)
def main(argv):
# Allow using -v for --version if it's the only argument, and --verbose if there are more.
if len(argv) == 2 and argv[1] == '-v':
print(version.get_version())
return
parser = make_argparser()
args = parser.parse_args(argv[1:])
logging.basicConfig(stream=args.log, level=args.volume, format='%(message)s')
tone_down_logger()
start_time = time.time()
# If the user requested, report back some data about the start of the run.
if args.phone_home:
call = phone.Call(__file__, version.get_version(), platform=args.platform, test=args.test,
fail='warn')
call.send_data('start')
call.send_data('prelim', run_data=gather_prelim_data(args.families, args.reads, args.sam))
# Execute as much of the script as possible in a try/except to catch any exception that occurs
# and report it via ET.phone.
try:
logging.info('Reading the fasta/q to map read names to barcodes..')
names_to_barcodes = map_names_to_barcodes(args.reads, args.limit)
logging.info('Reading the SAM to build the graph of barcode relationships..')
graph, reversed_barcodes, num_good_alignments = read_alignments(args.sam, names_to_barcodes,
args.pos, args.mapq,
args.dist, args.limit)
logging.info('Reading the families.tsv to get the counts of each family..')
family_counts, read_pairs = get_family_counts(args.families, limit=args.limit,
check_ids=args.check_ids)
if args.structures or args.visualize != 0:
logging.info('Counting the unique barcode networks..')
structures = count_structures(graph, family_counts)
if args.structures:
print_structures(structures, args.struct_human)
if args.visualize != 0:
logging.info('Generating a visualization of barcode networks..')
visualize([s['graph'] for s in structures], args.visualize, args.viz_format)
logging.info('Building the correction table from the graph..')
corrections = make_correction_table(graph, family_counts, args.choose_by)
logging.info('Reading the families.tsv again to print corrected output..')
with open_as_text_or_gzip(args.families.name) as families:
print_corrected_output(families, corrections, reversed_barcodes, args.prepend, args.limit,
args.output)
run_time = int(time.time() - start_time)
max_mem = resource.getrusage(resource.RUSAGE_SELF).ru_maxrss/1024
logging.info('Max memory usage: {:0.2f}MB'.format(max_mem))
logging.info('Wall clock time: {} seconds'.format(run_time))
except (Exception, KeyboardInterrupt) as exception:
if args.phone_home and call:
try:
exception_data = getattr(exception, 'child_context', parallel_tools.get_exception_data())
logging.critical(parallel_tools.format_traceback(exception_data))
exception_data = parallel_tools.scrub_tb_paths(exception_data, script_path=__file__)
except Exception:
exception_data = {}
raise
run_time = int(time.time() - start_time)
try:
run_data = {'barcodes':len(names_to_barcodes), 'good_alignments':num_good_alignments,
'read_pairs':read_pairs, 'max_mem':int(max_mem)}
except Exception:
run_data = {}
run_data['failed'] = True
if exception_data:
run_data['exception'] = exception_data
call.send_data('end', run_time=run_time, run_data=run_data)
raise exception
else:
raise
if args.phone_home:
run_data = {'barcodes':len(names_to_barcodes), 'good_alignments':num_good_alignments,
'read_pairs':read_pairs, 'max_mem':int(max_mem)}
call.send_data('end', run_time=run_time, run_data=run_data)
def main(argv):
parser = make_argparser()
args = parser.parse_args(argv[1:])
if args.help:
parser.print_help()
return 0
logging.basicConfig(stream=args.log,
level=args.volume,
format='%(message)s')
tone_down_logger()
start_time = time.time()
# If the user requested, report back some data about the start of the run.
if args.phone_home:
call = phone.Call(__file__,
version.get_version(),
platform=args.platform,
test=args.test,
fail='warn')
call.send_data('start')
data = {
'stdin': args.infile is sys.stdin,
'processes': args.processes,
'queue_size': args.queue_size,
}
if data['stdin']:
data['input_size'] = None
else:
data['input_size'] = os.path.getsize(args.infile.name)
call.send_data('prelim', run_data=data)
else:
call = None
# Execute as much of the script as possible in a try/except to catch any exception that occurs
# and report it via ET.phone.
try:
# Process and validate arguments.
if args.queue_size is not None and args.queue_size <= 0:
fail('Error: --queue-size must be greater than zero.')
qual_start = QUAL_OFFSETS[args.qual_format]
qual_thres = chr(args.qual + qual_start)
if args.fastq_out is None:
# Output FASTA.
output_qual = None
else:
# Output FASTQ.
if qual_start + args.fastq_out > 126:
fail(
'Error: --fastq-out PHRED score ({}) is too large.'.format(
args.fastq_out))
output_qual = chr(qual_start + args.fastq_out)
if args.min_cons_reads > args.min_reads:
fail(
'Error: --min-reads must be greater than --min-cons-reads (or you\'ll have a lot of '
'consensus sequences with only N\'s!). If you want to exclude families with fewer than X '
'reads, give --min-reads X instead of --min-cons-reads X.')
if not any((args.dcs1, args.dcs2, args.sscs1, args.sscs2)):
fail('Error: must specify an output file!')
# A dict of output filehandles.
# Indexed so we can do filehandles['dcs'][mate].
filehandles = {
'dcs': (args.dcs1, args.dcs2),
'sscs': (args.sscs1, args.sscs2),
}
# Open a pool of worker processes.
stats = {'time': 0, 'reads': 0, 'runs': 0, 'duplexes': 0}
static_kwargs = {
'min_reads': args.min_reads,
'cons_thres': args.cons_thres,
'min_cons_reads': args.min_cons_reads,
'qual_thres': qual_thres,
'output_qual': output_qual,
}
pool = parallel_tools.SyncAsyncPool(
process_duplex,
processes=args.processes,
static_kwargs=static_kwargs,
queue_size=args.queue_size,
callback=process_result,
callback_args=[filehandles, stats],
)
try:
total_reads = 0
duplex = collections.OrderedDict()
family = []
barcode = None
order = None
# Note: mate is a 0-indexed integer ("mate 1" from the input file is mate 0 here).
mate = None
for line in args.infile:
# Allow comments (e.g. for test input files).
if line.startswith('#'):
continue
fields = line.rstrip('\r\n').split('\t')
if len(fields) != 6:
continue
this_barcode, this_order, this_mate, name, seq, qual = fields
this_mate = int(this_mate) - 1
# If the barcode, order, and mate are the same, we're just continuing the add reads to the
# current family. Otherwise, store the current family, start a new one, and process the
# duplex if we're at the end of one.
new_barcode = this_barcode != barcode
new_order = this_order != order
new_mate = this_mate != mate
if new_barcode or new_order or new_mate:
if order is not None and mate is not None:
duplex[(order, mate)] = family
# If the barcode changed, process the last duplex and start a new one.
if new_barcode and barcode is not None:
assert len(duplex) <= 4, duplex.keys()
pool.compute(duplex, barcode)
stats['duplexes'] += 1
duplex = collections.OrderedDict()
barcode = this_barcode
order = this_order
mate = this_mate
family = []
read = {'name': name, 'seq': seq, 'qual': qual}
family.append(read)
total_reads += 1
# Process the last family.
if order is not None and mate is not None:
duplex[(order, mate)] = family
assert len(duplex) <= 4, duplex.keys()
pool.compute(duplex, barcode)
stats['duplexes'] += 1
# Retrieve the remaining results.
logging.info('Flushing remaining results from worker processes..')
pool.flush()
finally:
# If the root process encounters an exception and doesn't tell the workers to stop, it will
# hang forever.
pool.close()
pool.join()
# Close all open filehandles.
if args.infile is not sys.stdin:
args.infile.close()
for fh_group in filehandles.values():
for fh in fh_group:
if fh:
fh.close()
# Final stats on the run.
run_time = int(time.time() - start_time)
max_mem = get_max_mem()
logging.info(
'Processed {} reads and {} duplexes in {} seconds.'.format(
total_reads, stats['runs'], run_time))
if stats['reads'] > 0 and stats['runs'] > 0:
per_read = stats['time'] / stats['reads']
per_run = stats['time'] / stats['runs']
logging.info('{:0.3f}s per read, {:0.3f}s per run.'.format(
per_read, per_run))
logging.info('in {}s total time and {:0.2f}MB RAM.'.format(
run_time, max_mem))
except (Exception, KeyboardInterrupt) as exception:
if args.phone_home and call:
try:
exception_data = getattr(exception, 'child_context',
parallel_tools.get_exception_data())
logging.critical(
parallel_tools.format_traceback(exception_data))
exception_data = parallel_tools.scrub_tb_paths(
exception_data, script_path=__file__)
except Exception:
exception_data = {}
run_time = int(time.time() - start_time)
try:
run_data = get_run_data(stats, pool)
except (Exception, UnboundLocalError):
run_data = {}
try:
run_data['mem'] = get_max_mem()
except Exception:
pass
run_data['failed'] = True
if exception_data:
run_data['exception'] = exception_data
call.send_data('end', run_time=run_time, run_data=run_data)
raise exception
else:
raise
if args.phone_home and call:
run_data = get_run_data(stats, pool, max_mem)
call.send_data('end', run_time=run_time, run_data=run_data)
def main(argv):
parser = make_argparser()
args = parser.parse_args(argv[1:])
logging.basicConfig(stream=args.log_file, level=args.volume, format='%(message)s')
tone_down_logger()
start_time = time.time()
# If the user requested, report back some data about the start of the run.
if args.phone_home:
call = phone.Call(__file__, version.get_version(), platform=args.platform, test=args.test,
fail='warn')
call.send_data('start')
data = {
'stdin': args.infile is sys.stdin,
'aligner': args.aligner,
'processes': args.processes,
'queue_size': args.queue_size,
}
if data['stdin']:
data['input_size'] = None
else:
data['input_size'] = os.path.getsize(args.infile.name)
call.send_data('prelim', run_data=data)
# Execute as much of the script as possible in a try/except to catch any exception that occurs
# and report it via ET.phone.
try:
if args.queue_size is not None and args.queue_size <= 0:
fail('Error: --queue-size must be greater than zero.')
# If we're using mafft, check that we can execute it.
if args.aligner == 'mafft' and not distutils.spawn.find_executable('mafft'):
fail('Error: Could not find "mafft" command on $PATH.')
# Open a pool of worker processes.
stats = {'duplexes':0, 'time':0, 'pairs':0, 'runs':0, 'failures':0, 'aligned_pairs':0}
pool = parallel_tools.SyncAsyncPool(process_duplex,
processes=args.processes,
static_kwargs={'aligner':args.aligner},
queue_size=args.queue_size,
callback=process_result,
callback_args=[stats],
)
"""Now the main loop.
This processes whole duplexes (pairs of strands) at a time for a future option to align the
whole duplex at a time.
duplex data structure:
duplex = {
'ab': [
{'name1': 'read_name1a',
'seq1': 'GATT-ACA',
'qual1': 'sc!0 /J*',
'name2': 'read_name1b',
'seq2': 'ACTGACTA',
'qual2': '34I&SDF)'
},
{'name1': 'read_name2a',
...
},
...
],
'ba': [
...
]
}
e.g.:
seq = duplex[order][pair_num]['seq1']"""
try:
duplex = collections.OrderedDict()
family = []
barcode = None
order = None
for line in args.infile:
fields = line.rstrip('\r\n').split('\t')
if len(fields) != 8:
continue
(this_barcode, this_order, name1, seq1, qual1, name2, seq2, qual2) = fields
if args.check_ids and not read_ids_match(name1, name2):
raise ValueError('Read names "{}" and "{}" do not match.'.format(name1, name2))
# If the barcode or order has changed, we're in a new family.
# Process the reads we've previously gathered as one family and start a new family.
if this_barcode != barcode or this_order != order:
duplex[order] = family
# If the barcode is different, we're at the end of the whole duplex. Process the it and start
# a new one. If the barcode is the same, we're in the same duplex, but we've switched strands.
if this_barcode != barcode:
# logging.debug('processing {}: {} orders ({})'.format(barcode, len(duplex),
# '/'.join([str(len(duplex[o])) for o in duplex])))
if barcode is not None:
pool.compute(duplex, barcode)
stats['duplexes'] += 1
duplex = collections.OrderedDict()
barcode = this_barcode
order = this_order
family = []
pair = {'name1': name1, 'seq1':seq1, 'qual1':qual1, 'name2':name2, 'seq2':seq2, 'qual2':qual2}
family.append(pair)
stats['pairs'] += 1
# Process the last family.
duplex[order] = family
# logging.debug('processing {}: {} orders ({}) [last]'.format(barcode, len(duplex),
# '/'.join([str(len(duplex[o])) for o in duplex])))
pool.compute(duplex, barcode)
stats['duplexes'] += 1
# Retrieve the remaining results.
logging.info('Flushing remaining results from worker processes..')
pool.flush()
finally:
# If an exception occurs in the parent without stopping the child processes, this will hang.
# Make sure to kill the children in all cases.
pool.close()
pool.join()
# Close input filehandle if it's open.
if args.infile is not sys.stdin:
args.infile.close()
# Final stats on the run.
run_time = int(time.time() - start_time)
max_mem = get_max_mem()
logging.error('Processed {pairs} read pairs in {duplexes} duplexes, with {failures} alignment '
'failures.'.format(**stats))
if stats['aligned_pairs'] > 0 and stats['runs'] > 0:
per_pair = stats['time'] / stats['aligned_pairs']
per_run = stats['time'] / stats['runs']
logging.error('{:0.3f}s per pair, {:0.3f}s per run.'.format(per_pair, per_run))
logging.error('in {}s total time and {:0.2f}MB RAM.'.format(run_time, max_mem))
except (Exception, KeyboardInterrupt) as exception:
if args.phone_home and call:
try:
exception_data = getattr(exception, 'child_context', parallel_tools.get_exception_data())
logging.critical(parallel_tools.format_traceback(exception_data))
exception_data = parallel_tools.scrub_tb_paths(exception_data, script_path=__file__)
except Exception:
exception_data = {}
run_time = int(time.time() - start_time)
try:
run_data = get_run_data(stats, pool, args.aligner)
except (Exception, UnboundLocalError):
run_data = {}
try:
run_data['mem'] = get_max_mem()
except Exception:
pass
run_data['failed'] = True
if exception_data:
run_data['exception'] = exception_data
call.send_data('end', run_time=run_time, run_data=run_data)
raise exception
else:
raise
if args.phone_home and call:
run_data = get_run_data(stats, pool, args.aligner, max_mem)
call.send_data('end', run_time=run_time, run_data=run_data)
