Fix #6773: Don't do image extent adjustment... #6776
Conversation
...on non-rectilinear axes
|
I don't know if I'm running this PR wrong, but it seems like the Cartopy result is the same here; though somehow Travis thinks differently. Will have to investigate later. |
| if (round_to_pixel_border and | ||
| t.is_affine and | ||
| # deliberate exact class match | ||
| type(self.axes) in (maxes.Axes, maxes._subplots.Subplot) and |
There was a problem hiding this comment.
Does this mess up anything with wcaxes?
There was a problem hiding this comment.
It might. We don't really have a good way to determine if the edges of the axes are going to be snapped or not, but I'm starting to think that just by specific class is not good enough.
|
OK, I've tested this one out now. It's not quite as expected, but it doesn't go over the edge now. But it seems like the snapping works in reverse now? Expected: |
|
@QuLogic: What is the axes class used in your example. This fix uses the class of the Axes to determine whether it has rectilinear edges or not. I suspect that isn't good enough, and maybe we'll have to do something smarter here. |
|
|
|
I'm changed the test to be on the Axes.patch, not the Axes class itself. @QuLogic: How does this work for your use case? |
|
This does fix the images in #6773, though it breaks Cartopy tests in the opposite way. That is, the image is now "too small", but I think it's actually more correct in that the image is no longer (ever so slightly) outside the expected border. The example above still produces the white line on the right, but now I'm not sure if it's a bug in |
|
I'm closing this because #7465 has been merged. |
...on non-rectilinear axes