Sorry, something went wrong.
|
|
||
| * Pure ASCII strings are now normalized in constant time by :func:`unicodedata.normalize`. | ||
| (Contributed by Dong-hee Na in :issue:`44987`.) | ||
| * :func:`sum` is now nearly 30% faster for integers smaller than ``2**30``. |
There was a problem hiding this comment.
sum() is still much slower than 2.7.
Maybe we shouldn't mention this, until it is fixed in 3.12
Sorry, something went wrong.
There was a problem hiding this comment.
Sure, though on the other hand, it still a fairly substantial improvement over 3.10, the previous feature version, which seems to me is the main focus of this section—at least in the past, I'm not sure optimizations were only mentioned if they exceeded 2.7's speed for something, since that likely wasn't the case for quite a few of them. That said, I'm certainly not the expert here (unlike you all), and will of course defer to the judgement of @pablogsal as RM and you and rest of the core dev performance folks on whether this should be included, if you think its best left out.
Sorry, something went wrong.
There was a problem hiding this comment.
On discuss - release, Pable said "I don’t think there is anything wrong mentioning that is faster than the version in 3.10. After all there was work here to make the situation better and if true, I don’t see why a considerable improvement over the previous version should not be mentioned". Sum is not new, but since we have an optimization section, I agree.
Sorry, something went wrong.
|
Thanks @CAM-Gerlach for the PR, and @pablogsal for merging it 🌮🎉.. I'm working now to backport this PR to: 3.11. |
Sorry, something went wrong.
|
GH-98554 is a backport of this pull request to the 3.11 branch. |
Sorry, something went wrong.
Part of #95913
I've added Sphinx markup and cross references to the Optimizations section of the Python What's New document, copyedited the text to fix grammar errors, read better and be slightly clearer, and sorted it by language, then builtins, then standard library (sorted by module name).
Additionally, I've added a handful of additional optimizations to common functions with documented substantial (>=15%) performance increases mentioned in the changelog (that were unrelated to Faster CPython), and consistent with the existing optimizations mentioned here and in previous verisons. These are: