Do not set clip path if it exists #23199
Conversation
|
Illustration import matplotlib as mpl
from matplotlib import pyplot as plt
fig, ax = plt.subplots()
poly = mpl.patches.Polygon([[1,0], [0,1], [-1,0], [0,-1]],
facecolor="#ddffdd", edgecolor="#00ff00", linewidth=2, alpha=0.5)
ax.add_patch(poly)
line = mpl.lines.Line2D((-1, 1), (0.5, 0.5), clip_on=True, clip_path=poly)
ax.add_artist(line)
ax.set_xlim(-1, 1)
ax.set_ylim(-1, 1)gives on current with this PR: (no, there is no need to do |
25b9dfc to
9a86e82
Compare
| @@ -686,7 +687,7 @@ def annotate(self, text, xy, xytext=None, xycoords='data', textcoords=None, | |||
| textcoords=textcoords, arrowprops=arrowprops, | |||
| annotation_clip=annotation_clip, **kwargs) | |||
| a.set_transform(mtransforms.IdentityTransform()) | |||
| if 'clip_on' in kwargs: | |||
| if kwargs.get('clip_on', False) and a.get_clip_path() is None: | |||
There was a problem hiding this comment.
I guess that clip_on not only should be in kwargs, but also be set to True?
| @@ -369,7 +369,7 @@ def draw_path(self, renderer, gc, tpath, affine, rgbFace): | |||
| self.patch.set_transform(affine + self._offset_transform(renderer)) | |||
| self.patch.set_clip_box(gc.get_clip_rectangle()) | |||
| clip_path = gc.get_clip_path() | |||
| if clip_path: | |||
| if clip_path and self.patch.get_clip_path() is None: | |||
There was a problem hiding this comment.
This one is bit doubtful...
|
@oscargus is there a reason you have this in draft? I think it looks good! The only thing that could be added would be a quick test. Might be able to do a |
|
Also if you are interested in looking at the text portion, this PR seemed like a good start that could be reworked: #10811 |
|
I think it was probably the test issue. Which ideally should be testing all of the modifications and I may not really know how PathEffects work. Not sure that what you suggest will work, at least not with a rotated rectangle, as the "shorter" line must be quite exact to pass the test. Will rebase and see what happens for a start. |
24c3802 to
1d2b56d
Compare
|
This is the added test image on main: |
1d2b56d to
4632fac
Compare
| @@ -8455,6 +8455,43 @@ def test_zorder_and_explicit_rasterization(): | |||
| fig.savefig(b, format='pdf') | |||
|
|
|||
|
|
|||
| @image_comparison(["preset_clip_paths.png"], remove_text=True, style="mpl20") | |||
| def test_preset_clip_paths(): | |||
There was a problem hiding this comment.
I'm not sure we want to add an image here that we know is wrong just to have to change it again later?
Some possible alternatives:
- Adding all of the lines/patches completely outside of the rectangle so they are completely clipped? This should make it simpler to compare against a plain rectangle in a check_figures_equal comparison.
- Removing the annotation for now and adding one outside the axes later once that gets fixed.
- Use a Cairo backend if the text clipping works properly there...
- Perhaps simplest though, can you check the expected clip_path after you set it with
artists.get_clip_path() is axes.patch,artist.get_clip_path() is polygonfor the various cases.
There was a problem hiding this comment.
Comparing get_clip_path doesn't seem to work:
assert <matplotlib.transforms.TransformedPatchPath object at 0x000002BE4A717F70> is <matplotlib.patches.Polygon object at 0x000002BE4A5465E0>
Removing annotation and then adding it doesn't really seem better? Now, the test will fail once it is fixed so one will notice this and automatically have to update it (big risk that this test will not be updated.
For patheffects, this happens at draw time, so that will not be tested (not 100% sure that it is now as I do not fully understand that part).
I'd take it that one actually would like to take the intersection of e.g. the axes patch and the provided clip path (although I am not sure if that will ever become a problem). So even if the get_clip_path approach did work, it would break if this was fixed.
I think this is an example that shows that it is better to use the intersection:
import matplotlib as mpl
from matplotlib import pyplot as plt
fig, ax = plt.subplots()
poly3 = mpl.patches.Polygon([[-0.5, 0], [-0.5, 0.5], [0.5, 0.5], [0.5, 0]],
facecolor="g", edgecolor="y", linewidth=2, alpha=0.3)
fig.add_artist(poly3, clip=True)
line = mpl.lines.Line2D((-1, 1), (0.25, 0.25), color='r', clip_on=True, clip_path=poly3)
ax.add_artist(line)
# or
# fig.add_artist(line, clip=True)
With this PR one gets
while current master gives
or
There was a problem hiding this comment.
I'm not sure I totally follow these recent images... The first one looks like the Polygon isn't actually containing the entire line which I would have expected from the clip, so is there still an error present?
The images from main seem to show that the default clip paths are the axes and figure respectively which I expected.
Comparing get_clip_path doesn't seem to work:
Ahh, yes, if you look at the def set_clip_path implementation, I think you can just call TransformedPathPatch(artist)?
Sorry for sending you down this rabbit hole :)
There was a problem hiding this comment.
The first one looks like the Polygon isn't actually containing the entire line which I would have expected from the clip, so is there still an error present?
This is a new error. Although probably this case is artificial, it is the consequence of not setting the previous default clip. So as long as we cannot do an intersection of the clip paths, we have to select which error we get.
Then the question is if these compare the same? Will have to look, but pretty sure that one will have to overload the __eq__ method here. Or extract the path. (I guess is will not work as they will not be the same object.)
PR Summary
This fixes parts of #8270 and in general it does not update the clip path of an artist added with
add_artistif it already have a clip path set. Although that is not enough for #8270 (depending on backend).See: #8270 (comment)
Edit: more cases are added where a new clip path is not added if one is already set. I guess that in general one would like to have a union of the clip paths? Is that easily obtainable in some way?
Tests will be added.
PR Checklist
Tests and Styling
pytestpasses).flake8-docstringsand runflake8 --docstring-convention=all).Documentation
doc/users/next_whats_new/(follow instructions in README.rst there).doc/api/next_api_changes/(follow instructions in README.rst there).