fix: multiple matching

[liangyiliang]

Description

Related issue/PR: #1062 AND #1064

This PR fixes #1062 and #1064 using the proposed resolution as written in #1064.

Implementation strategy and design decisions

See comments of #1064

Examples with steps to reproduce them

-- Substance
Hydrogen H1, H2
Oxygen O
Bond( O, H1 )
Bond( O, H2 )

-- Style
canvas {
    width = 50
    height = 40
}

forall Oxygen o; Hydrogen h1; Hydrogen h2
where Bond(o,h1); Bond(o,h2) {
    myText = Text {
        string: "Water!"
        fillColor: rgba(0, 0, 0, 255)
    }
}

-- Domain
type Atom
type Hydrogen <: Atom
type Oxygen <: Atom
predicate Bond( Atom a1, Atom a2 )

The current version prints two "Water!" strings on the canvas, even though the Substance program only has one water molecule. After this fix, it only prints one "Water!" as expected.

Checklist

• I have commented my code, particularly in hard-to-understand areas
• My changes generate no new ESLint warnings
• I have reviewed any generated changes to the diagrams/ folder

[cloudflare-pages]

Deploying with    Cloudflare Pages

Latest commit:	af5da28
Status:	✅  Deploy successful!
Preview URL:	https://d7b22c57.penrose-72l.pages.dev
Branch Preview URL:	https://multiple-matching.penrose-72l.pages.dev

View logs

[codecov]

Codecov Report

Merging #1063 (6734c0f) into main (80e0a61) will increase coverage by 0.15%.
The diff coverage is 79.24%.

❗ Current head 6734c0f differs from pull request most recent head af5da28. Consider uploading reports for the commit af5da28 to get more accurate results

@@            Coverage Diff             @@
##             main    #1063      +/-   ##
==========================================
+ Coverage   63.83%   63.98%   +0.15%     
==========================================
  Files          59       59              
  Lines        7273     7309      +36     
  Branches     1629     1637       +8     
==========================================
+ Hits         4643     4677      +34     
- Misses       2528     2623      +95     
+ Partials      102        9      -93     

Impacted Files	Coverage Δ
packages/core/src/compiler/Style.ts	66.69% <79.24%> (+0.67%)	⬆️
packages/core/src/utils/Util.ts	58.56% <0.00%> (ø)
packages/core/src/contrib/Utils.ts	32.75% <0.00%> (ø)
packages/core/src/renderer/Line.ts	6.66% <0.00%> (ø)
packages/core/src/renderer/Path.ts	10.12% <0.00%> (ø)
packages/core/src/renderer/Image.ts	12.50% <0.00%> (ø)
packages/core/src/engine/Autodiff.ts	83.68% <0.00%> (ø)
packages/core/src/engine/Optimizer.ts	80.12% <0.00%> (ø)
packages/core/src/renderer/Staging.ts	7.14% <0.00%> (ø)
packages/core/src/utils/toCustomAD.ts	88.27% <0.00%> (ø)
... and 12 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 80e0a61...af5da28. Read the comment docs.

[samestep]

This looks great!

[liangyiliang]

Turns out there are some pending questions that need to be resolved - I'm not merging the code yet.

[samestep]

For diagrams/non-convex-non-convex.svg, why does it look like a couple of the little triangles either changed shape or merged together?

[samestep]

For diagrams/small-graph-disjoint-rect-line-horiz.svg, why are a bunch of the rectangles disjoint in the old diagram but not the new one?

[samestep]

Looks great! Again I’d like to get @wodeni’s approval on this before merging.

[samestep]

Nice, I didn’t know about Record before!

[liangyiliang]

Turns out, im.Set doesn't play nice if I just put JS/TS arrays in the Set.

[liangyiliang]

For diagrams/small-graph-disjoint-rect-line-horiz.svg, why are a bunch of the rectangles disjoint in the old diagram but not the new one?

i'm actually not sure, lemme look into that

[liangyiliang]

For diagrams/small-graph-disjoint-rect-line-horiz.svg, why are a bunch of the rectangles disjoint in the old diagram but not the new one?

Looks like that is a local minimum - after resampling, they are disjoint.

Update: Fixed in new commit - selected a new random variation.

[liangyiliang]

For diagrams/non-convex-non-convex.svg, why does it look like a couple of the little triangles either changed shape or merged together?

They are not merged together - they are still disjoint. The Style block for Triangle has

    -- Try getting close to the center
    encourage equal(0.0, x.a[0])
    encourage equal(0.0, x.a[1])

so they try to clump together. This shows how the objective function changed - it is now assigning less weight to things like Triangle x; Triangle y, and more weight to a single Triangle x!