Repository navigation
BUG: Fix structured_to_unstructured on single-field types - #13334
Conversation
There was a problem hiding this comment.
This is a strong argument for why #13326 is a good idea
There was a problem hiding this comment.
Previously returned the scalar 5.5
There was a problem hiding this comment.
If someone actually wants the result 5.5, they should just have been using np.mean(b['x'], axis=-1)
Previously a single-field type would decay, which is undesirable. The included test previously did not pass
7b4dd63 to
b5895be
Compare
|
Rebased for ease of backporting. |
|
@ahaldane Thoughts? |
|
Just got to this. Yes, it seems like the right change to make. I think it is worth a release note since it is technically a backward incompatible change. I just hope no one has depended on this yet! (seems unlikely since it a new function). After that, LGTM, I'd be happy to merge. |
eric-wieser
added
the
56 - Needs Release Note.
|
Oh I see, you are planning to put it in 1.16.3. In that case we can just go ahead and merge right? |
|
If its not too late to put in 1.16.3... |
|
Oh sorry, I was away the last week so may have caused the cutoff date to be missed. It would be nice to get it in 1.16.3... |
|
@charris, this should get 1.17.0 / 1.16.4 notes I assume? |
|
Note that our release notes for 1.16.0 barely mention
Every bug fix is a backwards-incompatible change for those depending on it. I'd perhaps argue that the changelog is sufficient here, but will let @charris arbitrate. |
Just 1.17.0, then do 1.16.4 for the backport. Note that there is no 1.17.0 in 1.16.x, so the file will need removal. |
|
Updated with a 1.17 release note, 1.16 release note created in #13396 |
mattip
removed
the
56 - Needs Release Note.
|
Thanks Eric |
charris
added
component: numpy.lib
and removed
09 - Backport-Candidate
Previously a single-field type would decay, which is undesirable.
The included test previously did not pass