languages/cpython
languages/cpython
1
0
Fork 0
mirror of https://github.com/python/cpython synced 2024-07-20 09:45:46 +00:00
Code Issues Actions 8 Packages Projects Releases Wiki Activity

GH-115060: Speed up pathlib.Path.glob() by removing redundant regex matching (#115061)

Browse source 
When expanding and filtering paths for a `**` wildcard segment, build an `re.Pattern` object from the subsequent pattern parts, rather than the entire pattern, and match against the `os.DirEntry` object prior to instantiating a path object. Also skip compiling a pattern when expanding a `*` wildcard segment.
This commit is contained in:
Barney Gale 2024-02-10 18:12:34 +00:00 committed by GitHub
parent 9d1a353230
commit 6f93b4df92
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 76 additions and 28 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
Lib/pathlib/__init__.py
View file

@ -587,9 +587,13 @@ def iterdir(self):
def _scandir(self):
return os.scandir(self)
def _make_child_entry(self, entry):
def _direntry_str(self, entry):
# Transform an entry yielded from _scandir() into a path string.
return entry.name if str(self) == '.' else entry.path
def _make_child_direntry(self, entry):
# Transform an entry yielded from _scandir() into a path object.
path_str = entry.name if str(self) == '.' else entry.path
path_str = self._direntry_str(entry)
path = self.with_segments(path_str)
path._str = path_str
path._drv = self.drive

82
Lib/pathlib/_abc.py
View file

@ -86,19 +86,29 @@ def _select_children(parent_paths, dir_only, follow_symlinks, match):
continue
except OSError:
continue
if match(entry.name):
yield parent_path._make_child_entry(entry)
# Avoid cost of making a path object for non-matching paths by
# matching against the os.DirEntry.name string.
if match is None or match(entry.name):
yield parent_path._make_child_direntry(entry)
def _select_recursive(parent_paths, dir_only, follow_symlinks):
"""Yield given paths and all their subdirectories, recursively."""
def _select_recursive(parent_paths, dir_only, follow_symlinks, match):
"""Yield given paths and all their children, recursively, filtering by
string and type.
"""
if follow_symlinks is None:
follow_symlinks = False
for parent_path in parent_paths:
if match is not None:
# If we're filtering paths through a regex, record the length of
# the parent path. We'll pass it to match(path, pos=...) later.
parent_len = len(str(parent_path._make_child_relpath('_'))) - 1
paths = [parent_path._make_child_relpath('')]
while paths:
path = paths.pop()
yield path
if match is None or match(str(path), parent_len):
# Yield *directory* path that matches pattern (if any).
yield path
try:
# We must close the scandir() object before proceeding to
# avoid exhausting file descriptors when globbing deep trees.
@ -108,14 +118,22 @@ def _select_recursive(parent_paths, dir_only, follow_symlinks):
pass
else:
for entry in entries:
# Handle directory entry.
try:
if entry.is_dir(follow_symlinks=follow_symlinks):
paths.append(path._make_child_entry(entry))
# Recurse into this directory.
paths.append(path._make_child_direntry(entry))
continue
except OSError:
pass
# Handle file entry.
if not dir_only:
yield path._make_child_entry(entry)
# Avoid cost of making a path object for non-matching
# files by matching against the os.DirEntry object.
if match is None or match(path._direntry_str(entry), parent_len):
# Yield *file* path that matches pattern (if any).
yield path._make_child_direntry(entry)
def _select_unique(paths):
@ -750,8 +768,14 @@ def _scandir(self):
from contextlib import nullcontext
return nullcontext(self.iterdir())
def _make_child_entry(self, entry):
def _direntry_str(self, entry):
# Transform an entry yielded from _scandir() into a path string.
# PathBase._scandir() yields PathBase objects, so use str().
return str(entry)
def _make_child_direntry(self, entry):
# Transform an entry yielded from _scandir() into a path object.
# PathBase._scandir() yields PathBase objects, so this is a no-op.
return entry
def _make_child_relpath(self, name):
@ -769,43 +793,49 @@ def glob(self, pattern, *, case_sensitive=None, follow_symlinks=None):
stack = pattern._pattern_stack
specials = ('', '.', '..')
filter_paths = False
deduplicate_paths = False
sep = self.pathmod.sep
paths = iter([self] if self.is_dir() else [])
while stack:
part = stack.pop()
if part in specials:
# Join special component (e.g. '..') onto paths.
paths = _select_special(paths, part)
elif part == '**':
# Consume adjacent '**' components.
# Consume following '**' components, which have no effect.
while stack and stack[-1] == '**':
stack.pop()
# Consume adjacent non-special components and enable post-walk
# regex filtering, provided we're treating symlinks consistently.
# Consume following non-special components, provided we're
# treating symlinks consistently. Each component is joined
# onto 'part', which is used to generate an re.Pattern object.
if follow_symlinks is not None:
while stack and stack[-1] not in specials:
filter_paths = True
stack.pop()
part += sep + stack.pop()
dir_only = bool(stack)
paths = _select_recursive(paths, dir_only, follow_symlinks)
# If the previous loop consumed pattern components, compile an
# re.Pattern object based on those components.
match = _compile_pattern(part, sep, case_sensitive) if part != '**' else None
# Recursively walk directories, filtering by type and regex.
paths = _select_recursive(paths, bool(stack), follow_symlinks, match)
# De-duplicate if we've already seen a '**' component.
if deduplicate_paths:
# De-duplicate if we've already seen a '**' component.
paths = _select_unique(paths)
deduplicate_paths = True
elif '**' in part:
raise ValueError("Invalid pattern: '**' can only be an entire path component")
else:
dir_only = bool(stack)
match = _compile_pattern(part, sep, case_sensitive)
paths = _select_children(paths, dir_only, follow_symlinks, match)
if filter_paths:
# Filter out paths that don't match pattern.
prefix_len = len(str(self._make_child_relpath('_'))) - 1
match = _compile_pattern(pattern._pattern_str, sep, case_sensitive)
paths = (path for path in paths if match(path._pattern_str, prefix_len))
# If the pattern component isn't '*', compile an re.Pattern
# object based on the component.
match = _compile_pattern(part, sep, case_sensitive) if part != '*' else None
# Iterate over directories' children filtering by type and regex.
paths = _select_children(paths, bool(stack), follow_symlinks, match)
return paths
def rglob(self, pattern, *, case_sensitive=None, follow_symlinks=None):
@ -854,7 +884,7 @@ def walk(self, top_down=True, on_error=None, follow_symlinks=False):
if is_dir:
if not top_down:
paths.append(path._make_child_entry(entry))
paths.append(path._make_child_direntry(entry))
dirnames.append(entry.name)
else:
filenames.append(entry.name)

13
Lib/test/test_pathlib/test_pathlib.py
View file

@ -1250,6 +1250,19 @@ def test_glob_pathlike(self):
self.assertEqual(expect, set(p.glob(P(pattern))))
self.assertEqual(expect, set(p.glob(FakePath(pattern))))
@needs_symlinks
def test_glob_dot(self):
P = self.cls
with os_helper.change_cwd(P(self.base, "dirC")):
self.assertEqual(
set(P('.').glob('*')), {P("fileC"), P("novel.txt"), P("dirD")})
self.assertEqual(
set(P('.').glob('**')), {P("fileC"), P("novel.txt"), P("dirD"), P("dirD/fileD"), P(".")})
self.assertEqual(
set(P('.').glob('**/*')), {P("fileC"), P("novel.txt"), P("dirD"), P("dirD/fileD")})
self.assertEqual(
set(P('.').glob('**/*/*')), {P("dirD/fileD")})
def test_rglob_pathlike(self):
P = self.cls
p = P(self.base, "dirC")

1
Misc/NEWS.d/next/Library/2024-02-06-03-55-46.gh-issue-115060.EkWRpP.rst Normal file
View file

@ -0,0 +1 @@
Speed up :meth:`pathlib.Path.glob` by removing redundant regex matching.