Lib/re/__init__.py

Source:

cpython 3.14 @ ab2d84fe1023/Lib/re/__init__.py ↗

Lib/re/__init__.py is the public API for Python's regular expressions. It delegates all actual regex operations to the _sre C extension (via the re._compiler and re._parser modules) and adds a pattern cache to avoid recompiling the same regex string repeatedly.

Map

Lines	Symbol	Role
1-50	imports, __all__	Flag constants; public name list
51-150	compile, _compile	Pattern compilation; LRU cache
151-250	match, search, fullmatch, findall, finditer	Convenience functions
251-330	sub, subn, split	Substitution and splitting
331-400	escape, purge, error	Utility functions

Reading

Pattern cache

_compile(pattern, flags) checks a module-level _cache dict keyed on (pattern, flags, type(pattern)). If the pattern is already compiled, the cached Pattern object is returned directly. The cache is capped at _MAXCACHE (512) entries; purge() clears it.

# CPython: Lib/re/__init__.py:51 _compile
def _compile(pattern, flags):
    if isinstance(flags, RegexFlag):
        flags = flags.value
    try:
        return _cache[type(pattern), pattern, flags]
    except KeyError:
        pass
    p = _compiler.compile(pattern, flags)
    _cache[type(pattern), pattern, flags] = p
    return p

Convenience functions

Each top-level function like re.match(pattern, string) calls _compile(pattern, 0).match(string). The pattern compilation is the expensive step; since the result is cached, repeated calls with the same pattern are nearly free.

# CPython: Lib/re/__init__.py:151 match
def match(pattern, string, flags=0):
    return _compile(pattern, flags).match(string)

Flags

re.IGNORECASE (alias re.I), re.MULTILINE (re.M), re.DOTALL (re.S), re.VERBOSE (re.X) etc. are RegexFlag enum members. They can be combined with |. The re.UNICODE flag is the default in Python 3; re.ASCII restricts \w, \d, \s to ASCII ranges.

gopy notes

Not yet ported. The planned package path is module/re/. Go's regexp package uses RE2 semantics (no backreferences, no lookaheads) while CPython's _sre uses a backtracking NFA. A faithful port would require either embedding a Python-compatible regex engine or documenting the semantic differences.