Issue 9396. Apply functools.lru_cache in the place of the

random flushing cache in the re module.

diff --git a/Lib/re.py b/Lib/re.py
index 2f1a76eaa521fe5139f4d7930dd66ac7fe40440d..269eaef8554b02dd52b4b4db1878558d66481e6a 100644 (file)
--- a/Lib/re.py
+++ b/Lib/re.py
@@ -118,6 +118,7 @@ This module also defines an exception 'error'.
 import sys
 import sre_compile
 import sre_parse
+import functools
 
 # public symbols
 __all__ = [ "match", "search", "sub", "subn", "split", "findall",
@@ -205,9 +206,9 @@ def compile(pattern, flags=0):
     return _compile(pattern, flags)
 
 def purge():
-    "Clear the regular expression cache"
-    _cache.clear()
-    _cache_repl.clear()
+    "Clear the regular expression caches"
+    _compile_typed.clear()
+    _compile_repl.clear()
 
 def template(pattern, flags=0):
     "Compile a template pattern, returning a pattern object"
@@ -289,12 +290,12 @@ def _shrink_cache(cache_dict, max_length, divisor=5):
             # Ignore problems if the cache changed from another thread.
             pass
 
-def _compile(*key):
+def _compile(*args):
+    return _compile_typed(type(args[0]), *args)
+
+@functools.lru_cache(maxsize=_MAXCACHE)
+def _compile_typed(type, *key):
     # internal: compile pattern
-    cachekey = (type(key[0]),) + key
-    p = _cache.get(cachekey)
-    if p is not None:
-        return p
     pattern, flags = key
     if isinstance(pattern, _pattern_type):
         if flags:
@@ -303,23 +304,14 @@ def _compile(*key):
         return pattern
     if not sre_compile.isstring(pattern):
         raise TypeError("first argument must be string or compiled pattern")
-    p = sre_compile.compile(pattern, flags)
-    if len(_cache) >= _MAXCACHE:
-        _shrink_cache(_cache, _MAXCACHE)
-    _cache[cachekey] = p
+    return sre_compile.compile(pattern, flags)
     return p
 
+@functools.lru_cache(maxsize=_MAXCACHE)
 def _compile_repl(*key):
     # internal: compile replacement pattern
-    p = _cache_repl.get(key)
-    if p is not None:
-        return p
     repl, pattern = key
-    p = sre_parse.parse_template(repl, pattern)
-    if len(_cache_repl) >= _MAXCACHE:
-        _shrink_cache(_cache_repl, _MAXCACHE)
-    _cache_repl[key] = p
-    return p
+    return sre_parse.parse_template(repl, pattern)
 
 def _expand(pattern, match, template):
     # internal: match.expand implementation hook

diff --git a/Lib/test/test_re.py b/Lib/test/test_re.py
index 6b11685e9c8804e954417bde5c49cf52a12686b7..96a83b88c1813836f590639a3b2e2b021e5e549b 100644 (file)
--- a/Lib/test/test_re.py
+++ b/Lib/test/test_re.py
@@ -875,70 +875,8 @@ def run_re_tests():
                     print('=== Fails on unicode-sensitive match', t)
 
 
-class ReCacheTests(unittest.TestCase):
-    """These tests are specific to the re._shrink_cache implementation."""
-
-    def setUp(self):
-        self._orig_maxcache = re._MAXCACHE
-
-    def tearDown(self):
-        re._MAXCACHE = self._orig_maxcache
-
-    def test_compile_cache_overflow(self):
-        # NOTE: If a profiler or debugger is tracing code and compiling
-        # regular expressions while tracing through this test... expect
-        # the test to fail.  This test is not concurrency safe.
-
-        # Explicitly fill the caches.
-        re._MAXCACHE = 20
-        max_cache = re._MAXCACHE
-        unique_chars = tuple(chr(char_num) for char_num in
-                             range(b'a'[0], b'a'[0]+max_cache))
-        re._cache.clear()
-        for char in unique_chars:
-            re._compile(char, 0)
-        self.assertEqual(max_cache, len(re._cache))
-        re._cache_repl.clear()
-        for char in unique_chars:
-            re._compile_repl(char*2, char)
-        self.assertEqual(max_cache, len(re._cache_repl))
-
-        # Overflow both caches and make sure they have extra room left
-        # afterwards as well as having more than a single entry.
-        re._compile('A', 0)
-        self.assertLess(len(re._cache), max_cache)
-        self.assertGreater(len(re._cache), 1)
-        re._compile_repl('A', 'A')
-        self.assertLess(len(re._cache_repl), max_cache)
-        self.assertGreater(len(re._cache_repl), 1)
-
-    def test_shrink_cache_at_limit(self):
-        cache = dict(zip(range(6), range(6)))
-        re._shrink_cache(cache, 6, divisor=3)
-        self.assertEqual(4, len(cache))
-
-    def test_shrink_cache_empty(self):
-        cache = {}
-        re._shrink_cache(cache, 6, divisor=3)
-        # Cache was empty, make sure we didn't raise an exception.
-        self.assertEqual(0, len(cache))
-
-    def test_shrink_cache_overflowing(self):
-        cache = dict(zip(range(6), range(6)))
-        re._shrink_cache(cache, 4, divisor=2)
-        # Cache was larger than the maximum, be sure we shrunk to smaller.
-        self.assertEqual(2, len(cache))
-
-    def test_shrink_cache_underflow(self):
-        cache = dict(zip(range(6), range(6)))
-        # No shrinking to do.
-        re._shrink_cache(cache, 9, divisor=3)
-        self.assertEqual(6, len(cache))
-
-
 def test_main():
     run_unittest(ReTests)
-    run_unittest(ReCacheTests)
     run_re_tests()
 
 if __name__ == "__main__":

diff --git a/Misc/NEWS b/Misc/NEWS
index 00188d4f946208a6e8cf8138d1f9be1d43bd3e1c..57bcf8ea5f3b3d58adf65c26dd7e3c1cc532e0de 100644 (file)
--- a/Misc/NEWS
+++ b/Misc/NEWS
@@ -697,8 +697,8 @@ Library
 
 - The default size of the re module's compiled regular expression cache has been
   increased from 100 to 500 and the cache replacement policy has changed from
-  simply clearing the entire cache on overflow to randomly forgetting 20% of the
-  existing cached compiled regular expressions.  This is a performance win for
+  simply clearing the entire cache on overflow to forgetting the least recently
+  used cached compiled regular expressions.  This is a performance win for
   applications that use a lot of regular expressions and limits the impact of
   the performance hit anytime the cache is exceeded.