Pandas sort deprecated in new charts, use sort_values instead

[chdoig]

Getting this deprecation warning when using the new charts.

/anaconda/envs/training/lib/python3.4/site-packages/bokeh/charts/_attributes.py:78: FutureWarning: sort(columns=....) is deprecated, use sort_values(by=.....)
  df = df.sort(columns=columns)

Switch to using sort_values instead.

cc: @fpliger

[spawnrider]

Just had the same waning. See below :

d:\dev\python34\lib\site-packages\bokeh\charts\_attributes.py:78: FutureWarning: sort(columns=....) is deprecated, use sort_values(by=.....)
  df = df.sort(columns=columns)

[bryevdv]

There's not much we can do about this now, Pandas changed upstream, which is out of our control. If we move to sort_values then still-very-recent versions of Pandas will no longer work, which seems more undesirable than the warning.

[fpliger]

wait... this should be fixed! I made a change to use sort_values and use sort in case of the exception for older versions of pandas...

(didn't do this for examples though)

[bryevdv]

Maybe something slipped though? If there is a "compat" layer (I do seem to recall that now) then it would be great to use it on these cases.

[fpliger]

@bryevdv here's part of the fixes that I'm referring to: https://github.com/bokeh/bokeh/blob/master/bokeh/charts/attributes.py#L149-L152

I'm confused... the PR that is referring to this issue ( #3637 ) has nothing to do with with it. It's instead mostly docs and flakes fixes... Maybe it's because of a PR title (and intent) change?

[bryevdv]

#3637 originally update to sort_values directly but that change was reverted.

Edit: it originally had this commit: 42c9505 But was reverted for the reason I gave (would break with older versions pandas) so now that PR has nothing to do with this issue anymore. It's not really relevant to fixing this issue, which should now happen separately, if at all.

The examples, I think we have to leave as-is but builder.py and donut_builder.py could be fixed in the way you show above.

[fpliger]

Aaaah, now I see.. Yeap, this slipped. We can create a compat at bokeh level and put it there.. Otherwise, as it's a charts only thing we can just use charts.utils.