Added the styler keyword to allow on-demand / callback to determine styling of geometries in the matplotlib interface.

[pelson]

Replaces #910. Closes #580.

[has2k1]

callbacks make for an API difference compared to MPL, is there any specific advantage for them.

For my use case, callbacks would not work because the decisions on the parameter values happen somewhere in an elaborate pipeline before the calls to ax.add_artist. So ax.add_geometries needs to handle parameters just like Matplotlib collection artists.

[pelson]

callbacks make for an API difference compared to MPL, is there any specific advantage for them.

It means we can style a feature that may not yet have all of the geometries to hand. For example, some features can return different geometries based on the zoom level.

For my use case, callbacks would not work because the decisions on the parameter values happen somewhere in an elaborate pipeline before the calls to ax.add_artist

That form would still work (untested):

geoms = [geom1, geom2, geom3, geom4, geom5, geom6, geom7]
facecolors = ['red', 'orange', 'yellow', 'green', 'blue', 'indigo', 'violet']  

def color_lookup(geom):
    index = geoms.index(geom)
    return {'facecolor': facecolors[index]}

ax.add_geometries(geoms, crs, styler=color_lookup)

If you really dislike that solution, I'd be supportive of creating a GeometryCollection object that follows the same interface as the PathCollection one. Namely:

ax.add_collection(GeometryCollection([geoms], facecolors=facecolors, transform=crs))

But I don't want to bleed in index information into Features because they are inherently limited to a fixed number of geometries.

[has2k1]

That solution would work, but indeed it is not friendly. If I had not been aware of the underlying issue, it would be mysterious. It requires a little too much context to fully understand. Also, it is probably of little consequence (as geometry lengths tend to be small) but that lookup is O(n^2), though it could be fixed with a 2nd lookup dictionary.

On the other hand, inspired by that code #910 could altered to take the index stuff out of the features, so that the fix is self contained, but having the styler and other parameters would make for a complicated API.

As such, a styler for ax.add_features and GeometryCollection for ax.add_collection is very good idea.

[dopplershift]

Looks fine to me. @pelson you still feel good about this one?

[pelson]

Looks fine to me. @pelson you still feel good about this one?

Sure do. If you are at hand, please feel free to merge at will. I'm going to build something on top of this change now, and when I'm done I'll merge (if you haven't beaten me to it) before opening a new PR.

[pelson]

Actually, given how old the branch is, building something on top of it is a pretty bad idea, so I'm going to merge this and move on.

[pelson]

Actually, given how old the branch is, building something on top of it is a pretty bad idea, so I'm going to merge this and move on.

🤣 - man, you're fast!

[dopplershift]

Had just opened the issue when you commented. 😁