def __repr__(self):
try:
if self._semlock._is_mine():
name = process.current_process().name
if threads_new.current_thread().name != 'MainThread':
name += '|' + threads_new.current_thread().name
elif self._semlock._get_value() == 1:
name = 'None'
elif self._semlock._count() > 0:
name = 'SomeOtherThread'
else:
name = 'SomeOtherProcess'
except Exception:
name = 'unknown'
return '<%s(owner=%s)>' % (self.__class__.__name__, name)
def __init__(self, path, threaded=True, timeout=None):
"""
>>> lock = LockBase('somefile')
>>> lock = LockBase('somefile', threaded=False)
"""
super(LockBase, self).__init__(path)
self.lock_file = os.path.abspath(path) + ".lock"
self.hostname = socket.gethostname()
self.pid = os.getpid()
if threaded:
t = threads_new.current_thread()
# Thread objects in Python 2.4 and earlier do not have ident
# attrs. Worm around that.
ident = getattr(t, "ident", hash(t))
self.tname = "-%x" % (ident & 0xffffffff)
else:
self.tname = ""
dirname = os.path.dirname(self.lock_file)
# unique name is mostly about the current process, but must
# also contain the path -- otherwise, two adjacent locked
# files conflict (one file gets locked, creating lock-file and
# unique file, the other one gets locked, creating lock-file
# and overwriting the already existing lock-file, then one
# gets unlocked, deleting both lock-file and unique file,
# finally the last lock errors out upon releasing.
self.unique_name = os.path.join(dirname,
"%s%s.%s%s" % (self.hostname,
self.tname,
self.pid,
hash(self.path)))
self.timeout = timeout
def concurrency_safe_write(object_to_write, filename, write_func):
"""Writes an object into a unique file in a concurrency-safe way."""
thread_id = id(threads_new.current_thread())
temporary_filename = '{}.thread-{}-pid-{}'.format(filename, thread_id,
os.getpid())
write_func(object_to_write, temporary_filename)
return temporary_filename
def _flag_current_thread_clean_exit():
"""Put a ``_clean_exit`` flag on the current thread"""
thread = threads_new.current_thread()
thread._clean_exit = True
def in_main_thread():
return isinstance(threads_new.current_thread(),
threads_new._MainThread)
def __call__(self, a):
m = _get_backing_memmap(a)
if m is not None and isinstance(m, np.memmap):
# a is already backed by a memmap file, let's reuse it directly
return _reduce_memmap_backed(a, m)
if (not a.dtype.hasobject and self._max_nbytes is not None
and a.nbytes > self._max_nbytes):
# check that the folder exists (lazily create the pool temp folder
# if required)
try:
os.makedirs(self._temp_folder)
os.chmod(self._temp_folder, FOLDER_PERMISSIONS)
except OSError as e:
if e.errno != errno.EEXIST:
raise e
try:
basename = self._memmaped_arrays.get(a)
except KeyError:
# Generate a new unique random filename. The process and thread
# ids are only useful for debugging purpose and to make it
# easier to cleanup orphaned files in case of hard process
# kill (e.g. by "kill -9" or segfault).
basename = "{}-{}-{}.pkl".format(
os.getpid(), id(threads_new.current_thread()),
uuid4().hex)
self._memmaped_arrays.set(a, basename)
filename = os.path.join(self._temp_folder, basename)
# In case the same array with the same content is passed several
# times to the pool subprocess children, serialize it only once
# XXX: implement an explicit reference counting scheme to make it
# possible to delete temporary files as soon as the workers are
# done processing this data.
if not os.path.exists(filename):
if self.verbose > 0:
print("Memmapping (shape={}, dtype={}) to new file {}".
format(a.shape, a.dtype, filename))
for dumped_filename in dump(a, filename):
os.chmod(dumped_filename, FILE_PERMISSIONS)
if self._prewarm:
# Warm up the data by accessing it. This operation ensures
# that the disk access required to create the memmapping
# file are performed in the reducing process and avoids
# concurrent memmap creation in multiple children
# processes.
load(filename, mmap_mode=self._mmap_mode).max()
elif self.verbose > 1:
print("Memmapping (shape={}, dtype={}) to old file {}".format(
a.shape, a.dtype, filename))
# The worker process will use joblib.load to memmap the data
return (load, (filename, self._mmap_mode))
else:
# do not convert a into memmap, let pickler do its usual copy with
# the default system pickler
if self.verbose > 1:
print("Pickling array (shape={}, dtype={}).".format(
a.shape, a.dtype))
return (loads, (dumps(a, protocol=HIGHEST_PROTOCOL), ))
def get_ident():
return threads_new.current_thread().ident
def exit(self):
self.was_killed.set()
if self is not current_thread():
self.join()
return self.report()
def in_main_thread() -> bool:
"""
True when the current thread is the main thread.
"""
return threads_new.current_thread().__class__.__name__ == "_MainThread"
