Summary
In #24545 it was identified that sometimes there is a problem with masked arrays and units. The solution in #24732 is rather straightforward, but it lead me to check the code for potential other locations.
Proposed fix
These are the locations I potentially found. It should probably be confirmed that it is a problem before modifying them (and those cases will be a good base for a test):
|
self._offsets = np.column_stack( |
|
(np.asarray(self.convert_xunits(offsets[:, 0]), float), |
|
np.asarray(self.convert_yunits(offsets[:, 1]), float))) |
(Here one should probably take a complete look at the
set_offsets/
get_offsets setup and see if one can remove redundant calls. See a bit of the discussions in
#24732)
|
offsets = np.column_stack([xs, ys]) |
(Not sure if it makes sense here.)
A bunch in quiver, but not sure how well that supports units and masks. Similar with projections.
As discussed in #24732 (comment) it seems like date2num does not support masked arrays, so making it do that is also a step in the right direction for this. (This is dealt with in #24734)
Summary
In #24545 it was identified that sometimes there is a problem with masked arrays and units. The solution in #24732 is rather straightforward, but it lead me to check the code for potential other locations.
Proposed fix
These are the locations I potentially found. It should probably be confirmed that it is a problem before modifying them (and those cases will be a good base for a test):
matplotlib/lib/matplotlib/collections.py
Lines 548 to 550 in cc85fb6
(Here one should probably take a complete look at the
set_offsets/get_offsetssetup and see if one can remove redundant calls. See a bit of the discussions in #24732)matplotlib/lib/matplotlib/collections.py
Line 2099 in cc85fb6
(Not sure if it makes sense here.)
A bunch in quiver, but not sure how well that supports units and masks. Similar with projections.
As discussed in #24732 (comment) it seems like
date2numdoes not support masked arrays, so making it do that is also a step in the right direction for this. (This is dealt with in #24734)