GH-73435: Add pathlib.PurePath.globmatch()

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 `globmatch()` method that: - Supports the recursive wildcard `**` - Matches the *entire* path when given a relative pattern As a result, `globmatch()`'s pattern language exactly matches that of `glob()`.
barneygale · Jan 26, 2024 · Jan 20, 2024 · Jan 20, 2024 · Jan 20, 2024 · Jan 22, 2024
commit 1c8bf813007e63930f0f89a980f3b848b1590372
diff --git a/Doc/library/glob.rst b/Doc/library/glob.rst
@@ -147,7 +147,7 @@ The :mod:`glob` module defines the following functions:
 
    .. seealso::
 
-     :meth:`pathlib.PurePath.match` and :meth:`pathlib.Path.glob` methods,
+     :meth:`pathlib.PurePath.globmatch` and :meth:`pathlib.Path.glob` methods,
      which call this function to implement pattern matching and globbing.
 
    .. versionadded:: 3.13

@@ -552,55 +552,54 @@ Pure paths provide the following methods and properties:
       PureWindowsPath('c:/Program Files')
 
 
-.. method:: PurePath.match(pattern, *, case_sensitive=None)
+.. method:: PurePath.globmatch(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').globmatch('a/*.py')
       True
-      >>> PurePath('/a/b/c.py').match('a/*.py')
+      >>> PurePath('a/b.py').globmatch('*.py')
       False
+      >>> PurePath('/a/b/c.py').globmatch('/a/**')
+      True
+      >>> PurePath('/a/b/c.py').globmatch('**/*.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').globmatch('*.PY')
       False
+      >>> PureWindowsPath('b.py').globmatch('*.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.globmatch`, but the recursive
+   wildcard "``**``" is not supported (it acts like non-recursive "``*``"),
+   and if a relative pattern is given, 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)
 

@@ -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.globmatch` 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`,

@@ -48,15 +48,15 @@ def _is_case_sensitive(pathmod):
 
 
 @functools.lru_cache(maxsize=256)
-def _compile_pattern(pat, sep, case_sensitive):
+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)
+    regex = glob.translate(pat, recursive=recursive, 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
@@ -450,13 +450,28 @@ def match(self, path_pattern, *, case_sensitive=None):
         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:
+        delta = len(self.parts) - len(path_pattern.parts)
+        if delta < 0:
+            return False  # Path is too short.
+        if delta > 0 and path_pattern.anchor:
+            return False  # Path is too long.
+        if not path_pattern.parts:
             raise ValueError("empty pattern")
-        match = _compile_pattern(pattern_str, sep, case_sensitive)
+        for path, pattern in zip(reversed(self.parts), reversed(path_pattern.parts)):
+            match = _compile_pattern(pattern, sep, case_sensitive, recursive=False)
+            if match(path) is None:
+                return False
+        return True
+
+    def globmatch(self, pattern, *, case_sensitive=None):
+        """
+        Return True if this path matches the given glob-style pattern.
+        """
+        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(str(pattern), pattern.pathmod.sep, case_sensitive)
         return match(str(self)) is not None
 
 

@@ -249,39 +249,84 @@ 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_globmatch_common(self):
+        P = self.cls
+        # Simple relative pattern.
+        self.assertTrue(P('b.py').globmatch('b.py'))
+        self.assertFalse(P('a/b.py').globmatch('b.py'))
+        self.assertFalse(P('/a/b.py').globmatch('b.py'))
+        self.assertFalse(P('a.py').globmatch('b.py'))
+        self.assertFalse(P('b/py').globmatch('b.py'))
+        self.assertFalse(P('/a.py').globmatch('b.py'))
+        self.assertFalse(P('b.py/c').globmatch('b.py'))
+        # Wildcard relative pattern.
+        self.assertTrue(P('b.py').globmatch('*.py'))
+        self.assertFalse(P('a/b.py').globmatch('*.py'))
+        self.assertFalse(P('/a/b.py').globmatch('*.py'))
+        self.assertFalse(P('b.pyc').globmatch('*.py'))
+        self.assertFalse(P('b./py').globmatch('*.py'))
+        self.assertFalse(P('b.py/c').globmatch('*.py'))
+        # Multi-part relative pattern.
+        self.assertTrue(P('ab/c.py').globmatch('a*/*.py'))
+        self.assertFalse(P('/d/ab/c.py').globmatch('a*/*.py'))
+        self.assertFalse(P('a.py').globmatch('a*/*.py'))
+        self.assertFalse(P('/dab/c.py').globmatch('a*/*.py'))
+        self.assertFalse(P('ab/c.py/d').globmatch('a*/*.py'))
+        # Absolute pattern.
+        self.assertTrue(P('/b.py').globmatch('/*.py'))
+        self.assertFalse(P('b.py').globmatch('/*.py'))
+        self.assertFalse(P('a/b.py').globmatch('/*.py'))
+        self.assertFalse(P('/a/b.py').globmatch('/*.py'))
+        # Multi-part absolute pattern.
+        self.assertTrue(P('/a/b.py').globmatch('/a/*.py'))
+        self.assertFalse(P('/ab.py').globmatch('/a/*.py'))
+        self.assertFalse(P('/a/b/c.py').globmatch('/a/*.py'))
+        # Multi-part glob-style pattern.
+        self.assertTrue(P('a').globmatch('**'))
+        self.assertTrue(P('c.py').globmatch('**'))
+        self.assertTrue(P('a/b/c.py').globmatch('**'))
+        self.assertTrue(P('/a/b/c.py').globmatch('**'))
+        self.assertTrue(P('/a/b/c.py').globmatch('/**'))
+        self.assertTrue(P('/a/b/c.py').globmatch('/a/**'))
+        self.assertTrue(P('/a/b/c.py').globmatch('**/*.py'))
+        self.assertTrue(P('/a/b/c.py').globmatch('/**/*.py'))
+        self.assertTrue(P('/a/b/c.py').globmatch('/a/**/*.py'))
+        self.assertTrue(P('/a/b/c.py').globmatch('/a/b/**/*.py'))
+        self.assertTrue(P('/a/b/c.py').globmatch('/**/**/**/**/*.py'))
+        self.assertFalse(P('c.py').globmatch('**/a.py'))
+        self.assertFalse(P('c.py').globmatch('c/**'))
+        self.assertFalse(P('a/b/c.py').globmatch('**/a'))
+        self.assertFalse(P('a/b/c.py').globmatch('**/a/b'))
+        self.assertFalse(P('a/b/c.py').globmatch('**/a/b/c'))
+        self.assertFalse(P('a/b/c.py').globmatch('**/a/b/c.'))
+        self.assertFalse(P('a/b/c.py').globmatch('**/a/b/c./**'))
+        self.assertFalse(P('a/b/c.py').globmatch('**/a/b/c./**'))
+        self.assertFalse(P('a/b/c.py').globmatch('/a/b/c.py/**'))
+        self.assertFalse(P('a/b/c.py').globmatch('/**/a/b/c.py'))
+        self.assertRaises(ValueError, P('a').globmatch, '**a/b/c')
+        self.assertRaises(ValueError, P('a').globmatch, 'a/b/c**')
+        # Case-sensitive flag
+        self.assertFalse(P('A.py').globmatch('a.PY', case_sensitive=True))
+        self.assertTrue(P('A.py').globmatch('a.PY', case_sensitive=False))
+        self.assertFalse(P('c:/a/B.Py').globmatch('C:/A/*.pY', case_sensitive=True))
+        self.assertTrue(P('/a/b/c.py').globmatch('/A/*/*.Py', case_sensitive=False))
+        # Matching against empty path
+        self.assertFalse(P('').globmatch('*'))
+        self.assertTrue(P('').globmatch('**'))
+        self.assertFalse(P('').globmatch('**/*'))
+
     def test_parts_common(self):
         # `parts` returns a tuple.
         sep = self.sep