Colorbar in FacetGrid

[demianw]

Hi! I've used the great FacetGrid class to plot a series of 2D plots (e.g. contourf).
However, I have not found a way to activate the colorbars such as those existing in matplotlib's AxesGrid toolkit (http://matplotlib.org/mpl_toolkits/axes_grid/users/overview.html).

Is there a way to do this that I'm missing, do you have a recommendation?

Thanks!

[mwaskom]

Can you show an example of how you're using FacetGird for this? It's not straightforward to begin with so that would be helpful.

[demianw]

Here you have an example. I could probably do a pull request to add a map_ method that extends the FacetGrid functionality (I will let you pick the right to put in)

import numpy as np
import pandas
import seaborn as sns

x, y = np.meshgrid(np.linspace(0, 100), np.linspace(0, 100))

df = pandas.DataFrame(dict(
       x=r_[x.ravel(), x.ravel()],
       y=r_[y.ravel(), y.ravel()],
       s=r_[
           np.ones_like(x.ravel()),
           np.ones_like(x.ravel()) * 10
       ]
))

df['value'] = (
   df.x *
   exp(-((df.x - df.y) ** 2 / df.s ** 2))
)

g = sns.FacetGrid(df, col='s')

def data_frame_to_matrix(fn):
   def dummy(*args, **kwargs):
       data_ = kwargs['data']
       kwargs_ = kwargs.copy()
       del kwargs_['data']
       del kwargs_['color']
       data_ = data_.pivot(args[1], args[0])[args[2]]
       X, Y = meshgrid(data_.columns, data_.index)
       return fn(X, Y, data_, *args[3:], **kwargs_)
   return dummy

g.map_dataframe(
   data_frame_to_matrix(contourf),
   'x', 'y', 'value', 5,
)

[mwaskom]

Interestingly, just putting plt.colorbar at the end of your code actually does a nice job: (I also added vmin and vmax in map to ensure that the colorbar limits are valid for all facets):

Though I doubt this will generally work for all functions that could use a colorbar.

I wonder why fig._gci() is a private method; that's what plt.colorbar is using to find the most sensible thing to draw a colorbar for, and it would be useful for seaborn to take advantage of, but it's private.

[mwaskom]

See also matplotlib/matplotlib#4517

[mwaskom]

Unfortunately I've gotten a completely unhelpful response from matplotlib core (no surprise there) so it doesn't look like this is going to go anywhere...

[demianw]

Thanks! What I wanted was one colorbar per grid plot that's why the plt.colorbar was not a feasible solution. Sorry if I didn't express this well. Probably mimicking some of the work on matplotlib axesgrid. But I don't know how compatible this is with your gridding method and it might be better strategy to wrap their axesgrid with some code that generates the grid using data frame info

Demian Wassermann, PhD
demian.wassermann@inria.fr
Research Scientist
Athena Project Team
INRIA Sophia Antipolis - Méditerranée
2004 route des lucioles - FR-06902

On Jun 12, 2015, at 17:58, Michael Waskom notifications@github.com wrote:

Unfortunately I've gotten a completely unhelpful response from matplotlib core (no surprise there) so it doesn't look like this is going to go anywhere...

—
Reply to this email directly or view it on GitHub.

[mwaskom]

Oh, that's actually somewhat easier. Here's what I would do:

g = sns.FacetGrid(df, col='s')

def contourplot(*args, **kwargs):

    data = kwargs.pop("data").pivot(args[1], args[0])[args[2]]
    X, Y = np.meshgrid(data.columns, data.index)
    ax = plt.gca()
    mappable = ax.contourf(X, Y, data, *args[3:], **kwargs)
    ax.figure.colorbar(mappable)

g.map_dataframe(contourplot, 'x', 'y', 'value', 5)

[demianw]

Great! It was the ax.figure bit that I was missing! Thanks

Demian Wassermann, PhD
demian.wassermann@inria.fr
Research Scientist
Athena Project Team
INRIA Sophia Antipolis - Méditerranée
2004 route des lucioles - FR-06902

On Jun 12, 2015, at 18:28, Michael Waskom notifications@github.com wrote:

Oh, that's actually somewhat easier. Here's what I would do:

g = sns.FacetGrid(df, col='s')

def contourplot(_args, *_kwargs):

data = kwargs.pop("data").pivot(args[1], args[0])[args[2]]
X, Y = np.meshgrid(data.columns, data.index)
ax = plt.gca()
mappable = ax.contourf(X, Y, data, *args[3:], **kwargs)
ax.figure.colorbar(mappable)

g.map_dataframe(contourplot, 'x', 'y', 'value', 5)

—
Reply to this email directly or view it on GitHub.

[mwaskom]

Glad it works!