DOC/API : color map change documentation #4238
Conversation
This is the first pass at documenting the decisions on how to change the default color map/cycles. The idea is to explain the reasoning in the documentation and to provide examples of all of the considered color maps. This will hopefully be a useful document both for posterity but for making the decision.
|
|
||
| - it should be a sequential colormap, because diverging colormaps are | ||
| really misleading unless you know where the "center" of the data is, | ||
| and for a default colormap we generally won't. |
There was a problem hiding this comment.
imho for signed data, a diverging colormap centered at 0 is a good default choice. But I understand that behavior might be too magical for some.
There was a problem hiding this comment.
for me, it ain't just about the 0, but also the data around 0. I have in some of my plots for simplicity sake forced the colormap to go between -A and +A where A = np.max(np.abs(data)), but then if the data say goes from -10 to +100, then it comes a bit more complicated so as to avoid the -10 looking quite faint...
This, I think, comes under the realm of the Normalize class, perhaps we need a few more normalisation classes and think about which one to use as default...
There was a problem hiding this comment.
There is #3858 which adds a normalizer for dealing with that situation.
Picking anything about the diverging color map (if to use it, where to center it, if the upper/lower limits should be symmetric) requires knowing something about the meaning of the data which is a business mpl should not be in (as it is too widely used). This sort of domain-specific defaults should be the job of down-stream packages (ex scikit-image, pandas, seaborn, ...)
|
I've added one comment, but otherwise I think this is an excellent document! |
|
|
||
| As discussed at length elsewhere [insert links], ``jet`` is an | ||
| empirically bad color map and should not be the default color map. | ||
| Due to the position that changing the appearance of plot is a back |
There was a problem hiding this comment.
maybe just say "Due to the position that changing the appearance of the plot breaks backward compatibility, ..."
|
👍 Awesome. |
|
Know what I would love to see? A side-by-side comparison of all the On Wed, Mar 18, 2015 at 11:03 AM, Michael Droettboom <
|
|
I would like to see this merged as soon as possible so that we can iterate on the text via PR in a way that does not make me the bottle neck and all of the PR issues live under the matplotlib organizations fork of mpl not mine. |
|
no problem. once travis passes, I'll merge it in. |
DOC/API : color map change documentation
This is the first pass at documenting the decisions on how to
change the default color map/cycles.
The idea is to explain the reasoning in the documentation and to provide
examples of all of the considered color maps. This will hopefully be a
useful document both for posterity but for making the decision.