Repository navigation
Creating colorbar without artist fails with LogNorm #13967
Description
#13171 documented how to create a colobar that is not associated with an artist. This recipe fails when the colorbar norm is a LogNorm however:
import matplotlib.pyplot as plt
import matplotlib.cm as cm
import matplotlib.colors as mcolor
fig, ax = plt.subplots()
sm = cm.ScalarMappable(norm=mcolor.LogNorm(), cmap='viridis')
fig.colorbar(sm)
plt.draw()gives
Traceback (most recent call last):
File "test.py", line 7, in <module>
fig.colorbar(sm)
File "/Users/dstansby/github/matplotlib/lib/matplotlib/figure.py", line 2208, in colorbar
cb = cbar.colorbar_factory(cax, mappable, **cb_kw)
File "/Users/dstansby/github/matplotlib/lib/matplotlib/colorbar.py", line 1641, in colorbar_factory
cb = Colorbar(cax, mappable, **kwargs)
File "/Users/dstansby/github/matplotlib/lib/matplotlib/colorbar.py", line 1183, in __init__
ColorbarBase.__init__(self, ax, **kw)
File "/Users/dstansby/github/matplotlib/lib/matplotlib/colorbar.py", line 460, in __init__
self.draw_all()
File "/Users/dstansby/github/matplotlib/lib/matplotlib/colorbar.py", line 483, in draw_all
self._process_values()
File "/Users/dstansby/github/matplotlib/lib/matplotlib/colorbar.py", line 932, in _process_values
b = self.norm.inverse(self._uniform_y(self.cmap.N + 1))
File "/Users/dstansby/github/matplotlib/lib/matplotlib/colors.py", line 1085, in inverse
return vmin * np.ma.power((vmax / vmin), val)
ZeroDivisionError: division by zeroChanging LogNorm to Normalize makes the example work fine again, as intended.
Activity
anntzer commentedon Apr 15, 2019
You need to pass in the norm bounds (LogNorm(vmin=..., vmax=...)), otherwise they default to 0..1 -> log(0) -> kaboom (note that unlike "standard" colorbars there's no data to autoscale the norm to).
I don't think this is a bug, but this could be better documented, perhaps.
dstansby commentedon Apr 15, 2019
Maybe it would be possible to default to say 0.1, 1 for
LogNorminstead?anntzer commentedon Apr 15, 2019
Probably rather 1..10 (10^(0..1)) but other than that, why not.
anntzer commentedon Apr 16, 2019
Ah, but the problem is not on LogNorm's side; it's colorbar() which tries to set the default values (in https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/colorbar.py#L923).
I actually think it would be fair to say, if the norm is not normalized (so it's not associated to a real mappable with data), then we just throw here.
Requesting the user to do
ScalarMappable(Normalize(42, 123))orScalarMappable(LogNorm(1, 10))instead of trying to guess what extents he wants seems fair...dstansby commentedon May 6, 2019
I think that's a fair expectation, closing since there is now a more sensible error message.