plot_bgcolor in notebook

[hughesadam87]

I'm running an ipython notebook that sets the default matplotlib background plot color as gray. The look fine; however, when I do:

py.iplot_mpl(fig)

The background color returns to white. Furthermore, if I explicitly pass a background color:

py.iplot_mpl(fig, plot_bgcolor='rgb(128,0,0)')

It is still white. Is py.iplot_mpl not passing the background color attribute through?

[etpinard]

I am confirming the bug. matplotlib background colors are not successfully converted to Plotly.

Use the following code instead:

import matplotlib.pyplot as plt
fig = plt.figure()
# --> your matplotlib functions <--

import plotly.plotly as py  
import plotly.tools as tls

# Convert matplotlib figure to plotly figure object
fig_plotly = tls.mpl_to_plotly(fig)

# Add 'plot_bgcolor' to layout
fig_plotly['layout'].update(plot_bgcolor='rgb(128,0,0)')

# Send updated figure object to plotly
py.iplot(fig_plotly,filename='matplotlylib-bgcolor-test1')

[theengineear]

Yes, you do have to manually add the plot_bgcolor attribute right now. There are a number of parameters that we currently don't support.

These get updated fairly often though, and when we feel confident that we have a new, large set of portable features from mpl, we'll likely shoot an email out to python users to discuss the new features.

Thanks for the feedback!

[hughesadam87]

Thanks for the workaround. Perhaps down the road, this behavior could be incorporated into iplot_mpl() through a special keyword like overload as in:

iplot_mpl(fig, overload={'plot_bgcolor':'rgb(128,0,0)'}, **kwds)

Of course, the solution above is fine, but I can imagine this happening quite often as plotly catches on, and you'll have to re-explain this a lot.

Thanks for the help.

[hughesadam87]

Any qualms if I put that aforementioned overload keyword into a wrapper in one of my libraries for the time being, and incorporate it into a tutorial I'm writing?

[theengineear]

If it will make the tutorial easier to work though, by all means, go for it!

I'm not sure that something quite like this would be adopted because of the ambiguity of the keywords in overload. There are many repeating keywords in plotly's figure objects since the organization is encapsulated in sub-dictionaries. Therefore, I'd be more in favor of a solution that looked like this:

figure_update = {'layout': {'plot_bgcolor':'rgb(128,0,0)'}}
iplot_mpl(fig, update=figure_update, **kwds)

In the background, then, it's clear that we're just going to update the resulting figure with the figure_update dictionary (with protection from nesting stuff).

We're trying to not do too much magic in the background after user calls. I'm certainly open to any solution here that doesn't introduce any ambiguity.

If this solution seems like it would work for you, I' be happy to write it in (even before your tutorial) if you give me a timeframe on it.

[hughesadam87]

Hi,

Yes, I think that solution is better. We aren't really hard pressed for time, but if it does get written in, we'll certainly start using it asap. I'll keep an eye on this thread. If it does get changed, can you drop me a line either here or @hughesadam87 on twitter?

Thanks

[theengineear]

Ok, I pushed a branch called 'mpl-updated' that should fix this. Here's an NB showing how it'll work:
http://nbviewer.ipython.org/gist/theengineear/0719fd0ac63c9b4ceba3

If you like this, let me know and i'll merge it and update our PyPI.

[theengineear]

@hugadams , just wanted to make sure you saw this update for the changes we've been talking about.

[hughesadam87]

@theengineear

Thanks. I'm going to try it now; looks exactly what I had hoped for. Will be back in 30 min or so and will share our updated plotly tutorial. BTB

[hughesadam87]

So, I think you are fine to update PyPI based on the notebook you shared, but after installing this, I cannot import plotly. I've completely removed plotly from my sitepackages prior to installing; however, I'm using Enthought canopy and it could be that there's some other vestige of the old version somewhere. Here's the error I get, not sure if its reproducible:

        ---------------------------------------------------------------------------
        ImportError                               Traceback (most recent call last)
        <ipython-input-3-bd46b6da962d> in <module>()
        ----> 1 import plotly.plotly as py
              2 #import plotly.tools as tls
              3 
              4 #tls.get_credentials()
              5 #py.sign_in('user', 'pw')

        /home/glue/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/plotly-1.0.26-py2.7.egg/plotly/__init__.py in <module>()
              1 from . version import __version__
        ----> 2 import graph_objs
              3 import plotly
              4 import tools
              5 import utils

        /home/glue/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/plotly-1.0.26-py2.7.egg/plotly/graph_objs/__init__.py in <module>()
             44 
             45 """
        ---> 46 from graph_objs import *
             47 
             48 __all__ = ["Data",

        /home/glue/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/plotly-1.0.26-py2.7.egg/plotly/graph_objs/graph_objs.py in <module>()
             27 import json
             28 import textwrap
        ---> 29 from .. import exceptions
             30 from .. import utils
             31 

        ImportError: cannot import name exceptions

[theengineear]

[theengineear]

@hugadams , i can't seem to reproduce it, but i'm going to go ahead an change our module name, excpetions looks to be causing a name clash.