BUG: groupby resample different results with .agg() vs .mean() #37905
Conversation
|
Hello @jalmaguer! 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 2020-12-21 05:42:45 UTC |
|
some CI Checks errors |
|
@jreback I made some commits a few days ago since your last comments so I just wanted to bump this pull request so that it doesn't get lost and forgotten about. |
… condition instead
| # pandas\core\groupby\grouper.py:348: error: Item "None" of | ||
| # "Optional[Any]" has no attribute "take" [union-attr] | ||
| ax = self._grouper.take(obj.index) # type: ignore[union-attr] | ||
| # Sometimes self._grouper will have been resorted while |
There was a problem hiding this comment.
can you track down exactly how this happens, this is a real problem, somthing is keeping state and now we are undoing it. Your fix is pretty clean, but this makes reasoning very about this very hard.
There was a problem hiding this comment.
It’s very complicated and you have to dig through a stack trace that is 18 method calls deep to understand it but I believe it’s because self._grouper which is meant to be the TimeGrouper though in this case is actually just a sorted DatetimeIndex is always sorted and obj if you go up the stack trace you’ll see that it is coming from a SeriesGroupby._selected_obj which I’m guessing just doesn’t get sorted when it’s created. The case where self._grouper has not been resorted I believe only happens when the dates were already in order and don’t need to be resorted.
|
can you merge master (been lots of changes in groupby recently) |
@jreback Sure no problem. Just merged it. Looks like it should be good to go. |
|
@jreback Just bumping this so it doesn't get forgotten |
|
ok we had some patches which might have fixed this can u merge master (and also test if your added tests pass w/o the patch) |
|
@jreback Just merged master. Looks like my added tests do not pass without the patch. |
|
thanks @jalmaguer |
|
@meeseeksdev backport 1.2.x |
… with .agg() vs .mean()
…() vs .mean() (#38633) Co-authored-by: Jose <jalmague@gmail.com>
black pandasgit diff upstream/master -u -- "*.py" | flake8 --diff