FIX: imshow arguments should be popped before passing upstream #2416
Conversation
|
Note that maybe we should add a warning to users that this keyword argument isn't being used rather than dropping it on the floor? Or, we could say "if regrid_shape is passed, don't take the fast-path"? |
98834c9 to
4e19b50
Compare
|
I'm trying to decide if there's any value in the regridding behavior when plotting in the original coordinates... |
We are thinking the same thing, hence the thrash in my PR pushing 😂 Would you prefer a warning and drop it and go through the fast-path then? Or this is currently not choosing the fast-path, but doesn't really do much. Although it is a much "larger" image technically and thus a larger file when saved to disk, but looks the same. |
lib/cartopy/mpl/geoaxes.py
Outdated
| @@ -1281,7 +1281,8 @@ def imshow(self, img, *args, **kwargs): | |||
| y0 - eps <= extent[3] <= y1 + eps)) | |||
|
|
|||
| if (transform is None or transform == self.transData or | |||
| same_projection and inside_bounds): | |||
| same_projection and inside_bounds and | |||
| kwargs.get("regrid_shape", None) is None): | |||
There was a problem hiding this comment.
This would forward on a manually-specified regrid_shape=None onto Matplotlib, no?
|
I think until we have an actual use case, we should keep the current semantics, avoid the error, and issue a warning. |
The regrid doesn't do anything, so we should warn the user that we are getting rid of it and continuing on.
4e19b50 to
b46c4d5
Compare
|
OK, I'm sold. I added a warning message to it now. |
We have extra keyword arguments in Cartopy that can't be passed upstream to Matplotlib, so remove them from the keyword arguments before going into any possible methods.
closes #2415