Repository navigation
[Bug]: to_rgba_array("none") returns empty array #28475
Description
Bug summary
mcolors.to_rgba_array("none") returns an empty array while mcolors.to_rgba("none") and mcolors.to_rgba_array(["none"]) return (0,0,0,0) (which is the documented value of None) and this leads to inconsistencies when trying to use mcolors.same_color
Actual outcome
>>> import matplotlib.colors as mcolors
>>> mcolors.to_rgba("none")
(0.0, 0.0, 0.0, 0.0)
>>> mcolors.to_rgba_array("none")
array([], shape=(0, 4), dtype=float64)
>>> mcolors.to_rgba_array(["none"])
array([[0., 0., 0., 0.]])
>>> mcolors.same_color("none", mcolors.to_rgba("none"))
False
>>> mcolors.same_color(["none"], mcolors.to_rgba("none"))
TrueExpected outcome
>>> import matplotlib.colors as mcolors
>>> mcolors.to_rgba("none")
(0.0, 0.0, 0.0, 0.0)
>>> mcolors.to_rgba_array("none")
array([[0., 0., 0., 0.]])
>>> mcolors.to_rgba_array(["none"])
array([[0., 0., 0., 0.]])
>>> mcolors.same_color("none", mcolors.to_rgba("none"))
True
>>> mcolors.same_color(["none"], mcolors.to_rgba("none"))
TrueAdditional information
I tried the quick fixes of changing the return here to array([[0., 0., 0., 0.]]) and np.array([to_rgba('none')], float)
matplotlib/lib/matplotlib/colors.py
Lines 482 to 483 in d347c32
That yielded a ton of test failures, mostly around colorbars/contours, hatches, and but roughly the diffs all looked something like this:
contourf_extend_patches (actual, expected)
colorbar_two_slope
Also I realize this could be not worth fixing and a quick fix is to cast none to a list in same_color
Activity
story645 commentedon Aug 12, 2024
Someone at scipy tracked this down to being an issue w/ eager conversion to color that relies on 'none'/empty array for if statements to preserve the correct alpha values. xref #28710
rcomer commentedon Mar 9, 2025
The existing behaviour is clearly deliberate but I agree with @story645 about what I expected to happen. Is it worth adding a boolean parameter so you could choose between the two behaviours?
story645 commentedon Mar 9, 2025
I've been thinking of a sort of true constant for none, but flagging would probably do the same trick - what I'm thinking is that it could be really useful to know which parts of the codebase need the empty behavior.