faceted plots

[clarkfitzg]

Opening this to provide visibility as I work on this PR.

Currently this has x and y string arguments for 2d plot working, but everything else still needs lots of work. I'll let people know when it's ready for review.

[clarkfitzg]

@shoyer Here's what we currently have:

In [15]: g = xray.plot.FacetGrid(t, col='time', col_wrap=2)

In [16]: g.map_dataarray(xray.plot.contourf, 'lon', 'lat')
Out[16]: <xray.plot.facetgrid.FacetGrid at 0x10de082d0>

Right now this assumes that the function for map_dataarray has the same signature as 2d xray plotting methods.

[clarkfitzg]

Here's what a 4d array looks like.

In [26]: t4d = xray.concat([t2, t2 + 50], pd.Index(['normal', 'hot'], name='fourth_dim'))

# This is a 4d array
In [27]: t4d.coords
Out[27]: 
Coordinates:
  * lat         (lat) float32 75.0 72.5 70.0 67.5 65.0 62.5 60.0 57.5 55.0 ...
  * lon         (lon) float32 200.0 202.5 205.0 207.5 210.0 212.5 215.0 ...
  * time        (time) datetime64[ns] 2013-01-01 2013-03-04T12:00:00
  * fourth_dim  (fourth_dim) object 'normal' 'hot'

In [28]: g = xray.plot.FacetGrid(t4d, col='time', row='fourth_dim')

In [29]: g.map_dataarray(xray.plot.imshow, 'lon', 'lat')
Out[29]: <xray.plot.facetgrid.FacetGrid at 0x10f5aad90>

[jhamman]

@clarkfitzg - These are looking really great. I think this is going to be a really useful tool.

The colorbar label is too close to the colorbar tick label, are we manually placing that or is this a matplotlib thing?

[clarkfitzg]

Thanks! The colorbar label spacing is a matplotlib thing- it only became a problem after rotating it to match the row labels. So I expect that I can fix it by changing the anchor point or placing it manually.

[shoyer]

Maybe you can write something like the facet_data method from Seaborn's FacetGrid? That seems like a nice separation of the iteration logic from the plotting logic

[clarkfitzg]

That's on my todo list- I better post that list here.

[clarkfitzg]

For this PR:

• facet on rows and columns
• Align label on colorbar
• Label figure axes
• Use xray plotting logic to make nice colorbar
• Make and test choices around colorbar

EDIT - going to revisit plt.contour(color='k') after #537 gets worked out.

• separate iteration over data from plotting - edge case: require the index to actually be unique .is_unique.
• Use core string formatting for titles
• tests for all of above
  • test col_wrap
• write and organize docs
  • docs for public attribute- axes and such

On this PR or possibly a different one:

• Convenience method to easily make faceted plots- something like darray.plot(col='time')
• port Seaborn's aspect capabilities
• matplotlib quiver plot for map example
• Create new xray.tutorial module with function like load_example_dataset for plotting examples

Different PR:

• Facets and maps
• Inline documentation and plots for API reference

[shoyer]

This is a minor point, but take a look at the PEP8 indentation guidelines:
https://www.python.org/dev/peps/pep-0008/#indentation

This is an example of the first "No" case :)

[clarkfitzg]

Just ran this through pep8 command line utility- lots of violations. Fixed.

[shoyer]

For the simple example in the doc -- use imshow instead of contourf? The result is definitely more aesthetically pleasing:

[shoyer]

My inclination is to not add this convenience shortcut, for two reasons:

1. It's a departure from Seaborn, so it's something new for users to learn.
2. It's not really necessary -- iterating over grid.axes.flat is much more explicit (generally a good thing), and only slightly more verbose.

[clarkfitzg]

I added this with the idea of using an abc with this class: https://docs.python.org/2/library/collections.html#collections-abstract-base-classes.

Reviewing the available abc's, I don't think we stand to gain any benefit from them. At the same time, it is conceptually a 2d array. I suppose that structure is captured well enough in the axes and name_dicts.

[shoyer]

Let's call this parameter data, like Seaborn. Eventually we might support dataset objects, too.

[clarkfitzg]

@shoyer Latest commit also added a little more to the docs on how to use the attributes data axes, and name_dicts.

[shoyer]

These variables are all duplicated between private versions (below) and public versions (above). For now, let's stick with the conservative choice of keeping them private like seaborn?

[clarkfitzg]

good catch

[clarkfitzg]

what is this doing? I think nothing.

[shoyer]

Maybe this is undoing the seaborn import? If it doesn't seem to be changing the doc build, take it out.

[shoyer]

one more suggestion: clarkfitzg#2

we can remove this methods and let them just show up on the doc page for FacetGrid.

[shoyer]

I think this is pretty much good to go. Merge at your discretion!

[clarkfitzg]

Ok, going to merge once the build passes. Will continue with the rest of the stuff next week. @shoyer should I rebase or squash these commits?

[shoyer]

I haven't been very careful about insisting on a clean git history so far, so don't worry about rebase/squash.