Fix scatter singlecolor #17499
Fix scatter singlecolor #17499
Conversation
Encourage user to use the *color* keyword in the case where they have only one color.
| @@ -4281,19 +4281,26 @@ def invalid_shape_exception(csize, xsize): | |||
| except ValueError: | |||
| pass # Failed to convert to float array; must be color specs. | |||
| else: | |||
| # handle the documented special case of a 2D array with 1 | |||
| # row which as RGB(A) to broadcast. | |||
| if c.shape == (1, 4) or c.shape == (1, 3): | |||
There was a problem hiding this comment.
OK, but do we want a 2-d array or a list of arrays as the special case?
There was a problem hiding this comment.
I don't understand. Would you elaborate, please? Why would there be a distinction here?
There was a problem hiding this comment.
The particular value for c is something like: c = [[0.1, 0.1, 0.4]]?
There was a problem hiding this comment.
We do c = np.asanyarray(c, dtype=float) a few lines above to normalize c. I do not think we want to get into the business of distinguishing between [[a, b, c]], [np.array([a, b, c])], and np.array([[a, b, c]]) as input.
There was a problem hiding this comment.
OK, but an unsqueezed 1x3 ndarray is a perfectly valid thing to pass for c if I have 3 data points. So how can we tell them apart?
There was a problem hiding this comment.
Before it is squeezed, the dimensions are checked so that a 1x3 or 1x4 can be interpreted as a color. The (3,) or (4,) 1-d cases, when the size matches the number of points, is interpreted as a sequence of values to be mapped. This is the documented way the ambiguity is resolved, based on a decision made long ago. I just fouled it up in the last change I made to the code. Thomas is cleaning up my mess.
|
Owee, I'm MrMeeseeks, Look at me. There seem to be a conflict, please backport manually. Here are approximate instructions:
And apply the correct labels and milestones. Congratulation you did some good work ! Hopefully your backport PR will be tested by the continuous integration and merged soon! If these instruction are inaccurate, feel free to suggest an improvement. |
Merge pull request matplotlib#17499 from tacaswell/fix_scatter_singlecolor Fix scatter singlecolor Conflicts: lib/matplotlib/axes/_axes.py - implicitly backport some extra docstring changes
…-v3.2.x Backport PR #17499: Fix scatter singlecolor
PR Summary
Restores support of the color-favoring special case documented in the docstring and warning. Independent of if we think this was a good design initially etc we pretty clearly document that this is the expected behavior and it was the behavior in 3.1 so it should be restored.
Also tweaked the docstring to direct users to the color kwarg which does not have the "should we colormap?" ambiguity.
Closes #17423
PR Checklist