def testAll(self):
"""Test FastaRandomReader.keys() and __getitem__."""
write_files_to_fofn([self.fa1, self.fa2], self.fofn)
reader = MetaSubreadFastaReader([self.fa1, self.fa2])
subread_1 = "m130812_185809_42141_c100533960310000001823079711101380_s1_p0/59/0_5071"
subread_2 = "m130812_random_random_s1_p0/440/13280_16126"
zmw_3 = "m130812_185809_42141_c100533960310000001823079711101380_s1_p0/70"
zmw_4 = "m130812_random_random_s1_p0/249"
r1 = reader[subread_1][0]
self.assertEqual(r1.name, subread_1)
self.assertEqual(hashlib.md5(r1.sequence).hexdigest(), "8128261dd851ae285d029618739559e9")
r2 = reader[subread_2][0]
self.assertEqual(r2.name, subread_2)
self.assertEqual(hashlib.md5(r2.sequence).hexdigest(), "451e5798a7f21cce80da27a03a8cb2c7")
r3, r4 = reader[zmw_3]
self.assertEqual(r3.name, "m130812_185809_42141_c100533960310000001823079711101380_s1_p0/70/0_5538")
self.assertEqual(r4.name, "m130812_185809_42141_c100533960310000001823079711101380_s1_p0/70/5587_5982")
self.assertEqual(hashlib.md5(r3.sequence).hexdigest(), "4db2f6e35c83dd279a8f71a51ac50445")
self.assertEqual(hashlib.md5(r4.sequence).hexdigest(), "1c1d080e9362a73ea2074f9a62fbd45e")
r5 = reader[zmw_4][0]
self.assertEqual(r5.name, "m130812_random_random_s1_p0/249/0_1339")
self.assertEqual(hashlib.md5(r5.sequence).hexdigest(), "b20d3723a136aedc2f96f6f498ad3da0")
def convert_fofn_to_fasta(fofn_filename,
out_filename,
fasta_out_dir,
force_overwrite=False):
"""
For each .bax.h5 file, create .bax.h5.fasta file and save paths to
out_filename, which should usually be 'input.fasta.fofn'
"""
logging.info("Converting fofn {fofn} to fasta.".format(fofn=fofn_filename))
in_fns = get_files_from_fofn(fofn_filename)
out_fns = []
mkdir(fasta_out_dir)
for in_fn in in_fns:
logging.debug("converting h5 file: {f}.".format(f=in_fn))
if not (in_fn.endswith('.bax.h5') or in_fn.endswith('.bas.h5')):
raise ValueError("fofn file {fofn} ".format(fofn=fofn_filename) +
"should only contain bax/bas.h5 files.")
# e.g. m111xxxx.1.bax.h5 ==>
# tmp_out_file = m11xxxx.1.bax.h5.fasta.tmp
# out_file = m11xxxx.1.bax.h5.fasta
in_basename = op.basename(in_fn)
tmp_out_file = op.join(fasta_out_dir, in_basename + '.fasta.tmp')
out_file = op.join(fasta_out_dir, in_basename + '.fasta')
if op.exists(out_file) and not force_overwrite:
logging.debug(
"File {0} already exists. skipping.".format(out_file))
else:
cmd = "pls2fasta {0} {1} ".format(in_fn, tmp_out_file) + \
"-minSubreadLength 300 -minReadScore 750 -trimByRegion"
logging.debug("CMD: {cmd}".format(cmd=cmd))
_out, _code, _msg = backticks(cmd)
if _code != 0:
raise RuntimeError("CMD failed: {cmd}\n".format(cmd=cmd) +
_msg)
trim_subread_flanks(tmp_out_file, out_file)
out_fns.append(out_file)
if op.exists(tmp_out_file):
os.remove(tmp_out_file)
write_files_to_fofn(out_fns, out_filename)
def convert_fofn_to_fasta(fofn_filename, out_filename, fasta_out_dir,
force_overwrite=False):
"""
For each .bax.h5 file, create .bax.h5.fasta file and save paths to
out_filename, which should usually be 'input.fasta.fofn'
"""
logging.info("Converting fofn {fofn} to fasta.".format(fofn=fofn_filename))
in_fns = get_files_from_fofn(fofn_filename)
out_fns = []
mkdir(fasta_out_dir)
for in_fn in in_fns:
logging.debug("converting h5 file: {f}.".format(f=in_fn))
if not (in_fn.endswith('.bax.h5') or in_fn.endswith('.bas.h5')):
raise ValueError("fofn file {fofn} ".format(fofn=fofn_filename) +
"should only contain bax/bas.h5 files.")
# e.g. m111xxxx.1.bax.h5 ==>
# tmp_out_file = m11xxxx.1.bax.h5.fasta.tmp
# out_file = m11xxxx.1.bax.h5.fasta
in_basename = op.basename(in_fn)
tmp_out_file = op.join(fasta_out_dir, in_basename + '.fasta.tmp')
out_file = op.join(fasta_out_dir, in_basename + '.fasta')
if op.exists(out_file) and not force_overwrite:
logging.debug("File {0} already exists. skipping.".format(out_file))
else:
cmd = "pls2fasta {0} {1} ".format(in_fn, tmp_out_file) + \
"-minSubreadLength 300 -minReadScore 750 -trimByRegion"
logging.debug("CMD: {cmd}".format(cmd=cmd))
_out, _code, _msg = backticks(cmd)
if _code != 0:
raise RuntimeError("CMD failed: {cmd}\n".format(cmd=cmd) + _msg)
trim_subread_flanks(tmp_out_file, out_file)
out_fns.append(out_file)
if op.exists(tmp_out_file):
os.remove(tmp_out_file)
write_files_to_fofn(out_fns, out_filename)
def convert_fofn_to_fasta(fofn_filename, out_filename, fasta_out_dir,
force_overwrite=False, cpus=1):
"""
For each .bax.h5 file, create .bax.h5.fasta file and save paths to
out_filename, which should usually be 'input.fasta.fofn'
"""
logging.info("Converting fofn {fofn} to fasta.".format(fofn=fofn_filename))
in_fns = get_files_from_fofn(fofn_filename)
#out_fns = []
mkdir(fasta_out_dir)
# multiprocessing worker stuff
manager = Manager()
in_queue = manager.Queue(len(in_fns))
in_queue_count = 0
outfile_track = {} # expected out file --> (cmd, tmp)
pool = []
out_fns = []
for in_fn in in_fns:
#print >> sys.stderr, "DEBUG: converting h5 file:", in_fn
logging.debug("converting h5 file: {f}.".format(f=in_fn))
if not (in_fn.endswith('.bax.h5') or in_fn.endswith('.bas.h5')):
raise ValueError("fofn file {fofn} ".format(fofn=fofn_filename) +
"should only contain bax/bas.h5 files.")
# e.g. m111xxxx.1.bax.h5 ==>
# tmp_out_file = m11xxxx.1.bax.h5.fasta.tmp
# out_file = m11xxxx.1.bax.h5.fasta
in_basename = op.basename(in_fn)
tmp_out_file = op.join(fasta_out_dir, in_basename + '.fasta.tmp')
out_file = op.join(fasta_out_dir, in_basename + '.fasta')
if op.exists(out_file) and not force_overwrite:
logging.debug("File {0} already exists. skipping.".format(out_file))
out_fns.append(out_file)
if op.exists(tmp_out_file):
os.remove(tmp_out_file)
else:
cmd = "pls2fasta {in_fn} ".format(in_fn=real_upath(in_fn)) + \
" {out} ".format(out=real_upath(tmp_out_file)) + \
"-minSubreadLength 300 -minReadScore 750 -trimByRegion"
print >> sys.stderr, "DEBUG: putting in queue:", cmd, tmp_out_file, out_file
in_queue.put((cmd, tmp_out_file, out_file))
in_queue_count += 1
outfile_track[out_file] = (cmd, tmp_out_file)
print >> sys.stderr, "DEBUG: put in queue:", cmd, tmp_out_file, out_file
cpus = min(cpus, in_queue_count) # cap max CPU if there's fewer files to convert
for i in xrange(cpus):
p = Process(target=convert_fofn_to_fasta_worker, args=(in_queue,))
pool.append(p)
#error_flag = False
# starting & joining pool worakers
for p in pool:
p.start()
#print >> sys.stderr, "Starting worker", p.name
for p in pool:
#print >> sys.stderr, "Waiting join", p.name
p.join(timeout=1200)
if p.is_alive(): p.terminate()
# check that all files exists
# if it does not, force to run locally
for out_file,(cmd, tmp_out_file) in outfile_track.iteritems():
in_queue.put((cmd, tmp_out_file, out_file))
convert_fofn_to_fasta_worker(in_queue)
out_fns.append(out_file)
#if error_flag:
# raise Exception, "Unable to successfuly run convert_fofn_to_fasta, ABORT!"
write_files_to_fofn(out_fns, out_filename)
def convert_fofn_to_fasta(fofn_filename, out_filename, fasta_out_dir,
force_overwrite=False, cpus=1):
"""
For each .bax.h5 file, create .bax.h5.fasta file and save paths to
out_filename, which should usually be 'input.fasta.fofn'
"""
logging.info("Converting fofn {fofn} to fasta.".format(fofn=fofn_filename))
in_fns = get_files_from_fofn(fofn_filename)
#out_fns = []
mkdir(fasta_out_dir)
# multiprocessing worker stuff
manager = Manager()
in_queue = manager.Queue(len(in_fns))
in_queue_count = 0
outfile_track = {} # expected out file --> (cmd, tmp)
pool = []
out_fns = []
for in_fn in in_fns:
#print >> sys.stderr, "DEBUG: converting h5 file:", in_fn
logging.debug("converting h5 file: {f}.".format(f=in_fn))
if not (in_fn.endswith('.bax.h5') or in_fn.endswith('.bas.h5')):
raise ValueError("fofn file {fofn} ".format(fofn=fofn_filename) +
"should only contain bax/bas.h5 files.")
# e.g. m111xxxx.1.bax.h5 ==>
# tmp_out_file = m11xxxx.1.bax.h5.fasta.tmp
# out_file = m11xxxx.1.bax.h5.fasta
in_basename = op.basename(in_fn)
tmp_out_file = op.join(fasta_out_dir, in_basename + '.fasta.tmp')
out_file = op.join(fasta_out_dir, in_basename + '.fasta')
if op.exists(out_file) and not force_overwrite:
logging.debug("File {0} already exists. skipping.".format(out_file))
out_fns.append(out_file)
if op.exists(tmp_out_file):
os.remove(tmp_out_file)
else:
cmd = "pls2fasta {in_fn} ".format(in_fn=real_upath(in_fn)) + \
" {out} ".format(out=real_upath(tmp_out_file)) + \
"-minSubreadLength 300 -minReadScore 750 -trimByRegion"
print >> sys.stderr, "DEBUG: putting in queue:", cmd, tmp_out_file, out_file
in_queue.put((cmd, tmp_out_file, out_file))
in_queue_count += 1
outfile_track[out_file] = (cmd, tmp_out_file)
print >> sys.stderr, "DEBUG: put in queue:", cmd, tmp_out_file, out_file
cpus = min(cpus, in_queue_count) # cap max CPU if there's fewer files to convert
for i in xrange(cpus):
p = Process(target=convert_fofn_to_fasta_worker, args=(in_queue,))
pool.append(p)
#error_flag = False
# starting & joining pool worakers
for p in pool:
p.start()
#print >> sys.stderr, "Starting worker", p.name
for p in pool:
#print >> sys.stderr, "Waiting join", p.name
p.join(timeout=1200)
if p.is_alive(): p.terminate()
# check that all files exists
# if it does not, force to run locally
for out_file,(cmd, tmp_out_file) in outfile_track.iteritems():
in_queue.put((cmd, tmp_out_file, out_file))
convert_fofn_to_fasta_worker(in_queue)
out_fns.append(out_file)
#if error_flag:
# raise Exception, "Unable to successfuly run convert_fofn_to_fasta, ABORT!"
write_files_to_fofn(out_fns, out_filename)
def convert_fofn_to_fasta(fofn_filename,
out_filename,
fasta_out_dir,
force_overwrite=False,
cpus=1):
"""
For each .bax.h5 file, create .bax.h5.fasta file and save paths to
out_filename, which should usually be 'input.fasta.fofn'
"""
logging.info("Converting fofn {fofn} to fasta.".format(fofn=fofn_filename))
in_fns = get_files_from_fofn(fofn_filename)
#out_fns = []
mkdir(fasta_out_dir)
# multiprocessing worker stuff
manager = Manager()
out_fns = manager.list()
in_queue = manager.Queue(99999)
pool = []
for i in xrange(cpus):
p = Process(target=convert_fofn_to_fasta_worker,
args=(in_queue, out_fns))
pool.append(p)
for in_fn in in_fns:
#print >> sys.stderr, "DEBUG: converting h5 file:", in_fn
logging.debug("converting h5 file: {f}.".format(f=in_fn))
if not (in_fn.endswith('.bax.h5') or in_fn.endswith('.bas.h5')):
raise ValueError("fofn file {fofn} ".format(fofn=fofn_filename) +
"should only contain bax/bas.h5 files.")
# e.g. m111xxxx.1.bax.h5 ==>
# tmp_out_file = m11xxxx.1.bax.h5.fasta.tmp
# out_file = m11xxxx.1.bax.h5.fasta
in_basename = op.basename(in_fn)
tmp_out_file = op.join(fasta_out_dir, in_basename + '.fasta.tmp')
out_file = op.join(fasta_out_dir, in_basename + '.fasta')
if op.exists(out_file) and not force_overwrite:
logging.debug(
"File {0} already exists. skipping.".format(out_file))
out_fns.append(out_file)
if op.exists(tmp_out_file):
os.remove(tmp_out_file)
else:
cmd = "pls2fasta {in_fn} ".format(in_fn=real_upath(in_fn)) + \
" {out} ".format(out=real_upath(tmp_out_file)) + \
"-minSubreadLength 300 -minReadScore 750 -trimByRegion"
print >> sys.stderr, "DEBUG: putting in queue:", cmd, tmp_out_file, out_file
in_queue.put((cmd, tmp_out_file, out_file))
print >> sys.stderr, "DEBUG: put in queue:", cmd, tmp_out_file, out_file
# logging.debug("CMD: {cmd}".format(cmd=cmd))
# _out, _code, _msg = backticks(cmd)
# if _code != 0:
# raise RuntimeError("CMD failed: {cmd}\n".format(cmd=cmd) + _msg)
# trim_subread_flanks(tmp_out_file, out_file)
# out_fns.append(out_file)
# if op.exists(tmp_out_file):
# os.remove(tmp_out_file)
# starting & joining pool worakers
for p in pool:
p.start()
#print >> sys.stderr, "Starting worker", p.name
for p in pool:
#print >> sys.stderr, "Waiting join", p.name
p.join()
write_files_to_fofn(out_fns, out_filename)
