Unable to disable autoscaling in pyplot.plot

[neilpeterman]

Bug report

Bug summary

Arguments to disable autoscaling are not working, so despite e.g. scalex=False the horizontal axis is re-scaled.

Code for reproduction

import matplotlib.pyplot as plt

plt.plot([0, 2], [0, 2], '-')
plt.plot([0, 10], [1, 1], '-', scalex=False)

Actual outcome

Output in-line in jupyter notebook:

Expected outcome

Expected autoscaling only for the first call, so the xscale is just a bit padded from [0, 2]. This works as expected in matplotlib==3.1.3:

Matplotlib version

• Operating system: Ubuntu 18.04.3
• Matplotlib version: 3.3.0
• Matplotlib backend (print(matplotlib.get_backend())): module://ipykernel.pylab.backend_inline
• Python version: 3.6.8
• Jupyter version (if applicable): 1.0.0
• Other libraries: N/A

Tools installed with pip

[QuLogic]

This comes from 160de56, or #13593. This may be an expected change; see the API note from that PR.

cc @anntzer

[neilpeterman]

Thanks for the clarification @QuLogic -- so the way to achieve the previous behavior would be to do something like this:

plt.plot([0, 2], [0, 2], '-')
plt.gcf().canvas.draw()
plt.plot([0, 10], [1, 1], '-', scalex=False)

The concept of delaying autoscaling makes sense to me, although in the example above the meaning of scalex=False is not so intuitive to me -- I would think anything with scalex=False should be excluded from the autoscaling process in one way or another. Does that sound like a reasonable expectation?

I'm not sure -- it could be that I'm just accustomed to the previous behavior.

[anntzer]

I guess what you really want is more along the lines of #15595?
In the meantime we could trigger an autoscale when scalex/scaley goes from True to False, i.e. in _request_autoscale_view, something like if self._stale_viewlim_x and not scalex or self._stale_viewlim_y and not scaley: self._unstale_viewlim()? (untested, but hopefully you get the idea...)

[neilpeterman]

It seems like the fixes you bring up would address the issue in a reasonable way, triggering autoscaling when needed.

[igurin-invn]

Just a little poke: this bug is still open. Whether or not it results from an intentional change in #13593, it's not reflected in the documentation.

scalex, scaleybool, default: True
These parameters determine if the view limits are adapted to the data limits. The values are passed on to autoscale_view.

[anntzer]

Good point. I think scalex/scaley should be deprecated in favor of #15595 once something like that gets in, but in the meantime the following seems to work OKish, going back to the old semantics:

diff --git i/lib/matplotlib/axes/_axes.py w/lib/matplotlib/axes/_axes.py
index 5369eadbde..2b08537ef1 100644
--- i/lib/matplotlib/axes/_axes.py
+++ w/lib/matplotlib/axes/_axes.py
@@ -1719,6 +1719,8 @@ class Axes(_AxesBase):
         additionally use any  `matplotlib.colors` spec, e.g. full names
         (``'green'``) or hex strings (``'#008000'``).
         """
+        if not scalex or not scaley:
+            self._unstale_viewLim()
         kwargs = cbook.normalize_kwargs(kwargs, mlines.Line2D)
         lines = [*self._get_lines(*args, data=data, **kwargs)]
         for line in lines:

(note that using scalex/scaley=False then implies some runtime cost, as explained in 160de56)

[igurin-invn]

Is there a better workaround than this?

_xlim = ax.get_xlim()
ax.plot(x_big, y)
ax.set_xlim(_xlim)