fixing pandas.DataFrame.plot(): labels do not appear in legend and label kwd #9574

schmohlio · 2015-03-03T03:39:15Z

The following behavior has been tested:

df.plot(y='sin(x)') -> gives a legend with label 'None' -> this should give no legend instead (as it plots one series, and then we don't automatically add a legend, see behaviour of df['sin(x)'].plot())
df.plot(y='sin(x)', legend=True) -> gives a legend with label 'None' -> this should of course give a legend with label 'sin(x)' (behaviour as df['sin(x)'].plot(legend=True))
df.plot(y='sin(x)', label='something else', legend=True) -> gives a legend with label 'None' -> should be a legend with label 'something else', as we want that the label kwarg overwrites the column name.

based on following data:
x=np.linspace(-10,10,201)
y,z=np.sin(x),np.cos(x)
x,y,z=pd.Series(x),pd.Series(y),pd.Series(z)
df=pd.concat([x,y,z],axis=1)
df.columns=['x','sin(x)','cos(x)']
df=df.set_index('x')

schmohlio · 2015-03-03T04:29:40Z

not sure why there would be a test_url failure in the build... also not used to Travis - did the build properly incorporate both commits?

jorisvandenbossche · 2015-03-03T08:56:16Z

@schmohlio I think you can ignore that test_url error, it does not seem to be related to this PR.

You can indeed put the tests in test_graphics.
I did not yet look in detail, but a quick feedback: I don't think it should be necessary to change the default value for legend. Further, is it needed to change the existing tests?

cc @TomAugspurger @sinhrks

schmohlio · 2015-03-03T14:42:35Z

Thanks for looking @jorisvandenbossche

Is there an elegant way to check for user supplied arg (legend=) when default is not None?

Didn't spend too much time on tests, but I made changes to align plot_frame with plot_series. Previously, label wasn't getting passed through at all + it was overwriting index name within plot_frame. Per #9542, it probably doesn't make sense to mutate the index name with label=, which is reflected in test changes. That previous design flaw was exacerbated when composite .plot calls are made, and the x axis only uses label= arg from the latest call.

Will await more feedback @TomAugspurger @sinhrks . Thanks again for letting me contribute!

TomAugspurger · 2015-03-04T13:37:41Z

pandas/tools/plotting.py

@@ -886,10 +891,11 @@ def _iter_data(self, data=None, keep_index=False, fillna=None):

        from pandas.core.frame import DataFrame
        if isinstance(data, (Series, np.ndarray, Index)):
+            label = self.label if self.label else data.name


Need to be careful here. We need to treat False differently than None, in case someone wants the actual label in the legend to be the text "False". do label = self.label if self.label is not None else data.name.

TomAugspurger · 2015-03-04T13:41:45Z

I'll try to look more later.

schmohlio · 2015-03-04T14:23:21Z

Good catch.

TomAugspurger · 2015-03-06T03:18:03Z

pandas/tools/plotting.py

@@ -816,7 +816,12 @@ def __init__(self, data, kind=None, by=None, subplots=False, sharex=True,
            grid = False if secondary_y else True

        self.grid = grid
-        self.legend = legend
+        if legend is not None:


Could you explain this if block? I'm not sure I see the point of it.

The original point of this was decide if a label should be rendered based on whether there was a valid label provided or if there was more than 1 series. I can revert back the defaults, though, and make legend show default series names if no label is provided.

TomAugspurger · 2015-03-06T03:18:21Z

I agree with Joris that the default for legend doesn't need to change.

TomAugspurger · 2015-03-06T03:21:34Z

pandas/tests/test_graphics.py

@@ -2518,12 +2512,12 @@ def test_df_legend_labels(self):
        self._check_legend_labels(ax, labels=['a', 'b (right)', 'c', 'g', 'h', 'i'])

        # scatter
-        ax = df.plot(kind='scatter', x='a', y='b', label='data1')
+        ax = df.plot(kind='scatter', x='a', y='b', label='data1', legend=True)


When you change the default for legend back to True you can remove the legend=Trues that you added.

Sounds good.

schmohlio · 2015-03-07T01:37:19Z

@jorisvandenbossche @TomAugspurger - here is a summary. Pushing the following in a few:

Revert to default legend=True.
- this means that even if no label= arg is provided by the user and only 1 series is plotted, the legend will still show something based on reindex_like function #2. This will just be a difference for Series and DataFrames.
- Previously I had introduced logic based on the number of series.
- will revert back legend=True on test edits
Ensure the following logic for coalescing the legend labels:
- check label= arg (and make sure it is passed through)--> check series name property (y label, column name, etc)
Keep index name test removed, and keep other tests that check legend. Maybe expand upon these.

TomAugspurger · 2015-03-07T02:35:16Z

Thanks for the summary. That all looks perfect.

When you push again could you squash things down to a single commit and add a note in doc/source/whatsnew/v0.16.0. Thanks.

schmohlio · 2015-03-07T03:13:26Z

I added tests to the proper check legend labels method. Altogether, this test method validates all the changes we discussed. Now the label= arg functions intuitively. I also added a check to make sure there is no column mutation. Thanks for patience.

Pushing to branch now @TomAugspurger @jorisvandenbossche

TomAugspurger · 2015-03-07T03:27:51Z

Great thanks. If you could just rebase + squash and add a release not I'll get it merged.

schmohlio · 2015-03-07T04:02:31Z

Hey @TomAugspurger Tom - just squashed my commits, rebased, pulled, and pushed. Hope that's okay.

jorisvandenbossche · 2015-03-07T09:59:09Z

Is it possible to keep the behaviour as before? So that when plotting one series by specifying the y kwarg no legend is shown? (although I don't know if this is the best behaviour, but when plotting one series, it looks to me as the name of the series should be the ylabel)

schmohlio · 2015-03-07T15:17:27Z

@jorisvandenbossche - right now a legend will appear with the ylabel as the label, unless the user explicitly supplies legend=False. I agreed before but seeing it in action now it almost seems intuitive to have the legend default to true. If we are going to keep the default to True it seems difficult to cleanly override defaults in case the user wants the legend.

schmohlio · 2015-03-07T15:20:05Z

The tests also line up, seemingly making this the smallest incremental change that fixes the labels not getting passed through. What if we added another ticket for adding a legend to series plots?

@TomAugspurger I think I see what you mean by release notes- will add those. does everything else look ok?

schmohlio · 2015-03-16T00:02:30Z

@TomAugspurger I added release notes and squashed my commits properly. Figured we could go from there in a cleaner manner.

TomAugspurger · 2015-03-16T01:31:18Z

doc/source/whatsnew/v0.16.0.txt

@@ -546,6 +546,37 @@ Bug Fixes
 - Bug in ``rank`` where comparing floats with tolerance will cause inconsistent behaviour (:issue:`8365`).
 - Fixed character encoding bug in ``read_stata`` and ``StataReader`` when loading data from a URL (:issue:`9231`).
 - Bug in adding ``offsets.Nano`` to other offets raises ``TypeError`` (:issue:`9284`)
+<<<<<<< HEAD


Thanks for rebasing. You can remove this line, the one below it, and the one on 579. git puts these in when the same line has been edited on your branch and what you're rebasing onto.

TomAugspurger · 2015-03-16T01:35:16Z

I think we're really close. Code-wise everything looks good. Just those couple things in the release notes. If you want to just rebase one more time :)

I'll be in our chat room at https://gitter.im/pydata/pandas Give me a shout if you have any trouble.

TomAugspurger · 2015-03-16T01:44:47Z

And @schmohlio looks like you'll be one of the first ones in 0.16.1. Could you make a new file at doc/source/whatsnew/v0.16.1.txt And paste this in:

EDIT: the formatting was a bit odd here. Paste in the text from this gist: https://gist.github.com/TomAugspurger/7f5061219243a24ac31f

Then you can move your section in doc/source/whatsnew/v0.16.0 to that new file. Thanks.

schmohlio · 2015-03-16T03:49:12Z

cool this looks good. will shorten and rebase with new template.

jreback · 2015-03-20T13:15:05Z

whats new is now available for 0.16.1 5ebf521

schmohlio · 2015-03-20T13:48:40Z

@jreback should i remove the whatsnew file I had added?

jreback · 2015-03-20T13:49:30Z

yes, when you pull and rebase you will get the new one

fixing pandas.DataFrame.plot(): labels do not appear in legend and label kwd

TomAugspurger · 2015-03-31T01:30:18Z

Thanks @schmohlio!

Looks like this was your first PR for pandas. Stick around if you find some more bugs :)

schmohlio · 2015-03-31T02:50:33Z

@TomAugspurger thanks to you! will look out for sure.

TomAugspurger reviewed Mar 4, 2015
View reviewed changes

TomAugspurger reviewed Mar 6, 2015
View reviewed changes

jreback added Visualization plotting Bug labels Mar 6, 2015

jreback added this to the 0.16.1 milestone Mar 6, 2015

TomAugspurger reviewed Mar 6, 2015
View reviewed changes

schmohlio force-pushed the fix-plot-legends branch from 1b62e49 to f3d06b2 Compare March 15, 2015 23:56

TomAugspurger reviewed Mar 16, 2015
View reviewed changes

schmohlio force-pushed the fix-plot-legends branch from f3d06b2 to 809bd2e Compare March 16, 2015 04:12

schmohlio force-pushed the fix-plot-legends branch from 809bd2e to f60a07e Compare March 21, 2015 17:14

schmohlio force-pushed the fix-plot-legends branch 2 times, most recently from 052ea56 to 594be2a Compare March 30, 2015 20:03

fixing pandas.DataFrame.plot(): Labels do not appear in Legend

c4b4cb7

schmohlio force-pushed the fix-plot-legends branch from 594be2a to c4b4cb7 Compare March 30, 2015 20:06

TomAugspurger pushed a commit that referenced this pull request Mar 31, 2015

Merge pull request #9574 from schmohlio/fix-plot-legends

a004c59

fixing pandas.DataFrame.plot(): labels do not appear in legend and label kwd

TomAugspurger merged commit a004c59 into pandas-dev:master Mar 31, 2015

schmohlio deleted the fix-plot-legends branch March 31, 2015 02:59

TomAugspurger mentioned this pull request Mar 31, 2015

pandas.DataFrame.plot(): Labels do not appear in legend #9542

Closed

This was referenced May 1, 2015

Plot label None in line plot #8905

Closed

BUG: Fixed plot label shows as None. #8905 #8945

Closed

This was referenced Jun 1, 2015

BUG: fix Series.plot label setting #10131

Merged

Series.plot() ignores label keyword #10119

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fixing pandas.DataFrame.plot(): labels do not appear in legend and label kwd #9574

fixing pandas.DataFrame.plot(): labels do not appear in legend and label kwd #9574

schmohlio commented Mar 3, 2015

schmohlio commented Mar 3, 2015

jorisvandenbossche commented Mar 3, 2015

schmohlio commented Mar 3, 2015

TomAugspurger Mar 4, 2015

TomAugspurger commented Mar 4, 2015

schmohlio commented Mar 4, 2015

TomAugspurger Mar 6, 2015

schmohlio Mar 6, 2015

TomAugspurger commented Mar 6, 2015

TomAugspurger Mar 6, 2015

schmohlio Mar 6, 2015

schmohlio commented Mar 7, 2015

TomAugspurger commented Mar 7, 2015

schmohlio commented Mar 7, 2015

TomAugspurger commented Mar 7, 2015

schmohlio commented Mar 7, 2015

jorisvandenbossche commented Mar 7, 2015

schmohlio commented Mar 7, 2015

schmohlio commented Mar 7, 2015

schmohlio commented Mar 16, 2015

TomAugspurger Mar 16, 2015

TomAugspurger commented Mar 16, 2015

TomAugspurger commented Mar 16, 2015

schmohlio commented Mar 16, 2015

jreback commented Mar 20, 2015

schmohlio commented Mar 20, 2015

jreback commented Mar 20, 2015

TomAugspurger commented Mar 31, 2015

schmohlio commented Mar 31, 2015

fixing pandas.DataFrame.plot(): labels do not appear in legend and label kwd #9574

fixing pandas.DataFrame.plot(): labels do not appear in legend and label kwd #9574

Conversation

schmohlio commented Mar 3, 2015

schmohlio commented Mar 3, 2015

jorisvandenbossche commented Mar 3, 2015

schmohlio commented Mar 3, 2015

TomAugspurger Mar 4, 2015

Choose a reason for hiding this comment

TomAugspurger commented Mar 4, 2015

schmohlio commented Mar 4, 2015

TomAugspurger Mar 6, 2015

Choose a reason for hiding this comment

schmohlio Mar 6, 2015

Choose a reason for hiding this comment

TomAugspurger commented Mar 6, 2015

TomAugspurger Mar 6, 2015

Choose a reason for hiding this comment

schmohlio Mar 6, 2015

Choose a reason for hiding this comment

schmohlio commented Mar 7, 2015

TomAugspurger commented Mar 7, 2015

schmohlio commented Mar 7, 2015

TomAugspurger commented Mar 7, 2015

schmohlio commented Mar 7, 2015

jorisvandenbossche commented Mar 7, 2015

schmohlio commented Mar 7, 2015

schmohlio commented Mar 7, 2015

schmohlio commented Mar 16, 2015

TomAugspurger Mar 16, 2015

Choose a reason for hiding this comment

TomAugspurger commented Mar 16, 2015

TomAugspurger commented Mar 16, 2015

schmohlio commented Mar 16, 2015

jreback commented Mar 20, 2015

schmohlio commented Mar 20, 2015

jreback commented Mar 20, 2015

TomAugspurger commented Mar 31, 2015

schmohlio commented Mar 31, 2015