This fixes a panic when undoing some bands.
180952c94f1c7c56c2ce5d2bdd3ab672aa0ccfec did not fix this problem
because we did not stop removing modified dots.
- feat(autorouter): Prepare for population of planar Topo-Navmesh with existing routes
See also issue #166.
- feat(topola-egui): Add dialog for topological navmesh layer selection
- feat(router/ng/eval): Optionally restrict set of allowed TopoNavmesh edges
- fix(router/ng/eval): Use poly_ext_handover
example usages would be marking the apex of a polygon explicitly,
or marking the nodes of a polygon which are part of the convex hull
- make ManageCompounds more generic
- removes unnecessary bounds on handles/refs
This change also propagates to `SpecctraMesadata`, which is a bit unfortunate,
because the code that needs this will never mutate the mesadata,
wrapping it in an `Arc` would be a possible alternative.
Reviewed-on: https://codeberg.org/topola/topola/pulls/161
Co-authored-by: Ellen Emilia Anna Zscheile <fogti+devel@ytrizja.de>
Co-committed-by: Ellen Emilia Anna Zscheile <fogti+devel@ytrizja.de>
This reverts commit b4fe7006b7.
The reverted commit introduced an autorouter regression, observable
in 4x_3rd_order_smd_lc_filters test using egui (see Matrix chat).
I have replaced all instances of `HashMap<...>` with `BTreeMap<...>`,
of `HashSet<...>` with `BTreeSet<...>`, and all the uses of the `Hash`
trait with `Ord`.
I have done a few manual tests and found the behavior to be
deterministic between the GUI application launches. However, undoing an
autoroute command and then manually executing it once again continues to
produce variable results. I suppose this is because of some bug in the
code where edits are applied. Hence, the issue
https://codeberg.org/topola/topola/issues/46
is only partially resolved.
I ran the following command in Fish shell:
```
reuse annotate --skip-unrecognised --copyright="Topola contributors" --license="MIT" **.{rs,md,toml}
```
The choice of year 2024 in the copyright statements is intentional.
The origin of this bug was found via git-bisect, commit
d6fe67a373
is buggy, which switches from GeometryWithRtree to RecordingGeometryWithRtree.
In PR #128, it was discovered that `add_to_compound` appears to
be missing the part that actually invokes the function that performs
the underlying action that is supposed to be diffed.
Fixes#127
These optimize out unnecessary code duplication, reserve vector capacity beforehand by leveraging `Iterator`s and avoid unnecessary double-lookups into HashMaps.
Reviewed-on: https://codeberg.org/topola/topola/pulls/128
Co-authored-by: Alain Emilia Anna Zscheile <fogti+devel@ytrizja.de>
Co-committed-by: Alain Emilia Anna Zscheile <fogti+devel@ytrizja.de>
Mikolaj Wielgus
Ellen Emilia Anna Zscheile