Repository navigation
Passing an incorrectly sized colour list to scatter should raise a relevant error #11373
Description
Bug report
Bug summary
When calling a scatter with a list for the c argument, which does not match the length of the data, an AttributeError: 'list' object has no attribute 'shape' is raised instead of informing the user about the non-matching length issue.
#7363 tried hard to make error message clearer but forgot to consider that input may be any sequence, not necessarily a numpy array.
A similar issue was raised in #9209.
Code for reproduction
import matplotlib.pyplot as plt x = [1,2,3,4,5] c = [1,2,3,4,5,6] # one color more than values plt.scatter(x,x, c=c) plt.show()
Actual outcome
An error
File "...\lib\site-packages\matplotlib\axes\_axes.py", line 4279, in scatter
.format(c.shape, x.size, y.size))
AttributeError: 'list' object has no attribute 'shape'
This error tells nothing about the true cause of the problem, but is rather an error caused by the error message itself.
Expected outcome
An error similar to the intended one
matplotlib/lib/matplotlib/axes/_axes.py
Lines 3946 to 3950 in 6ec80ea
but for a list, which does not have a shape attribute.
Note that I rather raised this issue, instead of directly fixing it, because I'm not sure if there should be some typechecking performed within the except clause or whether to add another try inside the except or - because both feels kind of strange - someone has an even better idea.
Matplotlib version
- Operating system: Windows 8.1
- Matplotlib version: 2.2.2
- Matplotlib backend: any
- Python version: 3.6
Activity
tacaswell commented on Jun 5, 2018
I also ran into this (but did not take the time to track it down) yesterday.
Running through np.asanyarray is also not great, but I think would be better than explicit type checking. We could also just put the repr of 'c' in the error message but that could get needlessly verbose.
mwaskom commented on Jun 5, 2018
Easiest fix would be doing np.shape(c) instead of c.shape, right?
ImportanceOfBeingErnest commented on Jun 5, 2018
Maybe it's worth mentioning that the same error is also shown when a nonexisting color is specified, i.e.
results in AttributeError: 'str' object has no attribute 'shape'.
When using np.shape(c) instead of c.shape, the error in the above case would be
which is of course better than what is currently raised, but doesn't really tell the user that the color they used does not exist.
ImportanceOfBeingErnest commented on Jul 25, 2018
Given that #11383 is merged, this issue can be closed.