`Lib/glob.py`

cpython 3.14 @ ab2d84fe1023/Lib/glob.py

Lib/glob.py implements glob() and iglob(), which expand Unix shell wildcard patterns against the filesystem. Patterns can contain * (any filename segment), ** (recursive directory walk), ? (single character), and [...] (character class). The implementation uses os.scandir for directory listing and fnmatch.filter for pattern matching.

Map

Lines	Symbol	Role
1-30	imports, `__all__`	Module setup
31-80	`glob`, `iglob`	Public API; `iglob` is a generator, `glob` materializes it
81-140	`_iglob`	Core generator; splits pattern on `os.sep` and recurses
141-200	`_iterdir`, `_iterdir_dirs`	`os.scandir`-based directory listing with `DirEntry` caching
201-250	`_glob1`, `_glob0`	Single-segment match with `fnmatch.filter`
251-300	`_rlistdir`	Recursive `**` walk yielding all descendant directories

Reading

Pattern splitting and recursion

_iglob splits the pattern on os.sep into a head (directory prefix) and tail (remaining pattern). It handles the three cases: no magic characters (literal path test), ** (recursive walk via _rlistdir), and single-segment with wildcards (_glob1).

# CPython: Lib/glob.py:90 _iglob
def _iglob(pathname, recursive, dironly):
    dirname, basename = os.path.split(pathname)
    if not has_magic(pathname):
        ...
        yield pathname
        return
    if _isrecursive(basename):
        yield from _rlistdir(dirname, dironly)
        return
    dirs = _iglob(dirname, recursive, True) if has_magic(dirname) else [dirname]
    for dirname in dirs:
        for name in _glob1(dirname, basename, dironly):
            yield os.path.join(dirname, name)

`_glob1`: fnmatch filtering

_glob1 lists the directory with os.scandir (filtering to dirs-only when dironly is True) and applies fnmatch.filter to the resulting names.

# CPython: Lib/glob.py:170 _glob1
def _glob1(dirname, pattern, dironly):
    if not dirname:
        dirname = os.curdir
    try:
        names = _iterdir(dirname, dironly)
    except OSError:
        return
    if not _ishidden(pattern):
        names = (x for x in names if not _ishidden(x))
    yield from fnmatch.filter(names, pattern)

`_rlistdir`: recursive `**` expansion

_rlistdir is a depth-first recursive generator that yields the starting directory and all descendants. It is used when ** appears in the pattern.

# CPython: Lib/glob.py:255 _rlistdir
def _rlistdir(dirname, dironly):
    yield dirname
    for x in _iterdir_dirs(dirname):
        path = os.path.join(dirname, x) if dirname else x
        yield from _rlistdir(path, dironly)

gopy notes

Not yet ported. The Go standard library's filepath.Glob provides basic * and ? support. A full port at module/glob/ would need to implement ** recursion and the [...] character class via filepath.Match or a custom walker.

CPython 3.14 changes

3.14 refactored glob into a Glob class (PEP 428 / pathlib unification) that backs both glob.glob() and pathlib.Path.glob(). The old functional API is preserved as a compatibility shim. iglob now accepts a root_dir keyword argument that sets the starting directory without changing the pattern.

Map​

Reading​

Pattern splitting and recursion​

_glob1: fnmatch filtering​

_rlistdir: recursive ** expansion​

gopy notes​

CPython 3.14 changes​

Map