Fix marker overlap #15594
Fix marker overlap #15594
Conversation
|
@jakeprojects are you still interested in finishing this? |
Sort points by z coordinate values similar to Poly3DCollection Added: Image comparison tests for marker overlap Fix PEP8 errors Refactored do_projection to use starred expression
e4b75a1 to
7cd8c4a
Compare
|
I took the liberty to fix the minor comments above and rebase on master. |
|
Not passing though... |
|
Tests were broken because the test and reference images appeared to have a different magnification (or layout margin). Not sure if we changed anything in this respect. Anyway, I've exchanged the image comparisons with figure comparison tests checking:
Commits should be squashed before merging to remove the test images from the earlier commits. I didn't want to force push a squash on another user's repo and remove part of the test code without approval. |
|
I'd rather they be rebased out instead of squashing, because the commits are written by different authors since this was taken up from a previous PR. I have not checked yet if the new test is okay though. |
|
Not sure how that rebase should work. The first commit contains the image tests and part of the relevant code. Do you wish to strip the tests from that commit? I think the best is to squash everything into one commit because we do not need the back-and-forth of the development history here. We can credit all the authors for the commit. |
| vys = vys[z_markers_idx] | ||
| fcs = fcs[z_markers_idx] | ||
| ecs = ecs[z_markers_idx] | ||
| vps = np.vstack((vxs, vys)).T |
There was a problem hiding this comment.
Why change to vstack.T instead of column_stack?
There was a problem hiding this comment.
Both are fine. I've changed to column_stack, which seems a bit clearer.
Interestingly, vstack.T is marginally faster for > 10k elements:
But 10ms difference is not the end of the world also for 1e6 elements. Probably all the surrounding code is much slower anyway.
|
Please merge #17378 instead: Same code but commits squashed and images stripped. (I didn't want to push the squash to another user's repo, so I had to open a separate PR. All three contributors are credited. |
|
Closing in favour of the above squashed PR. |
PR Summary
Performance optimization of sorting by z co-ordinates for scatter plots of Path3DCollection objects. Locally observed a 20x speed-up using this approach, unit tests test_mplot3d.py passed. Initially forked from xenoryt.
Reuses #10791 with performance optimization.
PR Checklist