DEPR: Deprecate tshift and integrate it to shift

[fujiaxiang]

• closes CLN: tshift & shift could be consolidated #11631
• xref BUG: groupby.tshift inconsistent behavior with other groupby transformations  #34452
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

[jreback]

thanks. would not be averse to actually re-writing tshift in terms of shift; I think this would be straightforward. Also could merge this and then followon to do that.

[fujiaxiang]

would not be averse to actually re-writing tshift in terms of shift; I think this would be straightforward

Agree it's rather straightforward, but just to clarify a few things:

1. we are still going to deprecate tshift after the re-write right? In other words, you don't mean to re-write and keep tshift right?
2. do you think it's better to move the current _tshift inside shift? i.e. we won't have a separate method _tshift, but unpack what's in it and put inside shift

[jreback]

would not be averse to actually re-writing tshift in terms of shift; I think this would be straightforward

Agree it's rather straightforward, but just to clarify a few things:

1. we are still going to deprecate tshift after the re-write right? In other words, you don't mean to re-write and keep tshift right?

correct, still deprecate

2. do you think it's better to move the current _tshift inside shift? i.e. we won't have a separate method _tshift, but unpack what's in it and put inside shift

yes if possible (ideally simplified as much as possible).

[fujiaxiang]

Have unpacked previous _tshift and moved into shift, also simplified the code a little bit. The deprecating tshift now calls shift. @jreback can you review?

[jreback]

@fujiaxiang can you rebase and some small comments, ping on green.

[fujiaxiang]

@jreback have split up the tests into PI and DTI and also separated the failure tests
I tried to parametrize the test cases into PI and DTI but I realize they only have a very small proportion of code in common, so I decided to separate them into different tests

[jreback]

@jreback have split up the tests into PI and DTI and also separated the failure tests
I tried to parametrize the test cases into PI and DTI but I realize they only have a very small proportion of code in common, so I decided to separate them into different tests

kk sounds good. ping on green (note that there maybe be a failure on master that is unrelated)

[jreback]

thanks @fujiaxiang very nice