gh-118761: Reduce import time of gettext.py by delaying re import

eli-schwartz · eli-schwartz · commit fd471b8785e8 · 2025-01-19T00:01:23.000-05:00
gettext is often imported in programs that may not end up translating
anything. In fact, the `struct` module already has a delayed import when
parsing GNUTranslations to speed up the no .mo files case. The re module
is also used in the same situation, but behind a function chain only
called by GNUTranslations.

cache the compiled regex globally the first time it is used. The
finditer function can be converted to a method call on the compiled
object (it always could) which is slightly more efficient and necessary
for the conditional re import.
diff --git a/Lib/gettext.py b/Lib/gettext.py
@@ -48,7 +48,6 @@
 
 import operator
 import os
-import re
 import sys
 
 
@@ -70,22 +69,26 @@
 # https://www.gnu.org/software/gettext/manual/gettext.html#Plural-forms
 # http://git.savannah.gnu.org/cgit/gettext.git/tree/gettext-runtime/intl/plural.y
 
-_token_pattern = re.compile(r"""
-        (?P<WHITESPACES>[ \t]+)                    | # spaces and horizontal tabs
-        (?P<NUMBER>[0-9]+\b)                       | # decimal integer
-        (?P<NAME>n\b)                              | # only n is allowed
-        (?P<PARENTHESIS>[()])                      |
-        (?P<OPERATOR>[-*/%+?:]|[><!]=?|==|&&|\|\|) | # !, *, /, %, +, -, <, >,
-                                                     # <=, >=, ==, !=, &&, ||,
-                                                     # ? :
-                                                     # unary and bitwise ops
-                                                     # not allowed
-        (?P<INVALID>\w+|.)                           # invalid token
-    """, re.VERBOSE|re.DOTALL)
-
+_token_pattern = None
 
 def _tokenize(plural):
-    for mo in re.finditer(_token_pattern, plural):
+    global _token_pattern
+    if _token_pattern is None:
+        import re
+        _token_pattern = re.compile(r"""
+                (?P<WHITESPACES>[ \t]+)                    | # spaces and horizontal tabs
+                (?P<NUMBER>[0-9]+\b)                       | # decimal integer
+                (?P<NAME>n\b)                              | # only n is allowed
+                (?P<PARENTHESIS>[()])                      |
+                (?P<OPERATOR>[-*/%+?:]|[><!]=?|==|&&|\|\|) | # !, *, /, %, +, -, <, >,
+                                                             # <=, >=, ==, !=, &&, ||,
+                                                             # ? :
+                                                             # unary and bitwise ops
+                                                             # not allowed
+                (?P<INVALID>\w+|.)                           # invalid token
+            """, re.VERBOSE|re.DOTALL)
+
+    for mo in _token_pattern.finditer(plural):
         kind = mo.lastgroup
         if kind == 'WHITESPACES':
             continue
diff --git a/Misc/NEWS.d/next/Library/2025-01-15-19-16-50.gh-issue-118761.cbW2ZL.rst b/Misc/NEWS.d/next/Library/2025-01-15-19-16-50.gh-issue-118761.cbW2ZL.rst
@@ -0,0 +1,3 @@
+Reduce import time of :mod:`gettext` by up to ten times, by importing
+:mod:`re` on demand. In particular, ``re`` is no longer implicitly
+exposed as ``gettext.re``. Patch by Eli Schwartz.

Original file line number	Diff line number	Diff line change
`@@ -0,0 +1,3 @@`
	`1`	+Reduce import time of :mod:`gettext` by up to ten times, by importing
	`2`	+:mod:`re` on demand. In particular, ``re`` is no longer implicitly
	`3`	+exposed as ``gettext.re``. Patch by Eli Schwartz.