Repository navigation
plt.plot does not plot anything with named arguments #12106
Description
Bug report
Bug summary
When using named arguments for x and y, the plots created with plt.plot respectively ax.plot are keeping empty. plt.errorbar (or ax.errorbar) are producing the desired output (see code below).
The documentation shows that x and y are valid named arguments.
Code for reproduction
import matplotlib.pyplot as plt
import numpy as np
x = np.arange(0, 12, .5)
y = np.random.rand(*x.shape)
plt.plot(x, y) # works as expected
plt.plot(x=x, y=y) # shows empty plot
plt.errorbar(x=x, y=y) # works as expectedActual outcome
Expected outcome
or
Matplotlib version
- Operating system: Windows 10x64
- Matplotlib version: 2.2.3 (had the same issue with 2.2.2)
- Matplotlib backend (
print(matplotlib.get_backend())): 'module://ipykernel.pylab.backend_inline' - Python version: 3.6.2
- Jupyter version (if applicable): 1.0.0 (Jupyter notebook 5.1.0rrc1)
- Other libraries: numpy 1.14.5+mkl
Python was installed using WinPython, packages installed with pip.
Activity
anntzer commentedon Sep 13, 2018
I am not particularly sure that we want to support passing x and y as kwargs, see recent discussion on a similar topic at #11526.
(But if we decide we indeed don't, then we should make the documentation clear and error out when they are passed in as kwargs.)
taxus13 commentedon Sep 13, 2018
Thank you for your reply!
A warning message and a clear documentation would definitely help.
Another suggestion would be a 'keyword-only' mode for the plot function (and other affected methods, today I stumbled upon the same issue with
loglog): Ifxand/oryare passed as kwargs, no positional arguments are allowed. Maybe this could be implemented as a layer around the current plot function, but I am sure if this breaks compatibility.phobson commentedon Sep 13, 2018
@taxus13 for an immediate work-around that's not writing your own function, seaborn might have you covered:
http://seaborn.pydata.org/tutorial/relational.html#relational-tutorial
anntzer commentedon Sep 13, 2018
fwiw I'm also against fully kw-only mode (also as discussed in the linked thread).
jklymak commentedon Sep 13, 2018
I think that
plt.errorbar(x=x, y=y)works is more accidental than anything.anntzer commentedon Sep 13, 2018
Well, errorbar doesn't have a completely
stupidweird signature...jklymak commentedon Sep 13, 2018
I guess it is a bit weird that this just doesn't plot anything, rather than give some sort of error.
anntzer commentedon Sep 13, 2018
I agree it should error out.
rudrathegreat commentedon Nov 8, 2018
If you are using a subplots, then you might want to consider the following -
I hope that helped!
anntzer commentedon Jan 3, 2019
Actually a duplicate of #4059.