/usr/share/pyshared/joblib/memory.py is in python-joblib 0.7.1-1.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 | """
A context object for caching a function's return value each time it
is called with the same input arguments.
"""
# Author: Gael Varoquaux <gael dot varoquaux at normalesup dot org>
# Copyright (c) 2009 Gael Varoquaux
# License: BSD Style, 3 clauses.
from __future__ import with_statement
import os
import shutil
import time
import pydoc
try:
import cPickle as pickle
except ImportError:
import pickle
import functools
import traceback
import warnings
import inspect
import json
# Local imports
from .hashing import hash
from .func_inspect import get_func_code, get_func_name, filter_args
from .logger import Logger, format_time
from . import numpy_pickle
from .disk import mkdirp, rm_subdirs
FIRST_LINE_TEXT = "# first line:"
# TODO: The following object should have a data store object as a sub
# object, and the interface to persist and query should be separated in
# the data store.
#
# This would enable creating 'Memory' objects with a different logic for
# pickling that would simply span a MemorizedFunc with the same
# store (or do we want to copy it to avoid cross-talks?), for instance to
# implement HDF5 pickling.
# TODO: Same remark for the logger, and probably use the Python logging
# mechanism.
def extract_first_line(func_code):
""" Extract the first line information from the function code
text if available.
"""
if func_code.startswith(FIRST_LINE_TEXT):
func_code = func_code.split('\n')
first_line = int(func_code[0][len(FIRST_LINE_TEXT):])
func_code = '\n'.join(func_code[1:])
else:
first_line = -1
return func_code, first_line
class JobLibCollisionWarning(UserWarning):
""" Warn that there might be a collision between names of functions.
"""
###############################################################################
# class `MemorizedFunc`
###############################################################################
class MemorizedFunc(Logger):
""" Callable object decorating a function for caching its return value
each time it is called.
All values are cached on the filesystem, in a deep directory
structure. Methods are provided to inspect the cache or clean it.
Attributes
----------
func: callable
The original, undecorated, function.
cachedir: string
Path to the base cache directory of the memory context.
ignore: list or None
List of variable names to ignore when choosing whether to
recompute.
mmap_mode: {None, 'r+', 'r', 'w+', 'c'}
The memmapping mode used when loading from cache
numpy arrays. See numpy.load for the meaning of the
arguments.
compress: boolean
Whether to zip the stored data on disk. Note that compressed
arrays cannot be read by memmapping.
verbose: int, optional
The verbosity flag, controls messages that are issued as
the function is evaluated.
"""
#-------------------------------------------------------------------------
# Public interface
#-------------------------------------------------------------------------
def __init__(self, func, cachedir, ignore=None, mmap_mode=None,
compress=False, verbose=1, timestamp=None):
"""
Parameters
----------
func: callable
The function to decorate
cachedir: string
The path of the base directory to use as a data store
ignore: list or None
List of variable names to ignore.
mmap_mode: {None, 'r+', 'r', 'w+', 'c'}, optional
The memmapping mode used when loading from cache
numpy arrays. See numpy.load for the meaning of the
arguments.
verbose: int, optional
Verbosity flag, controls the debug messages that are issued
as functions are evaluated. The higher, the more verbose
timestamp: float, optional
The reference time from which times in tracing messages
are reported.
"""
Logger.__init__(self)
self._verbose = verbose
self.cachedir = cachedir
self.func = func
self.mmap_mode = mmap_mode
self.compress = compress
if compress and mmap_mode is not None:
warnings.warn('Compressed results cannot be memmapped',
stacklevel=2)
if timestamp is None:
timestamp = time.time()
self.timestamp = timestamp
if ignore is None:
ignore = []
self.ignore = ignore
mkdirp(self.cachedir)
try:
functools.update_wrapper(self, func)
except:
" Objects like ufunc don't like that "
if inspect.isfunction(func):
doc = pydoc.TextDoc().document(func
).replace('\n', '\n\n', 1)
else:
# Pydoc does a poor job on other objects
doc = func.__doc__
self.__doc__ = 'Memoized version of %s' % doc
def __call__(self, *args, **kwargs):
# Compare the function code with the previous to see if the
# function code has changed
output_dir, argument_hash = self.get_output_dir(*args, **kwargs)
# FIXME: The statements below should be try/excepted
if not (self._check_previous_func_code(stacklevel=3) and
os.path.exists(output_dir)):
if self._verbose > 10:
_, name = get_func_name(self.func)
self.warn('Computing func %s, argument hash %s in '
'directory %s'
% (name, argument_hash, output_dir))
return self.call(*args, **kwargs)
else:
try:
t0 = time.time()
out = self.load_output(output_dir)
if self._verbose > 4:
t = time.time() - t0
_, name = get_func_name(self.func)
msg = '%s cache loaded - %s' % (name, format_time(t))
print(max(0, (80 - len(msg))) * '_' + msg)
return out
except Exception:
# XXX: Should use an exception logger
self.warn('Exception while loading results for '
'(args=%s, kwargs=%s)\n %s' %
(args, kwargs, traceback.format_exc()))
shutil.rmtree(output_dir, ignore_errors=True)
return self.call(*args, **kwargs)
def __reduce__(self):
""" We don't store the timestamp when pickling, to avoid the hash
depending from it.
In addition, when unpickling, we run the __init__
"""
return (self.__class__, (self.func, self.cachedir, self.ignore,
self.mmap_mode, self.compress, self._verbose))
#-------------------------------------------------------------------------
# Private interface
#-------------------------------------------------------------------------
def _get_func_dir(self, mkdir=True):
""" Get the directory corresponding to the cache for the
function.
"""
module, name = get_func_name(self.func)
module.append(name)
func_dir = os.path.join(self.cachedir, *module)
if mkdir:
mkdirp(func_dir)
return func_dir
def get_output_dir(self, *args, **kwargs):
""" Returns the directory in which are persisted the results
of the function corresponding to the given arguments.
The results can be loaded using the .load_output method.
"""
coerce_mmap = (self.mmap_mode is not None)
argument_hash = hash(filter_args(self.func, self.ignore,
args, kwargs),
coerce_mmap=coerce_mmap)
output_dir = os.path.join(self._get_func_dir(self.func),
argument_hash)
return output_dir, argument_hash
def _write_func_code(self, filename, func_code, first_line):
""" Write the function code and the filename to a file.
"""
func_code = '%s %i\n%s' % (FIRST_LINE_TEXT, first_line, func_code)
with open(filename, 'w') as out:
out.write(func_code)
def _check_previous_func_code(self, stacklevel=2):
"""
stacklevel is the depth a which this function is called, to
issue useful warnings to the user.
"""
# Here, we go through some effort to be robust to dynamically
# changing code and collision. We cannot inspect.getsource
# because it is not reliable when using IPython's magic "%run".
func_code, source_file, first_line = get_func_code(self.func)
func_dir = self._get_func_dir()
func_code_file = os.path.join(func_dir, 'func_code.py')
try:
with open(func_code_file) as infile:
old_func_code, old_first_line = \
extract_first_line(infile.read())
except IOError:
self._write_func_code(func_code_file, func_code, first_line)
return False
if old_func_code == func_code:
return True
# We have differing code, is this because we are referring to
# differing functions, or because the function we are referring as
# changed?
_, func_name = get_func_name(self.func, resolv_alias=False,
win_characters=False)
if old_first_line == first_line == -1 or func_name == '<lambda>':
if not first_line == -1:
func_description = '%s (%s:%i)' % (func_name,
source_file, first_line)
else:
func_description = func_name
warnings.warn(JobLibCollisionWarning(
"Cannot detect name collisions for function '%s'"
% func_description), stacklevel=stacklevel)
# Fetch the code at the old location and compare it. If it is the
# same than the code store, we have a collision: the code in the
# file has not changed, but the name we have is pointing to a new
# code block.
if not old_first_line == first_line and source_file is not None:
possible_collision = False
if os.path.exists(source_file):
_, func_name = get_func_name(self.func, resolv_alias=False)
num_lines = len(func_code.split('\n'))
with open(source_file) as f:
on_disk_func_code = f.readlines()[
old_first_line - 1
:old_first_line - 1 + num_lines - 1]
on_disk_func_code = ''.join(on_disk_func_code)
possible_collision = (on_disk_func_code.rstrip()
== old_func_code.rstrip())
else:
possible_collision = source_file.startswith('<doctest ')
if possible_collision:
warnings.warn(JobLibCollisionWarning(
'Possible name collisions between functions '
"'%s' (%s:%i) and '%s' (%s:%i)" %
(func_name, source_file, old_first_line,
func_name, source_file, first_line)),
stacklevel=stacklevel)
# The function has changed, wipe the cache directory.
# XXX: Should be using warnings, and giving stacklevel
if self._verbose > 10:
_, func_name = get_func_name(self.func, resolv_alias=False)
self.warn("Function %s (stored in %s) has changed." %
(func_name, func_dir))
self.clear(warn=True)
return False
def clear(self, warn=True):
""" Empty the function's cache.
"""
func_dir = self._get_func_dir(mkdir=False)
if self._verbose and warn:
self.warn("Clearing cache %s" % func_dir)
if os.path.exists(func_dir):
shutil.rmtree(func_dir, ignore_errors=True)
mkdirp(func_dir)
func_code, _, first_line = get_func_code(self.func)
func_code_file = os.path.join(func_dir, 'func_code.py')
self._write_func_code(func_code_file, func_code, first_line)
def call(self, *args, **kwargs):
""" Force the execution of the function with the given arguments and
persist the output values.
"""
start_time = time.time()
output_dir, argument_hash = self.get_output_dir(*args, **kwargs)
if self._verbose:
print(self.format_call(*args, **kwargs))
output = self.func(*args, **kwargs)
self._persist_output(output, output_dir)
self._persist_input(output_dir, *args, **kwargs)
duration = time.time() - start_time
if self._verbose:
_, name = get_func_name(self.func)
msg = '%s - %s' % (name, format_time(duration))
print(max(0, (80 - len(msg))) * '_' + msg)
return output
def format_call(self, *args, **kwds):
""" Returns a nicely formatted statement displaying the function
call with the given arguments.
"""
path, signature = self.format_signature(self.func, *args,
**kwds)
msg = '%s\n[Memory] Calling %s...\n%s' % (80 * '_', path, signature)
return msg
# XXX: Not using logging framework
#self.debug(msg)
def format_signature(self, func, *args, **kwds):
# XXX: This should be moved out to a function
# XXX: Should this use inspect.formatargvalues/formatargspec?
module, name = get_func_name(func)
module = [m for m in module if m]
if module:
module.append(name)
module_path = '.'.join(module)
else:
module_path = name
arg_str = list()
previous_length = 0
for arg in args:
arg = self.format(arg, indent=2)
if len(arg) > 1500:
arg = '%s...' % arg[:700]
if previous_length > 80:
arg = '\n%s' % arg
previous_length = len(arg)
arg_str.append(arg)
arg_str.extend(['%s=%s' % (v, self.format(i)) for v, i in
kwds.items()])
arg_str = ', '.join(arg_str)
signature = '%s(%s)' % (name, arg_str)
return module_path, signature
# Make make public
def _persist_output(self, output, dir):
""" Persist the given output tuple in the directory.
"""
try:
mkdirp(dir)
filename = os.path.join(dir, 'output.pkl')
numpy_pickle.dump(output, filename, compress=self.compress)
if self._verbose > 10:
print('Persisting in %s' % dir)
except OSError:
" Race condition in the creation of the directory "
def _persist_input(self, output_dir, *args, **kwargs):
""" Save a small summary of the call using json format in the
output directory.
"""
argument_dict = filter_args(self.func, self.ignore,
args, kwargs)
input_repr = dict((k, repr(v)) for k, v in argument_dict.items())
# This can fail do to race-conditions with multiple
# concurrent joblibs removing the file or the directory
try:
mkdirp(output_dir)
json.dump(
input_repr,
file(os.path.join(output_dir, 'input_args.json'), 'w'),
)
except:
pass
return input_repr
def load_output(self, output_dir):
""" Read the results of a previous calculation from the directory
it was cached in.
"""
if self._verbose > 1:
t = time.time() - self.timestamp
if self._verbose < 10:
print('[Memory]% 16s: Loading %s...' % (
format_time(t),
self.format_signature(self.func)[0]
))
else:
print('[Memory]% 16s: Loading %s from %s' % (
format_time(t),
self.format_signature(self.func)[0],
output_dir
))
filename = os.path.join(output_dir, 'output.pkl')
return numpy_pickle.load(filename,
mmap_mode=self.mmap_mode)
# XXX: Need a method to check if results are available.
#-------------------------------------------------------------------------
# Private `object` interface
#-------------------------------------------------------------------------
def __repr__(self):
return '%s(func=%s, cachedir=%s)' % (
self.__class__.__name__,
self.func,
repr(self.cachedir),
)
###############################################################################
# class `Memory`
###############################################################################
class Memory(Logger):
""" A context object for caching a function's return value each time it
is called with the same input arguments.
All values are cached on the filesystem, in a deep directory
structure.
see :ref:`memory_reference`
"""
#-------------------------------------------------------------------------
# Public interface
#-------------------------------------------------------------------------
def __init__(self, cachedir, mmap_mode=None, compress=False, verbose=1):
"""
Parameters
----------
cachedir: string or None
The path of the base directory to use as a data store
or None. If None is given, no caching is done and
the Memory object is completely transparent.
mmap_mode: {None, 'r+', 'r', 'w+', 'c'}, optional
The memmapping mode used when loading from cache
numpy arrays. See numpy.load for the meaning of the
arguments.
compress: boolean
Whether to zip the stored data on disk. Note that
compressed arrays cannot be read by memmapping.
verbose: int, optional
Verbosity flag, controls the debug messages that are issued
as functions are evaluated.
"""
# XXX: Bad explanation of the None value of cachedir
Logger.__init__(self)
self._verbose = verbose
self.mmap_mode = mmap_mode
self.timestamp = time.time()
self.compress = compress
if compress and mmap_mode is not None:
warnings.warn('Compressed results cannot be memmapped',
stacklevel=2)
if cachedir is None:
self.cachedir = None
else:
self.cachedir = os.path.join(cachedir, 'joblib')
mkdirp(self.cachedir)
def cache(self, func=None, ignore=None, verbose=None,
mmap_mode=False):
""" Decorates the given function func to only compute its return
value for input arguments not cached on disk.
Parameters
----------
func: callable, optional
The function to be decorated
ignore: list of strings
A list of arguments name to ignore in the hashing
verbose: integer, optional
The verbosity mode of the function. By default that
of the memory object is used.
mmap_mode: {None, 'r+', 'r', 'w+', 'c'}, optional
The memmapping mode used when loading from cache
numpy arrays. See numpy.load for the meaning of the
arguments. By default that of the memory object is used.
Returns
-------
decorated_func: MemorizedFunc object
The returned object is a MemorizedFunc object, that is
callable (behaves like a function), but offers extra
methods for cache lookup and management. See the
documentation for :class:`joblib.memory.MemorizedFunc`.
"""
if func is None:
# Partial application, to be able to specify extra keyword
# arguments in decorators
return functools.partial(self.cache, ignore=ignore)
if self.cachedir is None:
return func
if verbose is None:
verbose = self._verbose
if mmap_mode is False:
mmap_mode = self.mmap_mode
if isinstance(func, MemorizedFunc):
func = func.func
return MemorizedFunc(func, cachedir=self.cachedir,
mmap_mode=mmap_mode,
ignore=ignore,
compress=self.compress,
verbose=verbose,
timestamp=self.timestamp)
def clear(self, warn=True):
""" Erase the complete cache directory.
"""
if warn:
self.warn('Flushing completely the cache')
rm_subdirs(self.cachedir)
def eval(self, func, *args, **kwargs):
""" Eval function func with arguments `*args` and `**kwargs`,
in the context of the memory.
This method works similarly to the builtin `apply`, except
that the function is called only if the cache is not
up to date.
"""
if self.cachedir is None:
return func(*args, **kwargs)
return self.cache(func)(*args, **kwargs)
#-------------------------------------------------------------------------
# Private `object` interface
#-------------------------------------------------------------------------
def __repr__(self):
return '%s(cachedir=%s)' % (
self.__class__.__name__,
repr(self.cachedir),
)
def __reduce__(self):
""" We don't store the timestamp when pickling, to avoid the hash
depending from it.
In addition, when unpickling, we run the __init__
"""
# We need to remove 'joblib' from the end of cachedir
cachedir = self.cachedir[:-7] if self.cachedir is not None else None
return (self.__class__, (cachedir,
self.mmap_mode, self.compress, self._verbose))
|