BUG: Fix setitem with 1d matrix

[ivirshup]

• closes BUG: 1.3.0 column assignment via single columnnp.matrix behaviour change #42376
• tests added / passed
• Ensure all linting tests pass, see here for how to run them
• whatsnew entry

Fixes the case df["mean"] = sparse.random(100, 10).mean(axis=1) assigning a broken column.

Wasn't sure where the test should go, I feel like there must be a test for x = np.ones((10, 1)); df["x"] = x somewhere.

[ivirshup]

Tests fail with the new array manager, but don't with the block manager. Not sure what I should do about that, as it's unclear if this is intentional.

[ivirshup]

What should be done about the tests failing with the new data manager?

I, personally, would expect the two dimensional array that only has multiple indices along one dimension to keep working for column assignment.

[jreback]

pls add a whatsnew note. i think you can just skip for array manager for now. or fix it.

[ivirshup]

pls add a whatsnew note

Added

i think you can just skip for array manager for now. or fix it.

I'm a bit short on time for the moment, so I'd opt to skip it.

I've opened an issue for this: #44788

Is there documentation somewhere about what the new array manager is exactly? My assumption is it's around the BlockManager rewrite?

[jreback]

use @td.skip_array_manager_not_yet_implemented (and put the reference to the issue you are skipping).

or can do like this

pandas/tests/reshape/concat/test_concat.py: def test_concat_copy(self, using_array_manager): (see that test)

[ivirshup]

Added the skip, I think the remaining CI failures are unrelated.

[jbrockmendel]

Is there documentation somewhere about what the new array manager is exactly?

Not really. I think there has been reticence to put it in the public docs because it is still experimental and may be removed.

[ivirshup]

I think there has been reticence to put it in the public docs because it is still experimental and may be removed.

Might be useful to mention in the contributing docs. It was a little confusing why tests were failing on CI at first.

---

Was there anything else I should do here?

[jbrockmendel]

Was there anything else I should do here?

just merge master pls

[jreback]

thanks @ivirshup very nice. sorry for the delay, quite a number of ci issues came up of late.

[ivirshup]

No worries, thanks for the merge!