Simple form of y-transform

[OAGr]

Experimentation:

toPointSet(dist) = {
  dist: dist -> Tag.startOpen,
  transform: PointSet.yTransform(PointSet.make(dist))
    -> PointSet.mapY(
      {|e| (e > 0 ? log(e) : log(-e)) -> max(-0)}
    )
    -> Tag.startOpen,
}

{
  dist1: toPointSet(mx(Sym.normal(5, 2), Sym.triangular(50, 51, 52))),
  dist2: toPointSet(
    mx(Sym.uniform(0, 1), Sym.uniform(3, 6), Sym.uniform(10, 12))
  ),
  dist3: toPointSet(mx(1, 2, 3, 4, 4)),
  dist4: toPointSet(mx(Sym.triangular(1, 2, 3))),
}

Or maybe:

toPointSet(dist) = {
  transform1 = Danger.yTransform(PointSet.make(dist))
  transform2 = Danger.yTransform(transform1)
  transform3 = Danger.yTransform(transform2)

  {
    dist: dist -> Tag.startOpen,
    transform1: transform1 -> Tag.startOpen,
    transform2: transform2 -> Tag.startOpen,
    transform3: transform3 -> Tag.startOpen,
  }
}

{
  dist1: toPointSet(
    mx(
      PointSet(normal(5, 2)),
      PointSet(triangular(50, 51, 52)),
      PointSet(uniform(60, 70))
    )
  ),
  dist5: toPointSet(normal(5, 2)),
  dist2: toPointSet(
    mx(Sym.uniform(0, 1), Sym.uniform(3, 6), Sym.uniform(10, 12))
  ),
  dist3: toPointSet(mx(1, 2, 3, 4, 4)),
  dist4: toPointSet(mx(triangular(1, 2, 3))),
}

My impression is that it's doing okay for continuous, but breaking for discrete.

[changeset-bot]

🦋 Changeset detected

Latest commit: 03b8164

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 6 packages

Name	Type
@quri/squiggle-lang	Patch
@quri/squiggle-components	Patch
@quri/prettier-plugin-squiggle	Patch
@quri/versioned-squiggle-components	Patch
vscode-squiggle	Patch
@quri/squiggle-textmate-grammar	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Updated (UTC)
quri-hub	✅ Ready (Inspect)	Visit Preview	Mar 30, 2024 8:44pm
squiggle-components	✅ Ready (Inspect)	Visit Preview	Mar 30, 2024 8:44pm
squiggle-website	✅ Ready (Inspect)	Visit Preview	Mar 30, 2024 8:44pm

1 Ignored Deployment

Name	Status	Preview	Updated (UTC)
quri-ui	⬜️ Ignored (Inspect)	Visit Preview	Mar 30, 2024 8:44pm

[OAGr]

Seems like it's kind of working, but messing up with Uniform distributions.

[berekuk]

I haven't read the tests, but tested this in the playground and output seems mostly fine.

Because of EPSILON and numeric issues, the output will be buggy for multiple transforms applied in a row, but I can't find a quick way to fix that.

[berekuk]

I'm not that fond of overdoing FP, but in this case maybe having a common E_A.forEachPair(() => { ... }) would make sense.

(E_A.forEachPair : Array.forEach :: E_A.pairwise : Array.map)

[berekuk]

(minor)

I'd probably swap y1 and y2 early on to avoid these min/max calls. Like,

let y1 = ...;
let y2 = ...;
if (y1 > y2) [y1, y2] = [y2, y1];

[berekuk]

This should add Number.EPSILON; XYShapes with identical xs values are not valid, as we discussed in DMs.
(this is still hacky but we don't have real stepwise XYShapes yet)

This is my code but it works identical as your previous version, and changing this would require test updates, so I'll leave this for you to improve (or keep as is and deal with bugs later; both versions would be buggy, for different reasons).