Fixed categorical coloring of Contours in matplotlib #2259
Conversation
philippjfr
added
type: bug
|
I can't tell if this PR is ready yet, but at a glance it looks good. Happy to merge once you tell me it is ready and the tests are green. |
|
Can't be ready since the tests aren't green right :-) Locally it seems to pass but Travis isn't happy yet. Will try to figure this out today. |
Could have just been unlucky with transient failures you know... |
The unit test I added is failing. |
philippjfr
force-pushed
the
mpl_contours_categorical
branch
2 times, most recently
from
5f74236
to
cada665
Compare
philippjfr
force-pushed
the
mpl_contours_categorical
branch
from
cada665
to
921752a
Compare
| @@ -213,7 +213,7 @@ def values(cls, dataset, dim, expanded=True, flat=True): | |||
| if np.isscalar(values): | |||
| if not expanded: | |||
| return np.array([values]) | |||
| values = np.full(len(dataset), values) | |||
| values = np.full(len(dataset), values, dtype=np.array(values).dtype) | |||
There was a problem hiding this comment.
This is the most bizarre bug I've come across in a long time. Before I added this dtype approach this line would error (but only on Travis), given this input:np.full(10, 'B'), saying that:
ValueError: could not convert string to float: 'B'
But internally np.full does exactly the same thing I've now done here to fix it:
def full(shape, fill_value, dtype=None, order='C'):
if dtype is None:
dtype = array(fill_value).dtype
a = empty(shape, dtype, order)
multiarray.copyto(a, fill_value, casting='unsafe')
return aAbsolutely baffling, but it seems to work now.
|
Ready now. |
|
Weird 'fix'! Anyway, looks good. Merging. |
As the title says, Contours/Polygons in matplotlib did not handle categorical values for the
color_index.