GH-112855: Speed up pathlib.PurePath pickling (#112856)

barneygale · web-flow · commit 15fbd53ba96b · 2024-04-20T17:46:52.000+01:00
The second item in the tuple returned from `__reduce__()` is a tuple of arguments to supply to path constructor. Previously we returned the `parts` tuple here, which entailed joining, parsing and normalising the path object, and produced a compact pickle representation.

With this patch, we instead return a tuple of paths that were originally given to the path constructor. This makes pickling much faster (at the expense of compactness).

It's worth noting that, in the olden times, pathlib performed this parsing/normalization up-front in every case, and so using `parts` for pickling was almost free. Nowadays pathlib only parses/normalises paths when it's necessary or advantageous to do so (e.g. computing a path parent, or iterating over a directory, respectively).
diff --git a/Lib/pathlib/__init__.py b/Lib/pathlib/__init__.py
@@ -169,9 +169,7 @@ def __rtruediv__(self, key):
             return NotImplemented
 
     def __reduce__(self):
-        # Using the parts tuple helps share interned path parts
-        # when pickling related paths.
-        return (self.__class__, self.parts)
+        return self.__class__, tuple(self._raw_paths)
 
     def __repr__(self):
         return "{}({!r})".format(self.__class__.__name__, self.as_posix())
diff --git a/Misc/NEWS.d/next/Library/2023-12-07-20-05-54.gh-issue-112855.ph4ehh.rst b/Misc/NEWS.d/next/Library/2023-12-07-20-05-54.gh-issue-112855.ph4ehh.rst
@@ -0,0 +1,2 @@
+Speed up pickling of :class:`pathlib.PurePath` objects. Patch by Barney
+Gale.

Original file line number	Diff line number	Diff line change
`@@ -0,0 +1,2 @@`
	`1`	+Speed up pickling of :class:`pathlib.PurePath` objects. Patch by Barney
	`2`	`+Gale.`