CI: fix npdev build #41665
CI: fix npdev build #41665
Conversation
simonjayhawkins
added
CI
|
@jbrockmendel you are changing the code here for unreleased compat issues. IIUC numpy may make some further changes. would it be better to add something to the 54 failing tests for now? |
|
I'm holding out hope that numpy will handle this upstream, xref numpy/numpy#19078 |
| @@ -1583,7 +1584,9 @@ def searchsorted(arr, value, side="left", sorter=None) -> np.ndarray: | |||
| # and `value` is a pd.Timestamp, we may need to convert value | |||
| arr = ensure_wrapped_if_datetimelike(arr) | |||
|
|
|||
| return arr.searchsorted(value, side=side, sorter=sorter) | |||
| with catch_warnings(): | |||
| filterwarnings("ignore", category=FutureWarning) | |||
There was a problem hiding this comment.
why don't u use create_ndarray here?
There was a problem hiding this comment.
bc value can be a scalar here
|
@seberg should we just suppress these warnings for the time being in expectation of a fix/change upstream? |
|
Sorry about being away for the long weekend. I will look into this again, but I got the feeling that you will likely need some type of helper like |
|
We just discussed this in the numpy call, and I think the most likely solution will be to (unfortunately) delay the deprecation, to give you more time and hopefully a better solution to this. That would mean you don't have to worry about string promotion changes for this release. |
|
OK. Is there anything I/we can do short(ish)-term to make the process easier on your end? (e.g. reopen this PR) |
|
@jorisvandenbossche not sure. Maybe we can meet up briefly and discuss how well I have some hopes that we could both clean up pandas and make NumPy and pandas work better together in one go with just small modifications to NumPy (and exposure of the new API). And that would just solve this issue incidentally. |
Sure. I'm in UTC-7, can do any AM between now and Tuesday, inclusive. |
|
Great! Send you an invite, in case anyone is curious, put the details here also: https://hackmd.io/fe_VHCBjSwutSwhpPP8KSg |
hopefully numpy will provide an alternative that makes this unnecessary