Only register if cmap not in colormaps

[hoxbro]

This comes up sometimes when debugging with panel serve --autoreload and changing a file in the holoviews/hvplot repo.

This has only come up for me when debugging, not actual use.

Steps to reproduce:

1. Make a file called with import colorcet
2. panel serve --autoreload that file
3. Make some changes in colorcet/__init__.py and reload the page

ValueError: A colormap named "cet_diverging_isoluminant_cjm_75_c23" is already registered.

Traceback (most recent call last):
  File "/home/shh/miniconda3/envs/holoviz/lib/python3.10/site-packages/bokeh/application/handlers/code_runner.py", line 231, in run
    exec(self._code, module.__dict__)
  File "/home/shh/Development/holoviz/development/dev_hvplot/tmp.py", line 1, in <module>
    import colorcet  # noqa
  File "/home/shh/Repos/holoviz/colorcet/colorcet/__init__.py", line 551, in <module>
    m_diverging_isoluminant_cjm_75_c23 = mpl_cm('diverging_isoluminant_cjm_75_c23',diverging_isoluminant_cjm_75_c23)
  File "/home/shh/Repos/holoviz/colorcet/colorcet/__init__.py", line 92, in mpl_cm
    register_cmap("cet_"+name, cmap=cm[name])
  File "/home/shh/Repos/holoviz/colorcet/colorcet/__init__.py", line 70, in register_cmap
    colormaps.register(cmap, name=name)
  File "/home/shh/miniconda3/envs/holoviz/lib/python3.10/site-packages/matplotlib/cm.py", line 136, in register
    raise ValueError(
ValueError: A colormap named "cet_diverging_isoluminant_cjm_75_c23" is already registered.

[ianthomas23]

It might be worth adding a new test to confirm that a ValueError is raised as in the example.

[hoxbro]

Added unit test. The failing tests on Python 3.6 are unrelated.

[jbednar]

See #96 for issues with this approach. Can those be addressed?

[hoxbro]

I'm not sure I follow. Why is there an issue with this approach?

[jbednar]

See #96 (comment) . Skipping registering the map is dangerous unless you verify that the map is indeed the same, not just the same name.

[jbednar]

I'd be fine with skipping it if the name is the same and a Python equality test indicates that it's the same object, which should be true if it's just about reloading the same code. ETA: Looking again at the code, I now see a line that does check that it's the same object. In that case, this looks fine! Sorry for the noise!

[jbednar]

Python 3.6 tests were failing, which we'll have to consider when we do the next release. I assume that's to do with instantiating an environment rather than an actual problem, and I'm happy for this package to be installable on 3.6 even if there are aspects of the docs or other build steps that won't pass on 3.6.

[maximlt]

That package won't be installable on Python 3.6 after the next release #110

[jbednar]

Ah, yes. I guess it's fine; "colorcet" as a package should be available for Python 3.6, but the very latest versions need not be.