Closed
Description
There is a "communication" issue between the axes.locator_params and the LogLocator class:
plt_scale = 'log'
#plt_scale = 'linear'
fig, (ax1) = plt.subplots(1, 1)
ax1.set(yscale=plt_scale)
# works for linear scale, not for log:
ax1.axes.locator_params('y', nbins=5)Result:
AttributeError Traceback (most recent call last)
<ipython-input-74-3492af5f93c8> in <module>()
7
8 # works for linear scale, not for log:
----> 9 ax1.axes.locator_params('y', nbins=5)
/home/pierre/anaconda/lib/python2.7/site-packages/matplotlib/axes.pyc in locator_params(self, axis, tight, **kwargs)
2305 self.xaxis.get_major_locator().set_params(**kwargs)
2306 if _y:
-> 2307 self.yaxis.get_major_locator().set_params(**kwargs)
2308 self.autoscale_view(tight=tight, scalex=_x, scaley=_y)
2309
AttributeError: 'LogLocator' object has no attribute 'set_params'
What I get from the traceback is that locator_params method forwards the locator attributes (in my case nbins) to the actual locator object via its set_params method. And this method is missing from LogLocator. (set_params is indeed defined in MaxNLocator https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/ticker.py#L1297)
So my question is: shouldn't all Locator subclasses have a set_params method ?
(and then, my code example could raise a different error, or maybe a warning, complaining that LogLocator makes in fact no use of an nbins parameter!)
best,
Pierre