Skip to content

Proplot colormaps cause issues with default xarray plotting #2

Closed

Description

@bradyrx

Quick plots with xarray (e.g., ds.plot()) use 'viridis' as the default sequential colormap and 'RdBu' as the default diverging colormap.

Thus, if one is to use ds.plot() with negative and positive values, an error is thrown because 'RdBu' doesn't exist once proplot is imported. For whatever reason, it seems like the 'RdBu' convention from matplotlib (https://matplotlib.org/examples/color/colormaps_reference.html) was switched to 'BuRd' in proplot.

Some other cases here are 'RdYlBu' from matplotlib becomes 'BuYlRd'. I imagine reordering of these strings should be avoided so that errors like this aren't thrown from outside packages. Perhaps simple aliases can be made to take either BuRd or RdBu.

Activity

  1. lukelbd commented on Jan 9, 2019

    @lukelbd
    Collaborator

    Thanks for the heads up, have never used the built-in xarray plotting commands.

    Indeed I switched the names for some default colormaps because I thought 'RdBu' was a silly convention (99.9% of the time blue is "small/negative" and red is "large/positive"). Compatibility breaks didn't occur to me. I think I'll replace the colormap dictionary (stored in matplotlib.cm.cmap_d) with a new class that handles these naming issues -- total case insensitivity, and e.g. equivalency of 'RdBu_r' and 'BuRd'. Shouldn't take very long.

  2. bradyrx commented on Jan 9, 2019

    @bradyrx
    CollaboratorAuthor

    I completely agree with your thinking (and understood why you did it when I saw it). I imagine this compatibility issue would extend to any library that default plots a certain colormap like xarray.

  3. lukelbd commented on Jan 9, 2019

    @lukelbd
    Collaborator

    Decided to just tackle this right now since I'd been meaning to do something similar anyway. Latest commit should satisfy the behavior I described (hopefully) without any side-effects, using a (probably overkill) fancy new class called _CmapDict. See:

    a6af853

    Try this example (after updating to the latest version):

    f, axs = plot.subplots(ncols=6, axwidth=2, innercolorbars='b', innercolorbars_kw={'hspace':0.2})
    for i,cmap in enumerate(('ColdHot', 'HotCold', 'HotCold_r', 'RdYlGn', 'GnYlRd', 'GnYlRd_r')):
        m = axs[i].contourf(np.random.rand(10,10), cmap=cmap)
        axs[i].bottompanel.colorbar(m)

    Let me know if you discover any new issues and will re-open.

  4. lukelbd commented on Jan 9, 2019

    @lukelbd
    Collaborator

    (Accidentally left a print statement in there; latest commit will have deleted it.)

  5. bradyrx commented on Jan 10, 2019

    @bradyrx
    CollaboratorAuthor

    @lukelbd, this doesn't seem to solve it on my end. Note that I did upgrade the package to the most recent commits.

    Using your example I get the following error:

    TypeError                                 Traceback (most recent call last)
    <ipython-input-27-9b7beea59934> in <module>
    ----> 1 f, axs = plot.subplots(ncols=6, axwidth=2, innercolorbars='b', innercolorbars_kw={'hspace':0.2})
          2 for i,cmap in enumerate(('ColdHot', 'HotCold', 'HotCold_r', 'RdYlGn', 'GnYlRd', 'GnYlRd_r')):
          3     m = axs[i].contourf(np.random.rand(10,10), cmap=cmap)
          4     axs[i].bottompanel.colorbar(m)
    
    ~/Desktop/proplot/proplot/subplots.py in subplots(array, ncols, nrows, rowmajor, emptycols, emptyrows, tight, auto_adjust, rcreset, silent, span, share, spanx, spany, sharex, sharey, innerpanels, innercolorbars, innerpanels_kw, basemap, proj, projection, proj_kw, projection_kw, **kwargs)
        262 
        263     # Create gridspec for outer plotting regions (divides 'main area' from side panels)
    --> 264     figsize, offset, subplots_kw, gridspec_kw = _gridspec_kwargs(nrows, ncols, **kwargs)
        265     row_offset, col_offset = offset
        266     gs = FlexibleGridSpec(**gridspec_kw)
    
    TypeError: _gridspec_kwargs() got an unexpected keyword argument 'innercolorbars_kw'
    

    Running ds.plot() returns the same issue as before. Interestingly, I tried to run import proplot.colortools to just check that the new Class was showing up, and get the following error:

    ---------------------------------------------------------------------------
    ModuleNotFoundError                       Traceback (most recent call last)
    <ipython-input-30-bba849ec6737> in <module>
    ----> 1 import proplot.colortools
    
    ModuleNotFoundError: No module named 'proplot.colortools'
    

    This happens with all submodules. Not sure if you have some lock on importing submodules directly? Although not sure how one would do that. Note that import proplot works just fine, so this is isolated to submodules.

  6. lukelbd commented on Jan 10, 2019

    @lukelbd
    Collaborator

    Hmm, innercolorbars_kw (which is just an alias for innerpanels_kw) should have been added by latest commit.

    The second example doesn't work because I just use the individual files to organize things (__init__.py contains from .colortools import *; from .axistools import *; etc.. Wanted user to have access to all commands at the top-level (without submodules).

    Try accessing plot._CmapDict; does it show anything? If not, maybe try uninstalling/re-installing. I pushed the changes.

    I wonder if when you pip install from a git URL, if the version number in setup.py hasn't changed (which it hasn't), pip just won't update the repo?

  7. bradyrx commented on Jan 10, 2019

    @bradyrx
    CollaboratorAuthor

    It works now... can't tell you what changed. I installed via pip and your example as well as ds.plot() on anomalies works.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions