Fix axh{line,span} on polar axes. #26788
Conversation
|
Since it's not in the test, just wondering what happens when you pass |
|
Tests should hopefully be fixed now, and I detailed the changes a bit more in the behavior changes note -- I can add a separate whatsnew if you think it's needed. Limiting the "other" direction in spans produces the expected wedges, now included in the test. |
|
I think a what's new note is not required, but would be nice to highlight that this now works. |
|
Added one. |
| ... now draw circles and circular arcs (`~.axhline`) or full or partial annuli | ||
| (`.axhspan`). |
There was a problem hiding this comment.
Note that the references in this file are not resolved by sphinx (it finds both the Axes method and pyplot), and are causing the doc build to fail
| ... now draw circles and circular arcs (`~.axhline`) or full or partial annuli | |
| (`.axhspan`). | |
| ... now draw circles and circular arcs (`~.Axes.axhline`) or full or partial annuli | |
| (`.Axes.axhspan`). |
| # For Rectangles and non-separable transforms, add_patch can be buggy | ||
| # and update the x limits even though it shouldn't do so for an | ||
| # yaxis_transformed patch, so undo that update. |
There was a problem hiding this comment.
There is some cutout in _update_patch_limits for x/y axis transforms; is that something that should be changed instead?
There was a problem hiding this comment.
I'll look into that.
There was a problem hiding this comment.
The problem is that Rectangle.get_transform() is not equal to get_xaxis/yaxis_transform, because Rectangle is implemented as a unit rectangle plus a sizing and translation affine transform, which ends up being included (first!) in Rectangle.get_transform. I guess ideally we'd want to check something like (-patch.get_patch_transform() + patch.get_transform()) (the subtraction needs to be "on the left"), but the transform system can't even handle that, so for now I don't think that would be so easy.
For axhline, set the underlying path's _interpolation_steps to
GRIDLINE_INTERPOLATION_STEPS (180), which amounts to using the same
logic to draw it as for drawing gridlines. This ensures that a polar
axhline goes (due to interpolation) around the whole circle, rather than
being a trivial line connecting a point to itself. Also update axvline
for consistency. (Note that _interpolation_steps has is ignored for
rectilinear transforms and thus the change doesn't slow down the common,
rectilinear case.)
Increase the number of interpolation steps of ax{v,h}span likewise to
GRIDLINE_INTERPOLATION_STEPS (again for consistency), and more
importantly switch them to using Rectangle rather than Polygon, so that
a polar axhspan is drawn as an annulus. This is necessary due to a
separate "bug", whereby the CLOSEPOLY step on a Polygon would generate
(effectively) a segment that's unfortunately ignored by
_interpolation_steps (as it doesn't appear explicitly), whereas
Rectangle explicitly includes the closing (4th) segment before emitting
a CLOSEPOLY.
For axhline, set the underlying path's _interpolation_steps to GRIDLINE_INTERPOLATION_STEPS (180), which amounts to using the same logic to draw it as for drawing gridlines. This ensures that a polar axhline goes (due to interpolation) around the whole circle, rather than being a trivial line connecting a point to itself. Also update axvline for consistency. (Note that _interpolation_steps has is ignored for rectilinear transforms and thus the change doesn't slow down the common, rectilinear case.)
Increase the number of interpolation steps of ax{v,h}span likewise to GRIDLINE_INTERPOLATION_STEPS (again for consistency), and more importantly switch them to using Rectangle rather than Polygon, so that a polar axhspan is drawn as an annulus. This is necessary due to a separate "bug", whereby the CLOSEPOLY step on a Polygon would generate (effectively) a segment that's unfortunately ignored by _interpolation_steps (as it doesn't appear explicitly), whereas Rectangle explicitly includes the closing (4th) segment before emitting a CLOSEPOLY.
Closes #24996.
PR summary
PR checklist