Aagu/replace rust-geo-booleanop with geoclipper #41
Conversation
AnAverageGitUser
changed the title
|
Thanks! This bug is actually the reason I never merged that branch, it was too unstable 🙁 What I wanted to do at the time but got busy with other things (and also waiting for lelongg/geo-clipper#23 to be released) was to use geo-clipper in native, and geo-booleanop in wasm. Could you add that to your PR? |
|
oh lelongg/geo-offset#56 got merged but not yet released, this would be needed for the |
…compatible" the wasm-compatible variant includes the lib geo-booleanop that is currently in some cases not reliable the wasm-incompatible variant includes the lib geo-clipper seems to be more reliable, faster but not wasm-compatible
|
I've used 2 mutually exclusive feature-flags, because I couldn't find a way to use a "not(feature = "x")" in the Cargo.toml. Otherwise the two libs couldn't both be optional. Btw, thanks for all the work that you put into the bevy ecosystem. You're doing quite a lot. |
|
@mockersf Do you have outstanding requests for this pull request or is it fine as is? |
|
Sorry, I'm coming back to this now. I was hoping some crates would have been published by now, I'll try to move forward. but this can be merged anyway, thanks! |
* swapped out geo-booleanop through geo-clipper, because of issue 21re/rust-geo-booleanop#17 * made clippy happy * added mutually exclusive feature-flags "wasm-compatible" and "wasm-incompatible" the wasm-compatible variant includes the lib geo-booleanop that is currently in some cases not reliable the wasm-incompatible variant includes the lib geo-clipper seems to be more reliable, faster but not wasm-compatible * cargo fmt
AnAverageGitUser
deleted the
aagu/replace-rust-geo-booleanop-with-geo-clipper
branch
rust-geo-booleanop seems to have an unfixed bug ( 21re/rust-geo-booleanop#17 ). One possibility might be to switch to a similar library e.g. (Clipper2, https://angusj.com/clipper2/Docs/Overview.htm).
I ran into this while using bevy_pathmesh as one task thread kept on crashing after my application ran X seconds.
Note 1
The Clipper2 library seems to work on integer grid points, therefore one has to choose a precision for the coordinates that one is comfortable with. In this PR I chose 3 digits behind the decimal point.
Note 2
One of the tests
is_in_mesh_overlapping_simplifiedfailed initially, because the amount of polygons was not reduced from 16 to 8, but was stable from 8 to 8 after the update. The functionmerge_overlapping_obstaclesnow returns 8 polygons instead of the 16 previous. I haven't looked into this further, might be an issue.