BUG: merge_asof should be treated as a left join

[phofl]

• closes merge_asof(left_index=True, right_on=...) overwrites left index with right index #33463
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

The fix broke one test (test_merge_index_column_tz in test_merge_asof.py) which expected the right index instead of the left index.

The default index selection does not work in case of asof and left_index=True. I had to catch this case here.

[WillAyd]

@phofl is this still active? If so can you address @mroeschke comments / questions?

[pep8speaks]

Hello @phofl! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2021-04-11 11:47:41 UTC

[phofl]

@WillAyd Yeah, for sure. Missed that completely, sorry.

[phofl]

@jreback
Did you have something like this in mind?

I have a more general question about the _get_join_info part. The else part starting in line 884 is a bit weird in my opinion. The if condition if self.right_index is only reached in cases of inner, right and outer join. In case of a right join, we use the left index on purpose here. So the following code

import pandas as pd

index = pd.Index([1, 5, 10], name="test")
left = pd.DataFrame({"left": [1, 2, 3, 6, 7]}, index=[1, 2, 3, 6, 7])
right = pd.DataFrame({"right": ["a", "b", "c"], "right_time": [1, 4, 10]}, index=index)
result = pd.merge(left, right, right_index=True, left_on='left', how="right")

produces the left index.

     left right  right_time
1.0     1     a           1
NaN     5     b           4
NaN    10     c          10

although I would expect the right index here. Is this assumption correct? In this case I would modify the function calls of _create_join_index in a way to produce the desired results and keep the correct behavior in case of outer and inner joins. There are a few open issues about this topic.

[phofl]

@jreback Could you maybe have a look?

[github-actions]

This pull request is stale because it has been open for thirty days with no activity. Please update or respond to this comment if you're still interested in working on this.

[itamarst]

@jreback seems like all comments were addressed, could you take a look? Thanks!

[jreback]

can you merge master and move the note to 1.3 and will look.

[phofl]

Done

[jreback]

@phofl oldie, but if you can merge will take a look

[phofl]

@jreback merged, failure unrelated

[jreback]

ok let me look closer at this.

[mroeschke]

Could you write a comment here explaining why we're doing how="left" here?

[phofl]

Done

[mroeschke]

Generally this looks okay to me. Mind merging in master?

[phofl]

merged master

[jreback]

thanks @phofl very nice!