BUG: Stable ScalarType ordering
#29761
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Can we have a test that asserts the hard-coded sorted order? |
ngoldbaum
reviewed
Sep 18, 2025
| (np.bytes_, np.str_), | ||
| (np.str_, np.void), | ||
| # bouncy castles | ||
| (np.datetime64, np.timedelta64), |
There was a problem hiding this comment.
Total quibble but I think you can write this with just a single static list and then generate these tuples with e.g. itertools.pairwise.
There was a problem hiding this comment.
I wanted to do that at first. But I then decided against that because I didn't want to set the (IMO arbitrary) "kind" group ordering in stone (bool < complexfloating < np.floating < np.signedinteger < bouncy castles < object < flexible < unsignedinteger).
ngoldbaum
approved these changes
Sep 18, 2025
|
Thanks Joren. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The key used to sort
np.ScalarTypeis not unique (e.g.int64andlonglongon linux both have key('i', '8')). To remove duplicates,setis used, whose iteration order depends on the hashes of the containing scalartype classes. So in case of ties, the hash would be the tiebreaker. The problem is that these hashes are arbitrary, leading to differences inScalarTypeorder between platforms, python versions, machines, etc.In case you're wondering why I care: Stubtest is currently reporting errors for
ScalarTypefor certain platform and python versions: https://github.com/numpy/numtype/actions/runs/17778048694/job/50530549124?pr=706. And I'm sure you can imagine that it wasn't easy to figure out what was going on here (especially if you consider the stubtest error spaghetti output, but I digress). Hence the pedantic "backport candidate" label ;)