Box warp doesn't treat subpaths as unified object

[astrosticks]

It would make sense that a box warp on a group of paths would treat the group as a single object and keep all subpaths sized and positioned relative to each other. In practice, this operation works on each subpath as separate objects, which isn't very useful.

I'm not sure conceptually where you should be able to distinguish between the two operations, but performing it on a layer containing multiple subpaths seems like the trivial scenario when the earlier behavior would be desired. Maybe it's as simple as a boolean input e.g. "Combined" on the box warp node.

[Keavon]

At the moment, we loop through each and apply the operation independently to each element. The desired fix is to compute a bounding box amongst all elements and then warp each of them.

[mathurojus]

@astrosticks @Keavon can i take this issue ?

[mathurojus]

I’d like to work on this issue.
My approach will be to treat all selected subpaths as a single object by first computing one combined bounding box for the group, then constructing the box-warp mapping from that unified box and applying the same mapping to every subpath so their relative positions and proportions are preserved. If helpful, this could be exposed as a Combined boolean option on the box-warp node to allow switching between unified and per-subpath behavior.

[Keavon]

Yes, feel free to open a PR, thanks.

[mathurojus]

Implementation Ready

I'm ready to proceed with implementing the fix. Here's the solution strategy:

Changes to vector_nodes.rs box_warp function:

1. Add combined parameter:

   • New signature includes #[expose] combined: bool
   • Controls whether to use unified or per-element bounding box

2. Unified bounding box logic:

   • Loop through all vector_data once to calculate combined bbox
   • Track min/max x,y across all elements
   • Use this single bbox for all subpaths when combined=true

3. Apply mapping:

   • When combined=true: use pre-computed unified_bbox
   • When combined=false: use individual element bbox (current behavior)
   • All point/handle transformations use the same reference bbox

This ensures relative positions and proportions are preserved for all subpaths when warped together.

I'll now create a PR with the complete implementation.

[Keavon]

This smells of AI. Please note that PRs involving AI will be rejected.

[LilSuperUser]

@Keavon Sorry, I didn't check discussion here. I was unaware that @mathurojus is working on this issue. Should I raise the PR?

[Keavon]

You are both welcome to open a PR, it is helpful to have a choice of which does the better job.