Allow list-like for y in DataFrame.plot.

[TomAugspurger]

In pandas 0.22 and earlier, we passing multiple values for y worked by accident. The code clearly assumed that y was a scalar, but things usually worked, perhaps with an incidental warning:

In [12]: df = pd.DataFrame(np.random.uniform(size=(10, 2)), columns=['a', 'b'])

In [15]: df.plot(x='c', y=['a', 'b'])
/Users/taugspurger/miniconda3/envs/pandas-0.21.0/lib/python3.6/site-packages/pandas/plotting/_core.py:1714: UserWarning: Pandas doesn't allow columns to be created via a new attribute name - see https://pandas.pydata.org/pandas-docs/stable/indexing.html#attribute-access
  series.name = label
Out[15]: <matplotlib.axes._subplots.AxesSubplot at 0x10f83b198>

On master, this will currently raise. I think that we should explicitly support list-likes for y.

Tagging for 0.23 since I think this should be a blocker.

xref #18695 where we made the change.

[TomAugspurger]

@masongallo do you have any time / interest to work on this?

[masongallo]

@TomAugspurger yes I would be happy to help once I get some time, hopefully later this week.

Can you give a bit more details on what you would expect to return in your example above (or a similar example where the user supplies multiple y columns)?

[TomAugspurger]

Thanks! I think it should be equivalent to df.set_index('c')[['a', 'b']].plot(). So multiple lines, each plotted against the x.

[masongallo]

Gotcha - this only makes sense if a and b are on the "same" scale right?

The plot method has a secondary_y arg but I'm not sure it makes sense to use in a general case like this.

[TomAugspurger]

Gotcha - this only makes sense if a and b are on the "same" scale right?

To get a sensible plot, yes, but users are free to do whatever :) `secondary_y` is for creating a plot with two independent y-axes. Accepting a list-like for `y` would allow for an arbitrary number of lines all with the same y axis.

…

On Mon, Feb 26, 2018 at 11:29 AM, Mason Gallo ***@***.***> wrote: Gotcha - this only makes sense if a and b are on the "same" scale right? The plot method has a secondary_y arg but I'm not sure it makes sense to use in a general case like this. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#19699 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABQHIic1OXLIgvMTc6vIdrxOQ3f6sn6Jks5tYuoNgaJpZM4SFw5J> .

[masongallo]

Accepting a list-like for y would allow for an arbitrary number of lines all with
the same y axis.

This makes sense to me & I think the usecase is clear.