Make zdir key-word only for 3D-scatter and -bar #27538
Conversation
oscargus
changed the title
|
It seems okay, other than the missing tests? |
Missing tests as in tests to see that it actually warns or tests for |
| @@ -2653,7 +2653,7 @@ def scatter(self, xs, ys, zs=0, zdir='z', s=20, c=None, depthshade=True, | |||
| if np.may_share_memory(zs_orig, zs): # Avoid unnecessary copies. | |||
| zs = zs.copy() | |||
|
|
|||
| patches = super().scatter(xs, ys, s=s, c=c, *args, **kwargs) | |||
There was a problem hiding this comment.
Here and below at super.bar(), are we sure that the super class method can't use any *args? Don't want to break any plots that rely on it if so.
There was a problem hiding this comment.
In theory yes: https://matplotlib.org/devdocs/api/_as_gen/matplotlib.axes.Axes.scatter.html#matplotlib.axes.Axes.scatter
but very unlikely in practice. One would have to pass all 3D positional parameters
scatter(self, xs, ys, zs, zdir, s, c, depthshade, ...)
and then continue with the non-overlapping 2D positional parameters, minimally:
scatter(self, xs, ys, zs, zdir, s, c, depthshade, marker)
I would be very surprised if somebody took the effort to figure that out.
I'm inclined to allow tolerate minimal breakage here.
PR summary
Related to #27287 and #27368. Putting it here to see the consequences on examples etc and to discuss if it makes sense. Will naturally add release notes.
(As far as I can see,
zdiris not tested at all forscatter.baronly has a single test which passes zdir as a kwarg.)PR checklist