Fix zerolen intersect #16250
Merged
Fix zerolen intersect #16250
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
timhoffm
reviewed
Jan 18, 2020
anntzer
reviewed
Jan 20, 2020
| while (c2.vertex(&x22, &y22) != agg::path_cmd_stop) { | ||
| // if the segment in path 2 is 0 length, skip to next vertex | ||
| if ((x21 == x22) && (y21 == y22)) { | ||
| continue; |
There was a problem hiding this comment.
if you're really going to do the check to zero-len both here and in segments_intersect (which seems weird as well, I'd rather not(?)) at least the check should be the same in both places (len < epsilon)?
There was a problem hiding this comment.
My logic was this:
- computing
==is cheaper - if trip the
==condition, we will trigger the determinant condition, but the converse is not true. - do the cheaper, stricter thing here
- do the more expensive looser thing where we have to (see my discussion with @timhoffm above)
- keep all of the "isclose" logic together in the code so we are sure that the numbers don't get out of sync
There was a problem hiding this comment.
Well you could change the signature of isclose() to just isclose(a, b) and force atol and rtol as constants in the body of the function... I'm not going to block over that though.
One method to fix matplotlib#15842 In this case we check in `path_intersect_path` that as we iterate through the segments, if we hit a 0 length segment we grab the next vertex before checking if the segments from the two paths intersect.
One method to fix matplotlib#15842 In this case we check that if either segment is 0 length, it intersects with no other segments, not all other segments.
To make clear we won't be changing them.
Use length square being close to 0 rather than strict. I have a worry that because we are using "close" for identifying the determinant is 0, if we use strict equality for the length, then we may pass through the length check, but erroneously go through the den == 0 path.
Makes it a bit easier to identify failing cases.
- move the "close to 0" computation from segments_intersect to path_intersects_path so we only do it once. - change the signature of the c function isclose to back the atol and rtol into the function (to make sure the two places we use it stay in sync)
fef57d4 to
ffaf302
Compare
anntzer
approved these changes
Jan 23, 2020
jklymak
approved these changes
Jan 23, 2020
|
Owee, I'm MrMeeseeks, Look at me. There seem to be a conflict, please backport manually. Here are approximate instructions:
And apply the correct labels and milestones. Congratulation you did some good work ! Hopefully your backport PR will be tested by the continuous integration and merged soon! If these instruction are inaccurate, feel free to suggest an improvement. |
meeseeksmachine
pushed a commit
to meeseeksmachine/matplotlib
that referenced
this pull request
Jan 23, 2020
timhoffm
added a commit
that referenced
this pull request
Jan 23, 2020
…250-on-v3.2.x Backport PR #16250 on branch v3.2.x (Fix zerolen intersect)
tacaswell
pushed a commit
to tacaswell/matplotlib
that referenced
this pull request
Jan 25, 2020
Merge pull request matplotlib#16250 from tacaswell/fix_zerolen_intersect Fix zerolen intersect Also re-factors some tests. Conflicts: lib/matplotlib/tests/test_path.py - unrelated change to test above new test. Kept old version of test and added new test
6 tasks
timhoffm
added a commit
that referenced
this pull request
Jan 26, 2020
…-v3.1.x Backport PR #16250: Fix zerolen intersect
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PR Summary
This provides two fixes for #15842 which independently fix the reported bug.
The first merges segments when iterating through the paths, the second ensures that a 0 length segment intersects with no other segment (rather than all (or maybe just some?) segments.
The 0 length segment not intersecting any segments, including ones it is a point on, matches the behavior of shapely.
attn @TarasKuzyo
PR Checklist