feat(kernel): implement DSATUR coloring algorithm for vertices #245
Conversation
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #245 +/- ##
==========================================
+ Coverage 78.97% 79.77% +0.80%
==========================================
Files 52 55 +3
Lines 6834 7065 +231
==========================================
+ Hits 5397 5636 +239
+ Misses 1437 1429 -8 ☔ View full report in Codecov by Sentry. |
|
TODO:
|
imrn99
changed the title
there's an issue in neighbors computation bc nodes on the edge have a single oriented dart as links
| /// # Return | ||
| /// | ||
| /// The function will return the maximal used color, i.e. the `n_color_used - 1`. | ||
| pub fn color<T: CoordsFloat>(cmap: &mut CMap2<T>) -> u8 { |
There was a problem hiding this comment.
Not sure if it is a good name. Perhaps the algorithm name will be better. Add to the description it is a distance-1 coloring of the vertices.
There was a problem hiding this comment.
I re-export the function as color_dsatur in the coloring module. I can switch it to dsatur if you want
| pub fn color<T: CoordsFloat>(cmap: &mut CMap2<T>) -> u8 { | ||
| cmap.add_attribute_storage::<Color>(); | ||
|
|
||
| // build graph data as a collection of (Vertex, Vec<Neighbors>) |
There was a problem hiding this comment.
Could this part be helpful elsewhere?
You can add a way to create a graph from a CMap. And use things like Coupe later ;).
There was a problem hiding this comment.
Could this part be helpful elsewhere? You can add a way to create a graph from a CMap.
Yeah, I think I can also make it generic enough to handle all possible adjacency types (e.g. here we're using vertex-vertex, but we could fetch incident faces for another usage).
And use things like Coupe later ;).
I actually wanted to try and use it, but I think all coloring algorithms of coupe are focused on balancing weights, rather than making independent subsets. Is that right?
There was a problem hiding this comment.
I'm actually going to implement my idea for graphes in another Pr, and I'll sync this branch with master to use it after.
| // - not colored | ||
| // - of highest saturation | ||
| // - if there are multiple, take the one of highest degree | ||
| crt_node = nodes |
There was a problem hiding this comment.
Can you comment on the complexity?
There was a problem hiding this comment.
I'll add it to the doc, but it's O(n²)
Description
Associated issue: closes #236
Content description:
add_attribute_storagemethod toCMap2for convenienceFollow-up
Aside from using it, I think the algorithm implementation can be modified to handle coloring on all types of I-cells, using a trait and the
enum_dispatchcrate to avoid dyn object lookup costs.