BUG: Doctest doesn't have a SHOW_WARNINGS directive. #23819
Conversation
| @@ -154,7 +154,7 @@ def geterr(): | |||
| -------- | |||
| >>> np.geterr() | |||
| {'divide': 'warn', 'over': 'warn', 'under': 'ignore', 'invalid': 'warn'} | |||
| >>> np.arange(3.) / np.arange(3.) # doctest: +SHOW_WARNINGS | |||
There was a problem hiding this comment.
@bsipocz Hmm, removing that fails differently on the refguide checks. I think the best short term fix is to skip that test. Could you make sure to add it back in your PR?
| @@ -154,7 +154,7 @@ def geterr(): | |||
| -------- | |||
| >>> np.geterr() | |||
| {'divide': 'warn', 'over': 'warn', 'under': 'ignore', 'invalid': 'warn'} | |||
| >>> np.arange(3.) / np.arange(3.) # doctest: +SHOW_WARNINGS | |||
| >>> np.arange(3.) / np.arange(3.) | |||
| array([nan, 1., 1.]) | |||
| RuntimeWarning: invalid value encountered in divide | |||
There was a problem hiding this comment.
I believe the exception needs to be removed here in here.
| RuntimeWarning: invalid value encountered in divide |
There was a problem hiding this comment.
I think the warning is correct, but maybe warnings are set up differently in the two runs. With pytest, we have the problem that warnings are set to errors (when they should probably not be for the doctest part), and maybe the current checks don't handle them at all?
In any case, I tend towards the +SKIP. If we forget about it... this function should maybe be hidden away from public anyway...
acd6d02 to
37e14b7
Compare
|
Not sure that SKIP is ideal, but I do think the warning should be in the docs (it would show the warning in a normal setup there), so SKIP is hopefully OK for now and unblocks CI. Thanks for fixing this up Chuck! |
No description provided.