Shortcuts on this page
r m x toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
1"""
2APIs exposing metadata from third-party Python packages.
4This codebase is shared between importlib.metadata in the stdlib
5and importlib_metadata in PyPI. See
6https://github.com/python/importlib_metadata/wiki/Development-Methodology
7for more detail.
8"""
10from __future__ import annotations
12import abc
13import collections
14import email
15import functools
16import itertools
17import operator
18import os
19import pathlib
20import posixpath
21import re
22import sys
23import textwrap
24import types
25from collections.abc import Iterable, Mapping
26from contextlib import suppress
27from importlib import import_module
28from importlib.abc import MetaPathFinder
29from itertools import starmap
30from typing import Any
32from . import _meta
33from ._collections import FreezableDefaultDict, Pair
34from ._compat import (
35 NullFinder,
36 install,
37)
38from ._functools import method_cache, noop, pass_none, passthrough
39from ._itertools import always_iterable, bucket, unique_everseen
40from ._meta import PackageMetadata, SimplePath
41from ._typing import md_none
42from .compat import py39, py311
44__all__ = [
45 'Distribution',
46 'DistributionFinder',
47 'PackageMetadata',
48 'PackageNotFoundError',
49 'SimplePath',
50 'distribution',
51 'distributions',
52 'entry_points',
53 'files',
54 'metadata',
55 'packages_distributions',
56 'requires',
57 'version',
58]
61class PackageNotFoundError(ModuleNotFoundError):
62 """The package was not found."""
64 def __str__(self) -> str:
65 return f"No package metadata was found for {self.name}"
67 @property
68 def name(self) -> str: # type: ignore[override] # make readonly
69 (name,) = self.args
70 return name
73class Sectioned:
74 """
75 A simple entry point config parser for performance
77 >>> for item in Sectioned.read(Sectioned._sample):
78 ... print(item)
79 Pair(name='sec1', value='# comments ignored')
80 Pair(name='sec1', value='a = 1')
81 Pair(name='sec1', value='b = 2')
82 Pair(name='sec2', value='a = 2')
84 >>> res = Sectioned.section_pairs(Sectioned._sample)
85 >>> item = next(res)
86 >>> item.name
87 'sec1'
88 >>> item.value
89 Pair(name='a', value='1')
90 >>> item = next(res)
91 >>> item.value
92 Pair(name='b', value='2')
93 >>> item = next(res)
94 >>> item.name
95 'sec2'
96 >>> item.value
97 Pair(name='a', value='2')
98 >>> list(res)
99 []
100 """
102 _sample = textwrap.dedent(
103 """
104 [sec1]
105 # comments ignored
106 a = 1
107 b = 2
109 [sec2]
110 a = 2
111 """
112 ).lstrip()
114 @classmethod
115 def section_pairs(cls, text):
116 return (
117 section._replace(value=Pair.parse(section.value))
118 for section in cls.read(text, filter_=cls.valid)
119 if section.name is not None
120 )
122 @staticmethod
123 def read(text, filter_=None):
124 lines = filter(filter_, map(str.strip, text.splitlines()))
125 name = None
126 for value in lines:
127 section_match = value.startswith('[') and value.endswith(']')
128 if section_match:
129 name = value.strip('[]')
130 continue
131 yield Pair(name, value)
133 @staticmethod
134 def valid(line: str):
135 return line and not line.startswith('#')
138class _EntryPointMatch(types.SimpleNamespace):
139 module: str
140 attr: str
141 extras: str
144class EntryPoint:
145 """An entry point as defined by Python packaging conventions.
147 See `the packaging docs on entry points
148 <https://packaging.python.org/specifications/entry-points/>`_
149 for more information.
151 >>> ep = EntryPoint(
152 ... name=None, group=None, value='package.module:attr [extra1, extra2]')
153 >>> ep.module
154 'package.module'
155 >>> ep.attr
156 'attr'
157 >>> ep.extras
158 ['extra1', 'extra2']
160 If the value package or module are not valid identifiers, a
161 ValueError is raised on access.
163 >>> EntryPoint(name=None, group=None, value='invalid-name').module
164 Traceback (most recent call last):
165 ...
166 ValueError: ('Invalid object reference...invalid-name...
167 >>> EntryPoint(name=None, group=None, value='invalid-name').attr
168 Traceback (most recent call last):
169 ...
170 ValueError: ('Invalid object reference...invalid-name...
171 >>> EntryPoint(name=None, group=None, value='invalid-name').extras
172 Traceback (most recent call last):
173 ...
174 ValueError: ('Invalid object reference...invalid-name...
176 The same thing happens on construction.
178 >>> EntryPoint(name=None, group=None, value='invalid-name')
179 Traceback (most recent call last):
180 ...
181 ValueError: ('Invalid object reference...invalid-name...
183 """
185 pattern = re.compile(
186 r'(?P<module>[\w.]+)\s*'
187 r'(:\s*(?P<attr>[\w.]+)\s*)?'
188 r'((?P<extras>\[.*\])\s*)?$'
189 )
190 """
191 A regular expression describing the syntax for an entry point,
192 which might look like:
194 - module
195 - package.module
196 - package.module:attribute
197 - package.module:object.attribute
198 - package.module:attr [extra1, extra2]
200 Other combinations are possible as well.
202 The expression is lenient about whitespace around the ':',
203 following the attr, and following any extras.
204 """
206 name: str
207 value: str
208 group: str
210 dist: Distribution | None = None
212 def __init__(self, name: str, value: str, group: str) -> None:
213 vars(self).update(name=name, value=value, group=group)
214 self.module
216 def load(self) -> Any:
217 """Load the entry point from its definition. If only a module
218 is indicated by the value, return that module. Otherwise,
219 return the named object.
220 """
221 module = import_module(self.module)
222 attrs = filter(None, (self.attr or '').split('.'))
223 return functools.reduce(getattr, attrs, module)
225 @property
226 def module(self) -> str:
227 return self._match.module
229 @property
230 def attr(self) -> str:
231 return self._match.attr
233 @property
234 def extras(self) -> list[str]:
235 return re.findall(r'\w+', self._match.extras or '')
237 @functools.cached_property
238 def _match(self) -> _EntryPointMatch:
239 match = self.pattern.match(self.value)
240 if not match:
241 raise ValueError(
242 'Invalid object reference. '
243 'See https://packaging.python.org'
244 '/en/latest/specifications/entry-points/#data-model',
245 self.value,
246 )
247 return _EntryPointMatch(**match.groupdict())
249 def _for(self, dist):
250 vars(self).update(dist=dist)
251 return self
253 def matches(self, **params):
254 """
255 EntryPoint matches the given parameters.
257 >>> ep = EntryPoint(group='foo', name='bar', value='bing:bong [extra1, extra2]')
258 >>> ep.matches(group='foo')
259 True
260 >>> ep.matches(name='bar', value='bing:bong [extra1, extra2]')
261 True
262 >>> ep.matches(group='foo', name='other')
263 False
264 >>> ep.matches()
265 True
266 >>> ep.matches(extras=['extra1', 'extra2'])
267 True
268 >>> ep.matches(module='bing')
269 True
270 >>> ep.matches(attr='bong')
271 True
272 """
273 self._disallow_dist(params)
274 attrs = (getattr(self, param) for param in params)
275 return all(map(operator.eq, params.values(), attrs))
277 @staticmethod
278 def _disallow_dist(params):
279 """
280 Querying by dist is not allowed (dist objects are not comparable).
281 >>> EntryPoint(name='fan', value='fav', group='fag').matches(dist='foo')
282 Traceback (most recent call last):
283 ...
284 ValueError: "dist" is not suitable for matching...
285 """
286 if "dist" in params:
287 raise ValueError(
288 '"dist" is not suitable for matching. '
289 "Instead, use Distribution.entry_points.select() on a "
290 "located distribution."
291 )
293 def _key(self):
294 return self.name, self.value, self.group
296 def __lt__(self, other):
297 return self._key() < other._key()
299 def __eq__(self, other):
300 return self._key() == other._key()
302 def __setattr__(self, name, value):
303 raise AttributeError("EntryPoint objects are immutable.")
305 def __repr__(self):
306 return (
307 f'EntryPoint(name={self.name!r}, value={self.value!r}, '
308 f'group={self.group!r})'
309 )
311 def __hash__(self) -> int:
312 return hash(self._key())
315class EntryPoints(tuple):
316 """
317 An immutable collection of selectable EntryPoint objects.
318 """
320 __slots__ = ()
322 def __getitem__(self, name: str) -> EntryPoint: # type: ignore[override] # Work with str instead of int
323 """
324 Get the EntryPoint in self matching name.
325 """
326 try:
327 return next(iter(self.select(name=name)))
328 except StopIteration:
329 raise KeyError(name)
331 def __repr__(self):
332 """
333 Repr with classname and tuple constructor to
334 signal that we deviate from regular tuple behavior.
335 """
336 return '%s(%r)' % (self.__class__.__name__, tuple(self))
338 def select(self, **params) -> EntryPoints:
339 """
340 Select entry points from self that match the
341 given parameters (typically group and/or name).
342 """
343 return EntryPoints(ep for ep in self if py39.ep_matches(ep, **params))
345 @property
346 def names(self) -> set[str]:
347 """
348 Return the set of all names of all entry points.
349 """
350 return {ep.name for ep in self}
352 @property
353 def groups(self) -> set[str]:
354 """
355 Return the set of all groups of all entry points.
356 """
357 return {ep.group for ep in self}
359 @classmethod
360 def _from_text_for(cls, text, dist):
361 return cls(ep._for(dist) for ep in cls._from_text(text))
363 @staticmethod
364 def _from_text(text):
365 return (
366 EntryPoint(name=item.value.name, value=item.value.value, group=item.name)
367 for item in Sectioned.section_pairs(text or '')
368 )
371class PackagePath(pathlib.PurePosixPath):
372 """A reference to a path in a package"""
374 hash: FileHash | None
375 size: int
376 dist: Distribution
378 def read_text(self, encoding: str = 'utf-8') -> str:
379 return self.locate().read_text(encoding=encoding)
381 def read_binary(self) -> bytes:
382 return self.locate().read_bytes()
384 def locate(self) -> SimplePath:
385 """Return a path-like object for this path"""
386 return self.dist.locate_file(self)
389class FileHash:
390 def __init__(self, spec: str) -> None:
391 self.mode, _, self.value = spec.partition('=')
393 def __repr__(self) -> str:
394 return f'<FileHash mode: {self.mode} value: {self.value}>'
397class Distribution(metaclass=abc.ABCMeta):
398 """
399 An abstract Python distribution package.
401 Custom providers may derive from this class and define
402 the abstract methods to provide a concrete implementation
403 for their environment. Some providers may opt to override
404 the default implementation of some properties to bypass
405 the file-reading mechanism.
406 """
408 @abc.abstractmethod
409 def read_text(self, filename) -> str | None:
410 """Attempt to load metadata file given by the name.
412 Python distribution metadata is organized by blobs of text
413 typically represented as "files" in the metadata directory
414 (e.g. package-1.0.dist-info). These files include things
415 like:
417 - METADATA: The distribution metadata including fields
418 like Name and Version and Description.
419 - entry_points.txt: A series of entry points as defined in
420 `the entry points spec <https://packaging.python.org/en/latest/specifications/entry-points/#file-format>`_.
421 - RECORD: A record of files according to
422 `this recording spec <https://packaging.python.org/en/latest/specifications/recording-installed-packages/#the-record-file>`_.
424 A package may provide any set of files, including those
425 not listed here or none at all.
427 :param filename: The name of the file in the distribution info.
428 :return: The text if found, otherwise None.
429 """
431 @abc.abstractmethod
432 def locate_file(self, path: str | os.PathLike[str]) -> SimplePath:
433 """
434 Given a path to a file in this distribution, return a SimplePath
435 to it.
437 This method is used by callers of ``Distribution.files()`` to
438 locate files within the distribution. If it's possible for a
439 Distribution to represent files in the distribution as
440 ``SimplePath`` objects, it should implement this method
441 to resolve such objects.
443 Some Distribution providers may elect not to resolve SimplePath
444 objects within the distribution by raising a
445 NotImplementedError, but consumers of such a Distribution would
446 be unable to invoke ``Distribution.files()``.
447 """
449 @classmethod
450 def from_name(cls, name: str) -> Distribution:
451 """Return the Distribution for the given package name.
453 :param name: The name of the distribution package to search for.
454 :return: The Distribution instance (or subclass thereof) for the named
455 package, if found.
456 :raises PackageNotFoundError: When the named package's distribution
457 metadata cannot be found.
458 :raises ValueError: When an invalid value is supplied for name.
459 """
460 if not name:
461 raise ValueError("A distribution name is required.")
462 try:
463 return next(iter(cls._prefer_valid(cls.discover(name=name))))
464 except StopIteration:
465 raise PackageNotFoundError(name)
467 @classmethod
468 def discover(
469 cls, *, context: DistributionFinder.Context | None = None, **kwargs
470 ) -> Iterable[Distribution]:
471 """Return an iterable of Distribution objects for all packages.
473 Pass a ``context`` or pass keyword arguments for constructing
474 a context.
476 :context: A ``DistributionFinder.Context`` object.
477 :return: Iterable of Distribution objects for packages matching
478 the context.
479 """
480 if context and kwargs:
481 raise ValueError("cannot accept context and kwargs")
482 context = context or DistributionFinder.Context(**kwargs)
483 return itertools.chain.from_iterable(
484 resolver(context) for resolver in cls._discover_resolvers()
485 )
487 @staticmethod
488 def _prefer_valid(dists: Iterable[Distribution]) -> Iterable[Distribution]:
489 """
490 Prefer (move to the front) distributions that have metadata.
492 Ref python/importlib_resources#489.
493 """
494 buckets = bucket(dists, lambda dist: bool(dist.metadata))
495 return itertools.chain(buckets[True], buckets[False])
497 @staticmethod
498 def at(path: str | os.PathLike[str]) -> Distribution:
499 """Return a Distribution for the indicated metadata path.
501 :param path: a string or path-like object
502 :return: a concrete Distribution instance for the path
503 """
504 return PathDistribution(pathlib.Path(path))
506 @staticmethod
507 def _discover_resolvers():
508 """Search the meta_path for resolvers (MetadataPathFinders)."""
509 declared = (
510 getattr(finder, 'find_distributions', None) for finder in sys.meta_path
511 )
512 return filter(None, declared)
514 @property
515 def metadata(self) -> _meta.PackageMetadata | None:
516 """Return the parsed metadata for this Distribution.
518 The returned object will have keys that name the various bits of
519 metadata per the
520 `Core metadata specifications <https://packaging.python.org/en/latest/specifications/core-metadata/#core-metadata>`_.
522 Custom providers may provide the METADATA file or override this
523 property.
524 """
526 text = (
527 self.read_text('METADATA')
528 or self.read_text('PKG-INFO')
529 # This last clause is here to support old egg-info files. Its
530 # effect is to just end up using the PathDistribution's self._path
531 # (which points to the egg-info file) attribute unchanged.
532 or self.read_text('')
533 )
534 return self._assemble_message(text)
536 @staticmethod
537 @pass_none
538 def _assemble_message(text: str) -> _meta.PackageMetadata:
539 # deferred for performance (python/cpython#109829)
540 from . import _adapters
542 return _adapters.Message(email.message_from_string(text))
544 @property
545 def name(self) -> str:
546 """Return the 'Name' metadata for the distribution package."""
547 return md_none(self.metadata)['Name']
549 @property
550 def _normalized_name(self):
551 """Return a normalized version of the name."""
552 return Prepared.normalize(self.name)
554 @property
555 def version(self) -> str:
556 """Return the 'Version' metadata for the distribution package."""
557 return md_none(self.metadata)['Version']
559 @property
560 def entry_points(self) -> EntryPoints:
561 """
562 Return EntryPoints for this distribution.
564 Custom providers may provide the ``entry_points.txt`` file
565 or override this property.
566 """
567 return EntryPoints._from_text_for(self.read_text('entry_points.txt'), self)
569 @property
570 def files(self) -> list[PackagePath] | None:
571 """Files in this distribution.
573 :return: List of PackagePath for this distribution or None
575 Result is `None` if the metadata file that enumerates files
576 (i.e. RECORD for dist-info, or installed-files.txt or
577 SOURCES.txt for egg-info) is missing.
578 Result may be empty if the metadata exists but is empty.
580 Custom providers are recommended to provide a "RECORD" file (in
581 ``read_text``) or override this property to allow for callers to be
582 able to resolve filenames provided by the package.
583 """
585 def make_file(name, hash=None, size_str=None):
586 result = PackagePath(name)
587 result.hash = FileHash(hash) if hash else None
588 result.size = int(size_str) if size_str else None
589 result.dist = self
590 return result
592 @pass_none
593 def make_files(lines):
594 # Delay csv import, since Distribution.files is not as widely used
595 # as other parts of importlib.metadata
596 import csv
598 return starmap(make_file, csv.reader(lines))
600 @pass_none
601 def skip_missing_files(package_paths):
602 return list(filter(lambda path: path.locate().exists(), package_paths))
604 return skip_missing_files(
605 make_files(
606 self._read_files_distinfo()
607 or self._read_files_egginfo_installed()
608 or self._read_files_egginfo_sources()
609 )
610 )
612 def _read_files_distinfo(self):
613 """
614 Read the lines of RECORD.
615 """
616 text = self.read_text('RECORD')
617 return text and text.splitlines()
619 def _read_files_egginfo_installed(self):
620 """
621 Read installed-files.txt and return lines in a similar
622 CSV-parsable format as RECORD: each file must be placed
623 relative to the site-packages directory and must also be
624 quoted (since file names can contain literal commas).
626 This file is written when the package is installed by pip,
627 but it might not be written for other installation methods.
628 Assume the file is accurate if it exists.
629 """
630 text = self.read_text('installed-files.txt')
631 # Prepend the .egg-info/ subdir to the lines in this file.
632 # But this subdir is only available from PathDistribution's
633 # self._path.
634 subdir = getattr(self, '_path', None)
635 if not text or not subdir:
636 return
638 paths = (
639 py311
640 .relative_fix((subdir / name).resolve())
641 .relative_to(self.locate_file('').resolve(), walk_up=True)
642 .as_posix()
643 for name in text.splitlines()
644 )
645 return map('"{}"'.format, paths)
647 def _read_files_egginfo_sources(self):
648 """
649 Read SOURCES.txt and return lines in a similar CSV-parsable
650 format as RECORD: each file name must be quoted (since it
651 might contain literal commas).
653 Note that SOURCES.txt is not a reliable source for what
654 files are installed by a package. This file is generated
655 for a source archive, and the files that are present
656 there (e.g. setup.py) may not correctly reflect the files
657 that are present after the package has been installed.
658 """
659 text = self.read_text('SOURCES.txt')
660 return text and map('"{}"'.format, text.splitlines())
662 @property
663 def requires(self) -> list[str] | None:
664 """Generated requirements specified for this Distribution"""
665 reqs = self._read_dist_info_reqs() or self._read_egg_info_reqs()
666 return reqs and list(reqs)
668 def _read_dist_info_reqs(self):
669 return self.metadata.get_all('Requires-Dist')
671 def _read_egg_info_reqs(self):
672 source = self.read_text('requires.txt')
673 return pass_none(self._deps_from_requires_text)(source)
675 @classmethod
676 def _deps_from_requires_text(cls, source):
677 return cls._convert_egg_info_reqs_to_simple_reqs(Sectioned.read(source))
679 @staticmethod
680 def _convert_egg_info_reqs_to_simple_reqs(sections):
681 """
682 Historically, setuptools would solicit and store 'extra'
683 requirements, including those with environment markers,
684 in separate sections. More modern tools expect each
685 dependency to be defined separately, with any relevant
686 extras and environment markers attached directly to that
687 requirement. This method converts the former to the
688 latter. See _test_deps_from_requires_text for an example.
689 """
691 def make_condition(name):
692 return name and f'extra == "{name}"'
694 def quoted_marker(section):
695 section = section or ''
696 extra, sep, markers = section.partition(':')
697 if extra and markers:
698 markers = f'({markers})'
699 conditions = list(filter(None, [markers, make_condition(extra)]))
700 return '; ' + ' and '.join(conditions) if conditions else ''
702 def url_req_space(req):
703 """
704 PEP 508 requires a space between the url_spec and the quoted_marker.
705 Ref python/importlib_metadata#357.
706 """
707 # '@' is uniquely indicative of a url_req.
708 return ' ' * ('@' in req)
710 for section in sections:
711 space = url_req_space(section.value)
712 yield section.value + space + quoted_marker(section.name)
714 @property
715 def origin(self):
716 return self._load_json('direct_url.json')
718 def _load_json(self, filename):
719 # Deferred for performance (python/importlib_metadata#503)
720 import json
722 return pass_none(json.loads)(
723 self.read_text(filename),
724 object_hook=lambda data: types.SimpleNamespace(**data),
725 )
728class DistributionFinder(MetaPathFinder):
729 """
730 A MetaPathFinder capable of discovering installed distributions.
732 Custom providers should implement this interface in order to
733 supply metadata.
734 """
736 class Context:
737 """
738 Keyword arguments presented by the caller to
739 ``distributions()`` or ``Distribution.discover()``
740 to narrow the scope of a search for distributions
741 in all DistributionFinders.
743 Each DistributionFinder may expect any parameters
744 and should attempt to honor the canonical
745 parameters defined below when appropriate.
747 This mechanism gives a custom provider a means to
748 solicit additional details from the caller beyond
749 "name" and "path" when searching distributions.
750 For example, imagine a provider that exposes suites
751 of packages in either a "public" or "private" ``realm``.
752 A caller may wish to query only for distributions in
753 a particular realm and could call
754 ``distributions(realm="private")`` to signal to the
755 custom provider to only include distributions from that
756 realm.
757 """
759 name = None
760 """
761 Specific name for which a distribution finder should match.
762 A name of ``None`` matches all distributions.
763 """
765 def __init__(self, **kwargs):
766 vars(self).update(kwargs)
768 @property
769 def path(self) -> list[str]:
770 """
771 The sequence of directory path that a distribution finder
772 should search.
774 Typically refers to Python installed package paths such as
775 "site-packages" directories and defaults to ``sys.path``.
776 """
777 return vars(self).get('path', sys.path)
779 @abc.abstractmethod
780 def find_distributions(self, context=Context()) -> Iterable[Distribution]:
781 """
782 Find distributions.
784 Return an iterable of all Distribution instances capable of
785 loading the metadata for packages matching the ``context``,
786 a DistributionFinder.Context instance.
787 """
790@passthrough
791def _clear_after_fork(cached):
792 """Ensure ``func`` clears cached state after ``fork`` when supported.
794 ``FastPath`` caches zip-backed ``pathlib.Path`` objects that retain a
795 reference to the parent's open ``ZipFile`` handle. Re-using a cached
796 instance in a forked child can therefore resurrect invalid file pointers
797 and trigger ``BadZipFile``/``OSError`` failures (python/importlib_metadata#520).
798 Registering ``cache_clear`` with ``os.register_at_fork`` keeps each process
799 on its own cache.
800 """
801 getattr(os, 'register_at_fork', noop)(after_in_child=cached.cache_clear)
804class FastPath:
805 """
806 Micro-optimized class for searching a root for children.
808 Root is a path on the file system that may contain metadata
809 directories either as natural directories or within a zip file.
811 >>> FastPath('').children()
812 ['...']
814 FastPath objects are cached and recycled for any given root.
816 >>> FastPath('foobar') is FastPath('foobar')
817 True
818 """
820 @_clear_after_fork # type: ignore[misc]
821 @functools.lru_cache()
822 def __new__(cls, root):
823 return super().__new__(cls)
825 def __init__(self, root):
826 self.root = root
828 def joinpath(self, child):
829 return pathlib.Path(self.root, child)
831 def children(self):
832 with suppress(Exception):
833 return os.listdir(self.root or '.')
834 with suppress(Exception):
835 return self.zip_children()
836 return []
838 def zip_children(self):
839 # deferred for performance (python/importlib_metadata#502)
840 from zipp.compat.overlay import zipfile
842 zip_path = zipfile.Path(self.root)
843 names = zip_path.root.namelist()
844 self.joinpath = zip_path.joinpath
846 return dict.fromkeys(child.split(posixpath.sep, 1)[0] for child in names)
848 def search(self, name):
849 return self.lookup(self.mtime).search(name)
851 @property
852 def mtime(self):
853 with suppress(OSError):
854 return os.stat(self.root).st_mtime
855 self.lookup.cache_clear()
857 @method_cache
858 def lookup(self, mtime):
859 return Lookup(self)
862class Lookup:
863 """
864 A micro-optimized class for searching a (fast) path for metadata.
865 """
867 def __init__(self, path: FastPath):
868 """
869 Calculate all of the children representing metadata.
871 From the children in the path, calculate early all of the
872 children that appear to represent metadata (infos) or legacy
873 metadata (eggs).
874 """
876 base = os.path.basename(path.root).lower()
877 base_is_egg = base.endswith(".egg")
878 self.infos = FreezableDefaultDict(list)
879 self.eggs = FreezableDefaultDict(list)
881 for child in path.children():
882 low = child.lower()
883 if low.endswith((".dist-info", ".egg-info")):
884 # rpartition is faster than splitext and suitable for this purpose.
885 name = low.rpartition(".")[0].partition("-")[0]
886 normalized = Prepared.normalize(name)
887 self.infos[normalized].append(path.joinpath(child))
888 elif base_is_egg and low == "egg-info":
889 name = base.rpartition(".")[0].partition("-")[0]
890 legacy_normalized = Prepared.legacy_normalize(name)
891 self.eggs[legacy_normalized].append(path.joinpath(child))
893 self.infos.freeze()
894 self.eggs.freeze()
896 def search(self, prepared: Prepared):
897 """
898 Yield all infos and eggs matching the Prepared query.
899 """
900 infos = (
901 self.infos[prepared.normalized]
902 if prepared
903 else itertools.chain.from_iterable(self.infos.values())
904 )
905 eggs = (
906 self.eggs[prepared.legacy_normalized]
907 if prepared
908 else itertools.chain.from_iterable(self.eggs.values())
909 )
910 return itertools.chain(infos, eggs)
913class Prepared:
914 """
915 A prepared search query for metadata on a possibly-named package.
917 Pre-calculates the normalization to prevent repeated operations.
919 >>> none = Prepared(None)
920 >>> none.normalized
921 >>> none.legacy_normalized
922 >>> bool(none)
923 False
924 >>> sample = Prepared('Sample__Pkg-name.foo')
925 >>> sample.normalized
926 'sample_pkg_name_foo'
927 >>> sample.legacy_normalized
928 'sample__pkg_name.foo'
929 >>> bool(sample)
930 True
931 """
933 normalized = None
934 legacy_normalized = None
936 def __init__(self, name: str | None):
937 self.name = name
938 if name is None:
939 return
940 self.normalized = self.normalize(name)
941 self.legacy_normalized = self.legacy_normalize(name)
943 @staticmethod
944 def normalize(name):
945 """
946 PEP 503 normalization plus dashes as underscores.
947 """
948 return re.sub(r"[-_.]+", "-", name).lower().replace('-', '_')
950 @staticmethod
951 def legacy_normalize(name):
952 """
953 Normalize the package name as found in the convention in
954 older packaging tools versions and specs.
955 """
956 return name.lower().replace('-', '_')
958 def __bool__(self):
959 return bool(self.name)
962@install
963class MetadataPathFinder(NullFinder, DistributionFinder):
964 """A degenerate finder for distribution packages on the file system.
966 This finder supplies only a find_distributions() method for versions
967 of Python that do not have a PathFinder find_distributions().
968 """
970 @classmethod
971 def find_distributions(
972 cls, context=DistributionFinder.Context()
973 ) -> Iterable[PathDistribution]:
974 """
975 Find distributions.
977 Return an iterable of all Distribution instances capable of
978 loading the metadata for packages matching ``context.name``
979 (or all names if ``None`` indicated) along the paths in the list
980 of directories ``context.path``.
981 """
982 found = cls._search_paths(context.name, context.path)
983 return map(PathDistribution, found)
985 @classmethod
986 def _search_paths(cls, name, paths):
987 """Find metadata directories in paths heuristically."""
988 prepared = Prepared(name)
989 return itertools.chain.from_iterable(
990 path.search(prepared) for path in map(FastPath, paths)
991 )
993 @classmethod
994 def invalidate_caches(cls) -> None:
995 FastPath.__new__.cache_clear()
998class PathDistribution(Distribution):
999 def __init__(self, path: SimplePath) -> None:
1000 """Construct a distribution.
1002 :param path: SimplePath indicating the metadata directory.
1003 """
1004 self._path = path
1006 def read_text(self, filename: str | os.PathLike[str]) -> str | None:
1007 with suppress(
1008 FileNotFoundError,
1009 IsADirectoryError,
1010 KeyError,
1011 NotADirectoryError,
1012 PermissionError,
1013 ):
1014 return self._path.joinpath(filename).read_text(encoding='utf-8')
1016 return None
1018 read_text.__doc__ = Distribution.read_text.__doc__
1020 def locate_file(self, path: str | os.PathLike[str]) -> SimplePath:
1021 return self._path.parent / path
1023 @property
1024 def _normalized_name(self):
1025 """
1026 Performance optimization: where possible, resolve the
1027 normalized name from the file system path.
1028 """
1029 stem = os.path.basename(str(self._path))
1030 return (
1031 pass_none(Prepared.normalize)(self._name_from_stem(stem))
1032 or super()._normalized_name
1033 )
1035 @staticmethod
1036 def _name_from_stem(stem):
1037 """
1038 >>> PathDistribution._name_from_stem('foo-3.0.egg-info')
1039 'foo'
1040 >>> PathDistribution._name_from_stem('CherryPy-3.0.dist-info')
1041 'CherryPy'
1042 >>> PathDistribution._name_from_stem('face.egg-info')
1043 'face'
1044 >>> PathDistribution._name_from_stem('foo.bar')
1045 """
1046 filename, ext = os.path.splitext(stem)
1047 if ext not in ('.dist-info', '.egg-info'):
1048 return
1049 name, sep, rest = filename.partition('-')
1050 return name
1053def distribution(distribution_name: str) -> Distribution:
1054 """Get the ``Distribution`` instance for the named package.
1056 :param distribution_name: The name of the distribution package as a string.
1057 :return: A ``Distribution`` instance (or subclass thereof).
1058 """
1059 return Distribution.from_name(distribution_name)
1062def distributions(**kwargs) -> Iterable[Distribution]:
1063 """Get all ``Distribution`` instances in the current environment.
1065 :return: An iterable of ``Distribution`` instances.
1066 """
1067 return Distribution.discover(**kwargs)
1070def metadata(distribution_name: str) -> _meta.PackageMetadata | None:
1071 """Get the metadata for the named package.
1073 :param distribution_name: The name of the distribution package to query.
1074 :return: A PackageMetadata containing the parsed metadata.
1075 """
1076 return Distribution.from_name(distribution_name).metadata
1079def version(distribution_name: str) -> str:
1080 """Get the version string for the named package.
1082 :param distribution_name: The name of the distribution package to query.
1083 :return: The version string for the package as defined in the package's
1084 "Version" metadata key.
1085 """
1086 return distribution(distribution_name).version
1089_unique = functools.partial(
1090 unique_everseen,
1091 key=py39.normalized_name,
1092)
1093"""
1094Wrapper for ``distributions`` to return unique distributions by name.
1095"""
1098def entry_points(**params) -> EntryPoints:
1099 """Return EntryPoint objects for all installed packages.
1101 Pass selection parameters (group or name) to filter the
1102 result to entry points matching those properties (see
1103 EntryPoints.select()).
1105 :return: EntryPoints for all installed packages.
1106 """
1107 eps = itertools.chain.from_iterable(
1108 dist.entry_points for dist in _unique(distributions())
1109 )
1110 return EntryPoints(eps).select(**params)
1113def files(distribution_name: str) -> list[PackagePath] | None:
1114 """Return a list of files for the named package.
1116 :param distribution_name: The name of the distribution package to query.
1117 :return: List of files composing the distribution.
1118 """
1119 return distribution(distribution_name).files
1122def requires(distribution_name: str) -> list[str] | None:
1123 """
1124 Return a list of requirements for the named package.
1126 :return: An iterable of requirements, suitable for
1127 packaging.requirement.Requirement.
1128 """
1129 return distribution(distribution_name).requires
1132def packages_distributions() -> Mapping[str, list[str]]:
1133 """
1134 Return a mapping of top-level packages to their
1135 distributions.
1137 >>> import collections.abc
1138 >>> pkgs = packages_distributions()
1139 >>> all(isinstance(dist, collections.abc.Sequence) for dist in pkgs.values())
1140 True
1141 """
1142 pkg_to_dist = collections.defaultdict(list)
1143 for dist in distributions():
1144 for pkg in _top_level_declared(dist) or _top_level_inferred(dist):
1145 pkg_to_dist[pkg].append(md_none(dist.metadata)['Name'])
1146 return dict(pkg_to_dist)
1149def _top_level_declared(dist):
1150 return (dist.read_text('top_level.txt') or '').split()
1153def _topmost(name: PackagePath) -> str | None:
1154 """
1155 Return the top-most parent as long as there is a parent.
1156 """
1157 top, *rest = name.parts
1158 return top if rest else None
1161def _get_toplevel_name(name: PackagePath) -> str:
1162 """
1163 Infer a possibly importable module name from a name presumed on
1164 sys.path.
1166 >>> _get_toplevel_name(PackagePath('foo.py'))
1167 'foo'
1168 >>> _get_toplevel_name(PackagePath('foo'))
1169 'foo'
1170 >>> _get_toplevel_name(PackagePath('foo.pyc'))
1171 'foo'
1172 >>> _get_toplevel_name(PackagePath('foo/__init__.py'))
1173 'foo'
1174 >>> _get_toplevel_name(PackagePath('foo.pth'))
1175 'foo.pth'
1176 >>> _get_toplevel_name(PackagePath('foo.dist-info'))
1177 'foo.dist-info'
1178 """
1179 # Defer import of inspect for performance (python/cpython#118761)
1180 import inspect
1182 return _topmost(name) or inspect.getmodulename(name) or str(name)
1185def _top_level_inferred(dist):
1186 opt_names = set(map(_get_toplevel_name, always_iterable(dist.files)))
1188 def importable_name(name):
1189 return '.' not in name
1191 return filter(importable_name, opt_names)