Seaborn: FacetGrid legend missing colors in key

[msure]

Code Sample, a copy-pastable example if possible

import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.patches as patches
import seaborn as sns
sns.set_style('whitegrid')

test = sns.load_dataset("tips")
g = sns.FacetGrid(test,
                  col="sex",
                  hue="smoker",
                  size=4, aspect=1)
g = (g.map(sns.pointplot, "time", "total_bill").add_legend())

Problem description

For some reason I cannot get the colors from the plot to appear in the key:

Here is my setup:

INSTALLED VERSIONS

commit: None
python: 3.5.2.final.0
python-bits: 64
OS: Darwin
OS-release: 15.0.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: en_US.UTF-8

pandas: 0.19.0
seaborn: 0.7.1
matplotlib: 2.0.0
jupyter: 1.0.0
jupyter_client: 4.4.0
jupyter_console: 5.0.0
jupyter_core: 4.2.0

I was on a much earlier version of Matplotlib (1.5.1), and tried upgrading to 2.0 but that did not fix the issue.

[mwaskom]

Use factorplot.

[pbashivan]

Is there a solution for this? Can factorplot be used with a custom map function like in FacetGrid?.
I have a custom histogram function for plotting on each facet. It plots the data fine but I have the same issue with the legend missing the colors.

[ManavalanG]

I have same problem of missing colors in legend when using pairgrid plot in Jupyter notebook inline.

[jkarakas]

Use plt.legend(). I had a similar case and it solved it

[kning]

@jkarakas could you elaborate? I'm having a similar problem and would like to see your solution.

[jkarakas]

@kning It is very explicit but it works.There is obviously something wrong with getting the patches from the FacetGrid object.

[mwaskom]

As I've said previously: use factorplot.

[xguse]

OK well what if I want to use swarmplot?

[Rockbaron]

That works really good. thumbs up, jkarakas. However, I have 20 hue - datasets in my plot, but legend will show me only the first 6, because color palette has only length 6. How can I recycle the colors?

[mwaskom]

Folks I was terse in my response because there are numerous examples of this in the docs but you're really overthinking things with these hacks:

sns.factorplot(x="time", y="total_bill", hue="smoker", col="sex",
               size=4, aspect=1, hue_order=["Yes", "No"],
               data=tips)

[jkarakas]

@mwaskom, thank you for the answer. You have been of course correct about factorplot.

@Rockbaron you can expand your palette if you want to recycle deep like this: sns.set_palette(sns.color_palette("deep", 20)).

There are a lot of options and examples on working with color palettes at the seaborn documentation

[ksahlin]

I had the issue of missing only one of the colors in the legend when using PairGrid with pointplot. Making sure I put NaN first in my df solved this, as: df.sort_values(<column to sort>, na_position="first"). I don't know what is going on under the hood but a possible explanation is that matplotlib/seaborn takes the last row of each hue in the pandas dataframe and uses this as label. If this is a row with NaN the color will disappear in the legend.

[N0rbi]

Extending the answer of @jkarakas , you can pass a dictionary to the add_legend function of the FacetGrid object like so: g.add_legend(legend_data= {lab: hand for lab, hand in zip(labels, handles)})
then you won't need to call plt.legend

[ThisIsVenkatesh]

Extending the answer of @jkarakas , you can pass a dictionary to the add_legend function of the FacetGrid object like so: g.add_legend(legend_data= {lab: hand for lab, hand in zip(labels, handles)})
then you won't need to call plt.legend

Thanks for your response @N0rbi

[vpapaioannou]

What about using distplot? In the catplot documentation looks like that is not supported.