GH-73435: Add pathlib.PurePath.full_match() (#114350)

In 49f90ba we added support for the recursive wildcard `**` in
`pathlib.PurePath.match()`. This should allow arbitrary prefix and suffix
matching, like `p.match('foo/**')` or `p.match('**/foo')`, but there's a
problem: for relative patterns only, `match()` implicitly inserts a `**`
token on the left hand side, causing all patterns to match from the right.
As a result, it's impossible to match relative patterns from the left:
`PurePath('foo/bar').match('bar/**')` is true!

This commit reverts the changes to `match()`, and instead adds a new
`full_match()` method that:

- Allows empty patterns
- Supports the recursive wildcard `**`
- Matches the *entire* path when given a relative pattern

Doc/library/glob.rst

@@ -147,8 +147,9 @@ The :mod:`glob` module defines the following functions:
 
  .. seealso::
 
- :meth:`pathlib.PurePath.match` and :meth:`pathlib.Path.glob` methods,
- which call this function to implement pattern matching and globbing.
+ :meth:`pathlib.PurePath.full_match` and :meth:`pathlib.Path.glob`
+ methods, which call this function to implement pattern matching and
+ globbing.
 
  .. versionadded:: 3.13
 

Doc/library/pathlib.rst

@@ -559,55 +559,55 @@ Pure paths provide the following methods and properties:
  PureWindowsPath('c:/Program Files')
 
 
-.. method:: PurePath.match(pattern, *, case_sensitive=None)
+.. method:: PurePath.full_match(pattern, *, case_sensitive=None)
 
  Match this path against the provided glob-style pattern. Return ``True``
- if matching is successful, ``False`` otherwise.
-
- If *pattern* is relative, the path can be either relative or absolute,
- and matching is done from the right::
+ if matching is successful, ``False`` otherwise. For example::
 
- >>> PurePath('a/b.py').match('*.py')
- True
- >>> PurePath('/a/b/c.py').match('b/*.py')
+ >>> PurePath('a/b.py').full_match('a/*.py')
  True
- >>> PurePath('/a/b/c.py').match('a/*.py')
+ >>> PurePath('a/b.py').full_match('*.py')
  False
+ >>> PurePath('/a/b/c.py').full_match('/a/**')
+ True
+ >>> PurePath('/a/b/c.py').full_match('**/*.py')
+ True
 
- If *pattern* is absolute, the path must be absolute, and the whole path
- must match::
+ As with other methods, case-sensitivity follows platform defaults::
 
- >>> PurePath('/a.py').match('/*.py')
- True
- >>> PurePath('a/b.py').match('/*.py')
+ >>> PurePosixPath('b.py').full_match('*.PY')
  False
+ >>> PureWindowsPath('b.py').full_match('*.PY')
+ True
 
- The *pattern* may be another path object; this speeds up matching the same
- pattern against multiple files::
+ Set *case_sensitive* to ``True`` or ``False`` to override this behaviour.
 
- >>> pattern = PurePath('*.py')
- >>> PurePath('a/b.py').match(pattern)
- True
+ .. versionadded:: 3.13
 
- .. versionchanged:: 3.12
- Accepts an object implementing the :class:`os.PathLike` interface.
 
- As with other methods, case-sensitivity follows platform defaults::
+.. method:: PurePath.match(pattern, *, case_sensitive=None)
 
- >>> PurePosixPath('b.py').match('*.PY')
- False
- >>> PureWindowsPath('b.py').match('*.PY')
+ Match this path against the provided non-recursive glob-style pattern.
+ Return ``True`` if matching is successful, ``False`` otherwise.
+
+ This method is similar to :meth:`~PurePath.full_match`, but empty patterns
+ aren't allowed (:exc:`ValueError` is raised), the recursive wildcard
+ "``**``" isn't supported (it acts like non-recursive "``*``"), and if a
+ relative pattern is provided, then matching is done from the right::
+
+ >>> PurePath('a/b.py').match('*.py')
+ True
+ >>> PurePath('/a/b/c.py').match('b/*.py')
  True
+ >>> PurePath('/a/b/c.py').match('a/*.py')
+ False
 
- Set *case_sensitive* to ``True`` or ``False`` to override this behaviour.
+ .. versionchanged:: 3.12
+ The *pattern* parameter accepts a :term:`path-like object`.
 
  .. versionchanged:: 3.12
  The *case_sensitive* parameter was added.
 
- .. versionchanged:: 3.13
- Support for the recursive wildcard "``**``" was added. In previous
- versions, it acted like the non-recursive wildcard "``*``".
-
 
 .. method:: PurePath.relative_to(other, walk_up=False)
 

Doc/whatsnew/3.13.rst

@@ -336,7 +336,8 @@ pathlib
  object from a 'file' URI (``file:/``).
  (Contributed by Barney Gale in :gh:`107465`.)
 
-* Add support for recursive wildcards in :meth:`pathlib.PurePath.match`.
+* Add :meth:`pathlib.PurePath.full_match` for matching paths with
+ shell-style wildcards, including the recursive wildcard "``**``".
  (Contributed by Barney Gale in :gh:`73435`.)
 
 * Add *follow_symlinks* keyword-only argument to :meth:`pathlib.Path.glob`,

Lib/pathlib/__init__.py

@@ -490,6 +490,13 @@ def _pattern_stack(self):
  parts.reverse()
  return parts
 
+ @property
+ def _pattern_str(self):
+ """The path expressed as a string, for use in pattern-matching."""
+ # The string representation of an empty path is a single dot ('.'). Empty
+ # paths shouldn't match wildcards, so we change it to the empty string.
+ path_str = str(self)
+ return '' if path_str == '.' else path_str
 
 # Subclassing os.PathLike makes isinstance() checks slower,
 # which in turn makes Path construction slower. Register instead!

Lib/pathlib/_abc.py

@@ -47,19 +47,16 @@ def _is_case_sensitive(pathmod):
 re = glob = None
 
 
-@functools.lru_cache(maxsize=256)
-def _compile_pattern(pat, sep, case_sensitive):
+@functools.lru_cache(maxsize=512)
+def _compile_pattern(pat, sep, case_sensitive, recursive=True):
  """Compile given glob pattern to a re.Pattern object (observing case
  sensitivity)."""
  global re, glob
  if re is None:
  import re, glob
 
  flags = re.NOFLAG if case_sensitive else re.IGNORECASE
- regex = glob.translate(pat, recursive=True, include_hidden=True, seps=sep)
- # The string representation of an empty path is a single dot ('.'). Empty
- # paths shouldn't match wildcards, so we consume it with an atomic group.
- regex = r'(\.\Z)?+' + regex
+ regex = glob.translate(pat, recursive=recursive, include_hidden=True, seps=sep)
  return re.compile(regex, flags=flags).match
 
 
@@ -441,23 +438,48 @@ def _pattern_stack(self):
  raise NotImplementedError("Non-relative patterns are unsupported")
  return parts
 
+ @property
+ def _pattern_str(self):
+ """The path expressed as a string, for use in pattern-matching."""
+ return str(self)
+
  def match(self, path_pattern, *, case_sensitive=None):
  """
- Return True if this path matches the given pattern.
+ Return True if this path matches the given pattern. If the pattern is
+ relative, matching is done from the right; otherwise, the entire path
+ is matched. The recursive wildcard '**' is *not* supported by this
+ method.
  """
  if not isinstance(path_pattern, PurePathBase):
  path_pattern = self.with_segments(path_pattern)
  if case_sensitive is None:
  case_sensitive = _is_case_sensitive(self.pathmod)
  sep = path_pattern.pathmod.sep
- if path_pattern.anchor:
- pattern_str = str(path_pattern)
- elif path_pattern.parts:
- pattern_str = str('**' / path_pattern)
- else:
+ path_parts = self.parts[::-1]
+ pattern_parts = path_pattern.parts[::-1]
+ if not pattern_parts:
  raise ValueError("empty pattern")
- match = _compile_pattern(pattern_str, sep, case_sensitive)
- return match(str(self)) is not None
+ if len(path_parts) < len(pattern_parts):
+ return False
+ if len(path_parts) > len(pattern_parts) and path_pattern.anchor:
+ return False
+ for path_part, pattern_part in zip(path_parts, pattern_parts):
+ match = _compile_pattern(pattern_part, sep, case_sensitive, recursive=False)
+ if match(path_part) is None:
+ return False
+ return True
+
+ def full_match(self, pattern, *, case_sensitive=None):
+ """
+ Return True if this path matches the given glob-style pattern. The
+ pattern is matched against the entire path.
+ """
+ if not isinstance(pattern, PurePathBase):
+ pattern = self.with_segments(pattern)
+ if case_sensitive is None:
+ case_sensitive = _is_case_sensitive(self.pathmod)
+ match = _compile_pattern(pattern._pattern_str, pattern.pathmod.sep, case_sensitive)
+ return match(self._pattern_str) is not None
 
 
 
@@ -781,8 +803,8 @@ def glob(self, pattern, *, case_sensitive=None, follow_symlinks=None):
  if filter_paths:
  # Filter out paths that don't match pattern.
  prefix_len = len(str(self._make_child_relpath('_'))) - 1
- match = _compile_pattern(str(pattern), sep, case_sensitive)
- paths = (path for path in paths if match(str(path), prefix_len))
+ match = _compile_pattern(pattern._pattern_str, sep, case_sensitive)
+ paths = (path for path in paths if match(path._pattern_str, prefix_len))
  return paths
 
  def rglob(self, pattern, *, case_sensitive=None, follow_symlinks=None):

Lib/test/test_pathlib/test_pathlib_abc.py

@@ -249,39 +249,91 @@ def test_match_common(self):
  self.assertFalse(P('/ab.py').match('/a/*.py'))
  self.assertFalse(P('/a/b/c.py').match('/a/*.py'))
  # Multi-part glob-style pattern.
- self.assertTrue(P('a').match('**'))
- self.assertTrue(P('c.py').match('**'))
- self.assertTrue(P('a/b/c.py').match('**'))
- self.assertTrue(P('/a/b/c.py').match('**'))
- self.assertTrue(P('/a/b/c.py').match('/**'))
- self.assertTrue(P('/a/b/c.py').match('/a/**'))
- self.assertTrue(P('/a/b/c.py').match('**/*.py'))
- self.assertTrue(P('/a/b/c.py').match('/**/*.py'))
+ self.assertFalse(P('/a/b/c.py').match('/**/*.py'))
  self.assertTrue(P('/a/b/c.py').match('/a/**/*.py'))
- self.assertTrue(P('/a/b/c.py').match('/a/b/**/*.py'))
- self.assertTrue(P('/a/b/c.py').match('/**/**/**/**/*.py'))
- self.assertFalse(P('c.py').match('**/a.py'))
- self.assertFalse(P('c.py').match('c/**'))
- self.assertFalse(P('a/b/c.py').match('**/a'))
- self.assertFalse(P('a/b/c.py').match('**/a/b'))
- self.assertFalse(P('a/b/c.py').match('**/a/b/c'))
- self.assertFalse(P('a/b/c.py').match('**/a/b/c.'))
- self.assertFalse(P('a/b/c.py').match('**/a/b/c./**'))
- self.assertFalse(P('a/b/c.py').match('**/a/b/c./**'))
- self.assertFalse(P('a/b/c.py').match('/a/b/c.py/**'))
- self.assertFalse(P('a/b/c.py').match('/**/a/b/c.py'))
- self.assertRaises(ValueError, P('a').match, '**a/b/c')
- self.assertRaises(ValueError, P('a').match, 'a/b/c**')
  # Case-sensitive flag
  self.assertFalse(P('A.py').match('a.PY', case_sensitive=True))
  self.assertTrue(P('A.py').match('a.PY', case_sensitive=False))
  self.assertFalse(P('c:/a/B.Py').match('C:/A/*.pY', case_sensitive=True))
  self.assertTrue(P('/a/b/c.py').match('/A/*/*.Py', case_sensitive=False))
  # Matching against empty path
  self.assertFalse(P('').match('*'))
- self.assertTrue(P('').match('**'))
+ self.assertFalse(P('').match('**'))
  self.assertFalse(P('').match('**/*'))
 
+ def test_full_match_common(self):
+ P = self.cls
+ # Simple relative pattern.
+ self.assertTrue(P('b.py').full_match('b.py'))
+ self.assertFalse(P('a/b.py').full_match('b.py'))
+ self.assertFalse(P('/a/b.py').full_match('b.py'))
+ self.assertFalse(P('a.py').full_match('b.py'))
+ self.assertFalse(P('b/py').full_match('b.py'))
+ self.assertFalse(P('/a.py').full_match('b.py'))
+ self.assertFalse(P('b.py/c').full_match('b.py'))
+ # Wildcard relative pattern.
+ self.assertTrue(P('b.py').full_match('*.py'))
+ self.assertFalse(P('a/b.py').full_match('*.py'))
+ self.assertFalse(P('/a/b.py').full_match('*.py'))
+ self.assertFalse(P('b.pyc').full_match('*.py'))
+ self.assertFalse(P('b./py').full_match('*.py'))
+ self.assertFalse(P('b.py/c').full_match('*.py'))
+ # Multi-part relative pattern.
+ self.assertTrue(P('ab/c.py').full_match('a*/*.py'))
+ self.assertFalse(P('/d/ab/c.py').full_match('a*/*.py'))
+ self.assertFalse(P('a.py').full_match('a*/*.py'))
+ self.assertFalse(P('/dab/c.py').full_match('a*/*.py'))
+ self.assertFalse(P('ab/c.py/d').full_match('a*/*.py'))
+ # Absolute pattern.
+ self.assertTrue(P('/b.py').full_match('/*.py'))
+ self.assertFalse(P('b.py').full_match('/*.py'))
+ self.assertFalse(P('a/b.py').full_match('/*.py'))
+ self.assertFalse(P('/a/b.py').full_match('/*.py'))
+ # Multi-part absolute pattern.
+ self.assertTrue(P('/a/b.py').full_match('/a/*.py'))
+ self.assertFalse(P('/ab.py').full_match('/a/*.py'))
+ self.assertFalse(P('/a/b/c.py').full_match('/a/*.py'))
+ # Multi-part glob-style pattern.
+ self.assertTrue(P('a').full_match('**'))
+ self.assertTrue(P('c.py').full_match('**'))
+ self.assertTrue(P('a/b/c.py').full_match('**'))
+ self.assertTrue(P('/a/b/c.py').full_match('**'))
+ self.assertTrue(P('/a/b/c.py').full_match('/**'))
+ self.assertTrue(P('/a/b/c.py').full_match('/a/**'))
+ self.assertTrue(P('/a/b/c.py').full_match('**/*.py'))
+ self.assertTrue(P('/a/b/c.py').full_match('/**/*.py'))
+ self.assertTrue(P('/a/b/c.py').full_match('/a/**/*.py'))
+ self.assertTrue(P('/a/b/c.py').full_match('/a/b/**/*.py'))
+ self.assertTrue(P('/a/b/c.py').full_match('/**/**/**/**/*.py'))
+ self.assertFalse(P('c.py').full_match('**/a.py'))
+ self.assertFalse(P('c.py').full_match('c/**'))
+ self.assertFalse(P('a/b/c.py').full_match('**/a'))
+ self.assertFalse(P('a/b/c.py').full_match('**/a/b'))
+ self.assertFalse(P('a/b/c.py').full_match('**/a/b/c'))
+ self.assertFalse(P('a/b/c.py').full_match('**/a/b/c.'))
+ self.assertFalse(P('a/b/c.py').full_match('**/a/b/c./**'))
+ self.assertFalse(P('a/b/c.py').full_match('**/a/b/c./**'))
+ self.assertFalse(P('a/b/c.py').full_match('/a/b/c.py/**'))
+ self.assertFalse(P('a/b/c.py').full_match('/**/a/b/c.py'))
+ self.assertRaises(ValueError, P('a').full_match, '**a/b/c')
+ self.assertRaises(ValueError, P('a').full_match, 'a/b/c**')
+ # Case-sensitive flag
+ self.assertFalse(P('A.py').full_match('a.PY', case_sensitive=True))
+ self.assertTrue(P('A.py').full_match('a.PY', case_sensitive=False))
+ self.assertFalse(P('c:/a/B.Py').full_match('C:/A/*.pY', case_sensitive=True))
+ self.assertTrue(P('/a/b/c.py').full_match('/A/*/*.Py', case_sensitive=False))
+ # Matching against empty path
+ self.assertFalse(P('').full_match('*'))
+ self.assertTrue(P('').full_match('**'))
+ self.assertFalse(P('').full_match('**/*'))
+ # Matching with empty pattern
+ self.assertTrue(P('').full_match(''))
+ self.assertTrue(P('.').full_match('.'))
+ self.assertFalse(P('/').full_match(''))
+ self.assertFalse(P('/').full_match('.'))
+ self.assertFalse(P('foo').full_match(''))
+ self.assertFalse(P('foo').full_match('.'))
+
  def test_parts_common(self):
  # `parts` returns a tuple.
  sep = self.sep