gh-92550 - Fix regression in pathlib.Path.rglob() (GH-92583)

miss-islington · barneygale · web-flow · commit a51baec9ce0e · 2022-05-09T17:39:43.000-07:00
We could try to remedy this by taking a slice, but we then run into an issue where the empty string will match altsep on POSIX. That rabbit hole could keep getting deeper. A proper fix for the original issue involves making pathlib's path normalisation more configurable - in this case we want to retain trailing slashes, but in other we might want to preserve `./` prefixes, or elide `../` segments when we're sure we won't encounter symlinks. This reverts commit ea2f5bc. (cherry picked from commit dcdf250) Co-authored-by: Barney Gale <barney.gale@gmail.com>
diff --git a/Doc/library/pathlib.rst b/Doc/library/pathlib.rst
@@ -815,9 +815,6 @@ call fails (for example because the path doesn't exist).
 
    .. audit-event:: pathlib.Path.glob self,pattern pathlib.Path.glob
 
-   .. versionchanged:: 3.11
-      Return only directories if *pattern* ends with a pathname components
-      separator (:data:`~os.sep` or :data:`~os.altsep`).
 
 .. method:: Path.group()
 
@@ -1107,9 +1104,6 @@ call fails (for example because the path doesn't exist).
 
    .. audit-event:: pathlib.Path.rglob self,pattern pathlib.Path.rglob
 
-   .. versionchanged:: 3.11
-      Return only directories if *pattern* ends with a pathname components
-      separator (:data:`~os.sep` or :data:`~os.altsep`).
 
 .. method:: Path.rmdir()
 
diff --git a/Doc/whatsnew/3.11.rst b/Doc/whatsnew/3.11.rst
@@ -557,15 +557,6 @@ os
   instead of ``CryptGenRandom()`` which is deprecated.
   (Contributed by Dong-hee Na in :issue:`44611`.)
 
-
-pathlib
--------
-
-* :meth:`~pathlib.Path.glob` and :meth:`~pathlib.Path.rglob` return only
-  directories if *pattern* ends with a pathname components separator:
-  :data:`~os.sep` or :data:`~os.altsep`.
-  (Contributed by Eisuke Kawasima in :issue:`22276` and :issue:`33392`.)
-
 re
 --
 
diff --git a/Lib/pathlib.py b/Lib/pathlib.py
@@ -281,8 +281,6 @@ def make_uri(self, path):
 def _make_selector(pattern_parts, flavour):
     pat = pattern_parts[0]
     child_parts = pattern_parts[1:]
-    if not pat:
-        return _TerminatingSelector()
     if pat == '**':
         cls = _RecursiveWildcardSelector
     elif '**' in pat:
@@ -945,8 +943,6 @@ def glob(self, pattern):
         drv, root, pattern_parts = self._flavour.parse_parts((pattern,))
         if drv or root:
             raise NotImplementedError("Non-relative patterns are unsupported")
-        if pattern[-1] in (self._flavour.sep, self._flavour.altsep):
-            pattern_parts.append('')
         selector = _make_selector(tuple(pattern_parts), self._flavour)
         for p in selector.select_from(self):
             yield p
@@ -960,8 +956,6 @@ def rglob(self, pattern):
         drv, root, pattern_parts = self._flavour.parse_parts((pattern,))
         if drv or root:
             raise NotImplementedError("Non-relative patterns are unsupported")
-        if pattern[-1] in (self._flavour.sep, self._flavour.altsep):
-            pattern_parts.append('')
         selector = _make_selector(("**",) + tuple(pattern_parts), self._flavour)
         for p in selector.select_from(self):
             yield p
diff --git a/Lib/test/test_pathlib.py b/Lib/test/test_pathlib.py
@@ -1662,11 +1662,6 @@ def _check(glob, expected):
         else:
             _check(p.glob("*/fileB"), ['dirB/fileB', 'linkB/fileB'])
 
-        if not os_helper.can_symlink():
-            _check(p.glob("*/"), ["dirA", "dirB", "dirC", "dirE"])
-        else:
-            _check(p.glob("*/"), ["dirA", "dirB", "dirC", "dirE", "linkB"])
-
     def test_rglob_common(self):
         def _check(glob, expected):
             self.assertEqual(set(glob), { P(BASE, q) for q in expected })
@@ -1684,16 +1679,6 @@ def _check(glob, expected):
                                         "linkB/fileB", "dirA/linkC/fileB"])
         _check(p.rglob("file*"), ["fileA", "dirB/fileB",
                                   "dirC/fileC", "dirC/dirD/fileD"])
-        if not os_helper.can_symlink():
-            _check(p.rglob("*/"), [
-                "dirA", "dirB", "dirC", "dirC/dirD", "dirE",
-            ])
-        else:
-            _check(p.rglob("*/"), [
-                "dirA", "dirA/linkC", "dirB", "dirB/linkD", "dirC",
-                "dirC/dirD", "dirE", "linkB",
-            ])
-
         p = P(BASE, "dirC")
         _check(p.rglob("file*"), ["dirC/fileC", "dirC/dirD/fileD"])
         _check(p.rglob("*/*"), ["dirC/dirD/fileD"])
@@ -2719,7 +2704,6 @@ def test_glob(self):
         P = self.cls
         p = P(BASE)
         self.assertEqual(set(p.glob("FILEa")), { P(BASE, "fileA") })
-        self.assertEqual(set(p.glob("*a\\")), { P(BASE, "dirA") })
         self.assertEqual(set(p.glob("F*a")), { P(BASE, "fileA") })
         self.assertEqual(set(map(str, p.glob("FILEa"))), {f"{p}\\FILEa"})
         self.assertEqual(set(map(str, p.glob("F*a"))), {f"{p}\\fileA"})
@@ -2728,7 +2712,6 @@ def test_rglob(self):
         P = self.cls
         p = P(BASE, "dirC")
         self.assertEqual(set(p.rglob("FILEd")), { P(BASE, "dirC/dirD/fileD") })
-        self.assertEqual(set(p.rglob("*\\")), { P(BASE, "dirC/dirD") })
         self.assertEqual(set(map(str, p.rglob("FILEd"))), {f"{p}\\dirD\\FILEd"})
 
     def test_expanduser(self):
diff --git a/Misc/NEWS.d/next/Library/2022-05-09-23-36-19.gh-issue-92550.qZ4AhU.rst b/Misc/NEWS.d/next/Library/2022-05-09-23-36-19.gh-issue-92550.qZ4AhU.rst
@@ -0,0 +1,2 @@
+:meth:`pathlib.Path.rglob` raised :exc:`IndexError` when called with an
+empty string. This regression was introduced in 3.11b1.

Original file line number	Diff line number	Diff line change
`@@ -0,0 +1,2 @@`
	`1`	+:meth:`pathlib.Path.rglob` raised :exc:`IndexError` when called with an
	`2`	`+empty string. This regression was introduced in 3.11b1.`