Description
Bug summary
When plotting on a parasite axis using axes_grid1.parasite_axes, units are not automatically applied to the parasite axis.
Code for reproduction
#!/usr/bin/env python3
from mpl_toolkits.axes_grid1 import host_subplot
from mpl_toolkits import axisartist
import matplotlib.pyplot as plt
import matplotlib.units as units
import matplotlib.ticker as ticker
class Unit:
def __init__(self, val):
self._val = val
class Volt(Unit):
fmt = "%0.1f V"
class Amp(Unit):
fmt = "%0.1f A"
class UnitConverter(units.ConversionInterface):
@staticmethod
def convert(value, unit, axis):
return [x._val for x in value]
@staticmethod
def axisinfo(unit, axis):
return units.AxisInfo(majfmt=ticker.FormatStrFormatter(unit.fmt))
@staticmethod
def default_units(x, axis):
return x[0].__class__
units.registry[Volt] = UnitConverter()
units.registry[Amp] = UnitConverter()
host = host_subplot(111, axes_class=axisartist.Axes)
p1, = host.plot([0, 1, 2], [Volt(x) for x in (0, 1, 2)])
par1 = host.twinx()
par1.axis["right"].major_ticklabels.set_visible(True)
p2, = par1.plot([0, 1, 2], [Amp(x) for x in (0, 3, 2)])
plt.show()Actual outcome
Expected outcome
Additional information
As far as I can tell, this is because ParasiteAxesBase.cla contains this line:
self._get_lines = self._parent_axes._get_linesSince _get_lines contains a reference to its axes instance, this causes ax2.plot to attempt to call update_units on the host axes instead of the parasite axes. Removing this line appears to fix unit behavior for me, but I don't know why the line was there in the first place because it has been there since the very first commit of parasite_axes. Perhaps the goal was to make the axes share a color cycler?
I was able to preserve that behavior while fixing unit support by changing the line to
self._get_lines = functools.partial(self._parent_axes._get_lines, axes=self)and then changing _process_plot_var_args.__call__, _process_plot_var_args._makefill, and _process_plot_var_args._plot_args to use kwargs.get("axes", self.axes) instead of self.axes.
Operating system
OS X
Matplotlib Version
3.5.1
Matplotlib Backend
MacOSX
Python version
Python 3.10.1
Jupyter version
n/a
Installation
Linux package manager