Skip to content

BUG: Undo result type change of quantile/percentile but keep q precision#30601

Merged
charris merged 2 commits intonumpy:mainfrom
seberg:weak-q-precision
Jan 9, 2026
Merged

BUG: Undo result type change of quantile/percentile but keep q precision#30601
charris merged 2 commits intonumpy:mainfrom
seberg:weak-q-precision

Conversation

@seberg
Copy link
Copy Markdown
Member

@seberg seberg commented Jan 7, 2026 •
edited
Loading

Defer cast to input type to the end of the computation by passing through the weak_q information.

This fixes gh-30586, but I am not sure it is great to backport. Although, compared to the other PR it should just align the nan-functions fully and otherwise revert anything that isn't just a precision fix.

If we don't want to backport this, I would be tempted to just revert the original PR for 2.5, TBH.

This still needs tests, although it passes all tests added in gh-29105.

@seberg
BUG: Fix result type change and make weak q safe
f1cb7c5 
Defer cast to input type to the end of the computation by
passing through the `weak_q` information.

This fixes numpygh-30586, but I am not sure it is great to backport.
Although, compared to the other PR it should just align the nan-functions
fully and otherwise revert anything that isn't just a precision fix.
@seberg seberg requested review from eendebakpt and mhvk January 7, 2026 14:47
@seberg seberg added this to the 2.4.1 Release milestone Jan 7, 2026
@seberg
Copy link
Copy Markdown
Member Author

seberg commented Jan 7, 2026

(One tiny effect here is that previously, for arrays of longdouble, i.e. high precision, the q computation also used that higher precision. Since it now sticks to it's own default precision until the end -- we do lose that part, but I doubt there is much point in trying that hard...)

@seberg seberg added the 09 - Backport-Candidate PRs tagged should be backported label Jan 7, 2026
@seberg seberg changed the title BUG: Fix result type change and make weak q safe BUG: Undo result type change of quantile/percentile but keeping q precision Jan 8, 2026
@seberg seberg changed the title BUG: Undo result type change of quantile/percentile but keeping q precision BUG: Undo result type change of quantile/percentile but keep q precision Jan 8, 2026
@charris
Copy link
Copy Markdown
Member

charris commented Jan 9, 2026

I will put this before 2.4.1.

mhvk
mhvk approved these changes Jan 9, 2026
View reviewed changes
Copy link
Copy Markdown
Contributor

@mhvk mhvk left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks good. The direct test for in (int, float) make me wonder, though, if this isn't something that can be done with _array_converter? But you probably considered that, since you added that... Anyway, if that doesn't work for some reason, or is just too big an overhaul for this fixup, fine to go with this.

@charris charris merged commit b91f456 into numpy:main Jan 9, 2026
75 checks passed
@charris
Copy link
Copy Markdown
Member

charris commented Jan 9, 2026

Thanks Sebastian.

@charris charris mentioned this pull request Jan 9, 2026
Merged
@charris charris removed the 09 - Backport-Candidate PRs tagged should be backported label Jan 9, 2026
charris added a commit that referenced this pull request Jan 9, 2026
@charris
Merge pull request #30623 from charris/backport-30601
abe3d86 
BUG: Undo result type change of quantile/percentile but keep q precision (#30601)
@seberg
Copy link
Copy Markdown
Member Author

seberg commented Jan 10, 2026

Yeah, I considered it and thought it doesn't matter. _array_converter also solves the __array_wrap__ which I guess doesn't apply (at least not for this fix).
OTOH, it might still be a neat thing, just to try to remove explicit type checks...

@seberg
Copy link
Copy Markdown
Member Author

seberg commented Jan 22, 2026

Dang it, I forgot to undo one line of code and the tests didn't notice!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@mhvk mhvk mhvk approved these changes

@eendebakpt eendebakpt Awaiting requested review from eendebakpt

Assignees

No one assigned

Labels

00 - Bug

Projects

None yet

Milestone

2.4.1 Release

Development

Successfully merging this pull request may close these issues.

BUG: Return type changed in percentile for higher precision q

3 participants

@seberg @charris @mhvk