plot_date() after axhline() doesn't rescale axes

If I plot an axhline first, and then two points with plot_date, the x-axis is not rescaled to the two new data points:

import matplotlib.pyplot as plt
from datetime import datetime

fig, axs = plt.subplots(2, 1)

axs[0].axhline(1.5)
axs[0].plot([datetime(2016, 1, 1, 0, 0, 0), datetime(2016, 1, 2, 0, 0, 0)], [1, 2])

axs[1].plot([datetime(2016, 1, 1, 0, 0, 0), datetime(2016, 1, 2, 0, 0, 0)], [1, 2])
axs[1].axhline(1.5)

plt.show()

I would expect the two plots to be the same, and the plotting order not to matter.

Present on current master (2.0.0rc2.post2912+g015d8d5) and on 1.5

I think I have a fix for this, after making the fix, the script above results in:

But I'm not sure how to write tests for it. Are they needed?

@Yasaman-Mah

@Yasaman-Mah
In this is case, I would write a test confirming that both axes have the expected limits:

import matplotlib.pyplot as plt
from datetime import datetime

dates = [datetime(2016, 1, 1, 0, 0, 0), datetime(2016, 1, 2, 0, 0, 0)]
values = [1, 2]
expected_limits = (xmin, xmax) # fill this in

fig, (ax1, ax2) = plt.subplots(2, 1)
ax1.axhline(1.5)
ax1.plot(dates, values)

ax2.plot(dates, values)
ax2.axhline(1.5)

assert ax1.get_xlim() == expected_limits
assert ax1.get_xlim() == ax2.get_xlim()

I think it'd be fine to place this in matplotlib/lib/tests/test_axes.py

@Yasaman-Mah

@Yasaman-Mah Looking forward to seeing your PR!

fix matplotlib#7742

@Yasaman-Mah

@Yasaman-Mah
Plotting again after axhline causes the x-axis to go only up to 2016-01-10 instead of expected 2016-01-30

axs[1].plot([datetime(2016, 1, 1, 0, 0, 0), datetime(2016, 1, 30, 0, 0, 0)], [1, 2])
axs[1].axhline(1.5)
axs[1].plot([datetime(2016, 1, 1, 0, 0, 0), datetime(2016, 1, 10, 0, 0, 0)], [0, 5])

Fixes matplotlib#7742 - axhline rescale issue

@Shang-Jia

I also encountered what @Shang-Jia said and I have submitted a PR #8210 that fixes all issues discussed.

@Shang-Jia

@Shang-Jia
I understand that you are in UTSC, I just wanted to remind you that we are all doing this as a part of course work, and it would be against course (and possibly university's) policies to look at another student's code before submission deadline. Your comment above indicates that this is what you did.

To matplotlib developers: Thank you for all the feedback. Unfortunately, to me it seems that some of my classmates are trying to take credit for most of my fix by making minor improvements to it. In fact, I do have a better solution in mind, which I will start working on once the situation above is resolved.

Merge pull request #3 from sughandj/issue7742_axhline_limits

Merge pull request #4 from ibnIrshad/fix7742

In [3]: matplotlib.__version__
Out[3]: '3.3.0rc1.post273+gb803891a4b'

It looks like we have code in master that is designed to handle this (autoscaling with the mixed transform should be leaving x limits scaling alone), but the original example now yields a traceback:

TypeError                                 Traceback (most recent call last)
<ipython-input-1-01cdab9ed12b> in <module>
      5
      6 axs[0].axhline(1.5)
----> 7 axs[0].plot([datetime(2016, 1, 1, 0, 0, 0), datetime(2016, 1, 2, 0, 0, 0)], [1, 2])
      8
      9 axs[1].plot([datetime(2016, 1, 1, 0, 0, 0), datetime(2016, 1, 2, 0, 0, 0)], [1, 2])

~/work/programs/py/mpl/matplotlib/lib/matplotlib/axes/_axes.py in plot(self, scalex, scaley, data, *args, **kwargs)
   1743         lines = [*self._get_lines(*args, data=data, **kwargs)]
   1744         for line in lines:
-> 1745             self.add_line(line)
   1746         self._request_autoscale_view(scalex=scalex, scaley=scaley)
   1747         return lines

~/work/programs/py/mpl/matplotlib/lib/matplotlib/axes/_base.py in add_line(self, line)
   1975             line.set_clip_path(self.patch)
   1976
-> 1977         self._update_line_limits(line)
   1978         if not line.get_label():
   1979             line.set_label('_line%d' % len(self.lines))

~/work/programs/py/mpl/matplotlib/lib/matplotlib/axes/_base.py in _update_line_limits(self, line)
   1997         Figures out the data limit of the given line, updating self.dataLim.
   1998         """
-> 1999         path = line.get_path()
   2000         if path.vertices.size == 0:
   2001             return

~/work/programs/py/mpl/matplotlib/lib/matplotlib/lines.py in get_path(self)
   1009         """
   1010         if self._invalidy or self._invalidx:
-> 1011             self.recache()
   1012         return self._path
   1013

~/work/programs/py/mpl/matplotlib/lib/matplotlib/lines.py in recache(self, always)
    651         if always or self._invalidx:
    652             xconv = self.convert_xunits(self._xorig)
--> 653             x = _to_unmasked_float_array(xconv).ravel()
    654         else:
    655             x = self._x

~/work/programs/py/mpl/matplotlib/lib/matplotlib/cbook/__init__.py in _to_unmasked_float_array(x)
   1296         return np.ma.asarray(x, float).filled(np.nan)
   1297     else:
-> 1298         return np.asarray(x, float)
   1299
   1300
~/miniconda3/envs/mpl1/lib/python3.7/site-packages/numpy/core/_asarray.py in asarray(a, dtype, order)
     83
     84     """
---> 85     return array(a, dtype, copy=False, order=order)
     86
     87

TypeError: float() argument must be a string or a number, not 'datetime.datetime'

@efiring

@efiring, that's because on master the date unit converter is not being registered. If you merge #17869 that will be fixed 😉

(ahem, or if you register the date converter manually)

Merged and verified: the bug reported here still exists.

This isn't just dates, this is all units. Basically if the units change on an axes (as they do here from no units to date units) axis.update_info is called, before the new plot artists are added. This calls axis.set_default_interval, and the dataLim is set to the default for the converter, which for dates is 2000-2010.

This doesn't happen if the new plot has no units. the xlims correctly remain -np.Inf, +np.Inf and the relimit works as expected.

I don't know why axis.set_default_intervals is called. And this is the only place in the code where it is called. So I think we should just not call it here either.

Thanks to everyone who worked on fixing this! I remember opening this in the early days of my contributions to Matplotlib, so great to finally see it fixed!

Navigation Menu

Repository navigation

plot_date() after axhline() doesn't rescale axes #7742

9 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

plot_date() after axhline() doesn't rescale axes #7742

Description

Activity

Yasaman-Mah commented on Feb 28, 2017

phobson commented on Feb 28, 2017

tacaswell commented on Mar 1, 2017

Shang-Jia commented on Mar 6, 2017

sughandj commented on Mar 7, 2017

Yasaman-Mah commented on Mar 8, 2017

9 remaining items

efiring commented on Jul 14, 2020

jklymak commented on Jul 14, 2020

jklymak commented on Jul 14, 2020

efiring commented on Jul 15, 2020

jklymak commented on May 6, 2021

dstansby commented on Jun 1, 2021

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions