From ec78492ef1ae8db70d6b85c03310fad27af6170d Mon Sep 17 00:00:00 2001
From: Adam Turner <9087854+aa-turner@users.noreply.github.com>
Date: Thu, 3 Apr 2025 01:35:47 +0100
Subject: [PATCH 1/6] Optimise import time for ``string``
---
Lib/string.py | 64 +++++++++++--------
...-04-03-01-35-02.gh-issue-118761.VQcj70.rst | 2 +
2 files changed, 41 insertions(+), 25 deletions(-)
create mode 100644 Misc/NEWS.d/next/Library/2025-04-03-01-35-02.gh-issue-118761.VQcj70.rst
diff --git a/Lib/string.py b/Lib/string.py
index c4f05c7223ce8a..e3416135990179 100644
--- a/Lib/string.py
+++ b/Lib/string.py
@@ -49,29 +49,20 @@ def capwords(s, sep=None):
####################################################################
-import re as _re
-from collections import ChainMap as _ChainMap
-
+_sentinel_flags = object()
_sentinel_dict = {}
-class Template:
- """A string class for supporting $-substitutions."""
- delimiter = '$'
- # r'[a-z]' matches to non-ASCII letters when used with IGNORECASE, but
- # without the ASCII flag. We can't add re.ASCII to flags because of
- # backward compatibility. So we use the ?a local flag and [a-z] pattern.
- # See https://bugs.python.org/issue31672
- idpattern = r'(?a:[_a-z][_a-z0-9]*)'
- braceidpattern = None
- flags = _re.IGNORECASE
-
- def __init_subclass__(cls):
- super().__init_subclass__()
- if 'pattern' in cls.__dict__:
+class _TemplatePattern:
+ def __get__(self, instance, cls=None):
+ if cls is None:
+ return self
+ import re
+ if ('pattern' in cls.__dict__
+ and not isinstance(cls.__dict__['pattern'], _TemplatePattern)):
pattern = cls.pattern
else:
- delim = _re.escape(cls.delimiter)
+ delim = re.escape(cls.delimiter)
id = cls.idpattern
bid = cls.braceidpattern or cls.idpattern
pattern = fr"""
@@ -82,7 +73,32 @@ def __init_subclass__(cls):
(?P<invalid>) # Other ill-formed delimiter exprs
)
"""
- cls.pattern = _re.compile(pattern, cls.flags | _re.VERBOSE)
+ if cls.flags is _sentinel_flags:
+ cls.flags = re.IGNORECASE
+ pattern = re.compile(pattern, cls.flags | re.VERBOSE)
+ # replace this descriptor with the compiled pattern
+ setattr(cls, 'pattern', pattern)
+ return pattern
+
+
+class Template:
+ """A string class for supporting $-substitutions."""
+
+ delimiter = '$'
+ # r'[a-z]' matches to non-ASCII letters when used with IGNORECASE, but
+ # without the ASCII flag. We can't add re.ASCII to flags because of
+ # backward compatibility. So we use the ?a local flag and [a-z] pattern.
+ # See https://bugs.python.org/issue31672
+ idpattern = r'(?a:[_a-z][_a-z0-9]*)'
+ braceidpattern = None
+ flags = _sentinel_flags # default: re.IGNORECASE
+
+ # use a descriptor to be able to defer the import of `re`, for performance
+ pattern = _TemplatePattern()
+
+ def __init_subclass__(cls):
+ super().__init_subclass__()
+ cls.pattern = _TemplatePattern()
def __init__(self, template):
self.template = template
@@ -105,7 +121,8 @@ def substitute(self, mapping=_sentinel_dict, /, **kws):
if mapping is _sentinel_dict:
mapping = kws
elif kws:
- mapping = _ChainMap(kws, mapping)
+ from collections import ChainMap
+ mapping = ChainMap(kws, mapping)
# Helper function for .sub()
def convert(mo):
# Check the most common path first.
@@ -124,7 +141,8 @@ def safe_substitute(self, mapping=_sentinel_dict, /, **kws):
if mapping is _sentinel_dict:
mapping = kws
elif kws:
- mapping = _ChainMap(kws, mapping)
+ from collections import ChainMap
+ mapping = ChainMap(kws, mapping)
# Helper function for .sub()
def convert(mo):
named = mo.group('named') or mo.group('braced')
@@ -170,10 +188,6 @@ def get_identifiers(self):
self.pattern)
return ids
-# Initialize Template.pattern. __init_subclass__() is automatically called
-# only for subclasses, not for the Template class itself.
-Template.__init_subclass__()
-
########################################################################
# the Formatter class
diff --git a/Misc/NEWS.d/next/Library/2025-04-03-01-35-02.gh-issue-118761.VQcj70.rst b/Misc/NEWS.d/next/Library/2025-04-03-01-35-02.gh-issue-118761.VQcj70.rst
new file mode 100644
index 00000000000000..257ad7ece7d18a
--- /dev/null
+++ b/Misc/NEWS.d/next/Library/2025-04-03-01-35-02.gh-issue-118761.VQcj70.rst
@@ -0,0 +1,2 @@
+Improve import times by up to 27x for the :mod:`string` module.
+Patch by Adam Turner.
From 2b542c76445ec12ffaa4b6b869d6be8192d3e044 Mon Sep 17 00:00:00 2001
From: Adam Turner <9087854+aa-turner@users.noreply.github.com>
Date: Thu, 3 Apr 2025 02:52:55 +0100
Subject: [PATCH 2/6] Refactor to a common classmethod
---
Lib/string.py | 55 +++++++++++++++++++++++++++------------------------
1 file changed, 29 insertions(+), 26 deletions(-)
diff --git a/Lib/string.py b/Lib/string.py
index e3416135990179..1a62b699dc2ac6 100644
--- a/Lib/string.py
+++ b/Lib/string.py
@@ -49,36 +49,15 @@ def capwords(s, sep=None):
####################################################################
-_sentinel_flags = object()
_sentinel_dict = {}
+_sentinel_flags = object()
class _TemplatePattern:
def __get__(self, instance, cls=None):
if cls is None:
return self
- import re
- if ('pattern' in cls.__dict__
- and not isinstance(cls.__dict__['pattern'], _TemplatePattern)):
- pattern = cls.pattern
- else:
- delim = re.escape(cls.delimiter)
- id = cls.idpattern
- bid = cls.braceidpattern or cls.idpattern
- pattern = fr"""
- {delim}(?:
- (?P<escaped>{delim}) | # Escape sequence of two delimiters
- (?P<named>{id}) | # delimiter and a Python identifier
- {{(?P<braced>{bid})}} | # delimiter and a braced identifier
- (?P<invalid>) # Other ill-formed delimiter exprs
- )
- """
- if cls.flags is _sentinel_flags:
- cls.flags = re.IGNORECASE
- pattern = re.compile(pattern, cls.flags | re.VERBOSE)
- # replace this descriptor with the compiled pattern
- setattr(cls, 'pattern', pattern)
- return pattern
+ return cls._compile_pattern()
class Template:
@@ -93,12 +72,36 @@ class Template:
braceidpattern = None
flags = _sentinel_flags # default: re.IGNORECASE
- # use a descriptor to be able to defer the import of `re`, for performance
- pattern = _TemplatePattern()
+ pattern = _TemplatePattern() # use a descriptor to compile the pattern
def __init_subclass__(cls):
super().__init_subclass__()
- cls.pattern = _TemplatePattern()
+ cls._compile_pattern()
+
+ @classmethod
+ def _compile_pattern(cls):
+ import re # deferred import, for performance
+
+ cls_pattern = cls.__dict__.get('pattern')
+ if cls_pattern and not isinstance(cls_pattern, _TemplatePattern):
+ # Prefer a pattern defined on the class.
+ pattern = cls_pattern
+ else:
+ delim = re.escape(cls.delimiter)
+ id = cls.idpattern
+ bid = cls.braceidpattern or cls.idpattern
+ pattern = fr"""
+ {delim}(?:
+ (?P<escaped>{delim}) | # Escape sequence of two delimiters
+ (?P<named>{id}) | # delimiter and a Python identifier
+ {{(?P<braced>{bid})}} | # delimiter and a braced identifier
+ (?P<invalid>) # Other ill-formed delimiter exprs
+ )
+ """
+ if cls.flags is _sentinel_flags:
+ cls.flags = re.IGNORECASE
+ pat = cls.pattern = re.compile(pattern, cls.flags | re.VERBOSE)
+ return pat
def __init__(self, template):
self.template = template
From e28e43ff093256eade1fe965e6e72eb352c75477 Mon Sep 17 00:00:00 2001
From: Adam Turner <9087854+aa-turner@users.noreply.github.com>
Date: Thu, 3 Apr 2025 02:55:36 +0100
Subject: [PATCH 3/6] Add comment
---
Lib/string.py | 1 +
1 file changed, 1 insertion(+)
diff --git a/Lib/string.py b/Lib/string.py
index 1a62b699dc2ac6..af762894667e18 100644
--- a/Lib/string.py
+++ b/Lib/string.py
@@ -57,6 +57,7 @@ class _TemplatePattern:
def __get__(self, instance, cls=None):
if cls is None:
return self
+ # This descriptor is overwritten in ``_compile_pattern()``.
return cls._compile_pattern()
From 056dd07bb50834ee07601661ff00de6f6bcd72fd Mon Sep 17 00:00:00 2001
From: Adam Turner <9087854+aa-turner@users.noreply.github.com>
Date: Sun, 6 Apr 2025 18:11:27 +0100
Subject: [PATCH 4/6] Use None as the sentinel
---
Lib/string.py | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/Lib/string.py b/Lib/string.py
index af762894667e18..c4fb69c4ec7a04 100644
--- a/Lib/string.py
+++ b/Lib/string.py
@@ -50,7 +50,6 @@ def capwords(s, sep=None):
####################################################################
_sentinel_dict = {}
-_sentinel_flags = object()
class _TemplatePattern:
@@ -71,7 +70,7 @@ class Template:
# See https://bugs.python.org/issue31672
idpattern = r'(?a:[_a-z][_a-z0-9]*)'
braceidpattern = None
- flags = _sentinel_flags # default: re.IGNORECASE
+ flags = None # default: re.IGNORECASE
pattern = _TemplatePattern() # use a descriptor to compile the pattern
@@ -99,7 +98,7 @@ def _compile_pattern(cls):
(?P<invalid>) # Other ill-formed delimiter exprs
)
"""
- if cls.flags is _sentinel_flags:
+ if cls.flags is None:
cls.flags = re.IGNORECASE
pat = cls.pattern = re.compile(pattern, cls.flags | re.VERBOSE)
return pat
From bb3605a7d89acb288cabdf5e29dffc11b2567251 Mon Sep 17 00:00:00 2001
From: Adam Turner <9087854+aa-turner@users.noreply.github.com>
Date: Mon, 7 Apr 2025 21:52:14 +0100
Subject: [PATCH 5/6] Make _TemplatePattern a singleton
---
Lib/string.py | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/Lib/string.py b/Lib/string.py
index c4fb69c4ec7a04..9ba0769e93eee3 100644
--- a/Lib/string.py
+++ b/Lib/string.py
@@ -53,11 +53,12 @@ def capwords(s, sep=None):
class _TemplatePattern:
+ # This descriptor is overwritten in ``Template._compile_pattern()``.
def __get__(self, instance, cls=None):
if cls is None:
return self
- # This descriptor is overwritten in ``_compile_pattern()``.
return cls._compile_pattern()
+_TemplatePattern = _TemplatePattern()
class Template:
@@ -72,7 +73,7 @@ class Template:
braceidpattern = None
flags = None # default: re.IGNORECASE
- pattern = _TemplatePattern() # use a descriptor to compile the pattern
+ pattern = _TemplatePattern # use a descriptor to compile the pattern
def __init_subclass__(cls):
super().__init_subclass__()
@@ -83,7 +84,7 @@ def _compile_pattern(cls):
import re # deferred import, for performance
cls_pattern = cls.__dict__.get('pattern')
- if cls_pattern and not isinstance(cls_pattern, _TemplatePattern):
+ if cls_pattern is not None and cls_pattern is not _TemplatePattern:
# Prefer a pattern defined on the class.
pattern = cls_pattern
else:
From 6fb9b5d65c2822012efe47bf7158227bcfde4ce0 Mon Sep 17 00:00:00 2001
From: Adam Turner <9087854+AA-Turner@users.noreply.github.com>
Date: Tue, 8 Apr 2025 10:39:27 +0100
Subject: [PATCH 6/6] Serhiy's suggestion
Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
---
Lib/string.py | 7 ++-----
1 file changed, 2 insertions(+), 5 deletions(-)
diff --git a/Lib/string.py b/Lib/string.py
index 9ba0769e93eee3..eab5067c9b133e 100644
--- a/Lib/string.py
+++ b/Lib/string.py
@@ -83,11 +83,8 @@ def __init_subclass__(cls):
def _compile_pattern(cls):
import re # deferred import, for performance
- cls_pattern = cls.__dict__.get('pattern')
- if cls_pattern is not None and cls_pattern is not _TemplatePattern:
- # Prefer a pattern defined on the class.
- pattern = cls_pattern
- else:
+ pattern = cls.__dict__.get('pattern', _TemplatePattern)
+ if pattern is _TemplatePattern:
delim = re.escape(cls.delimiter)
id = cls.idpattern
bid = cls.braceidpattern or cls.idpattern