[codex] Reuse cached 3D model loads#899
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Quick review note for maintainers: this PR is intentionally scoped to the STEP cache lifecycle gap I mentioned in the attempt comment. It checks the in-memory conversion registry before hydrating from the saved GLB cache, so remounting the same STEP model can reuse the existing converted file instead of creating another blob URL. I kept the fetch/conversion path unchanged and added focused regression coverage for registry-first reuse plus one-time saved-cache hydration. Current GitHub checks are green, and the PR body lists the local verification commands I ran. |
|
Update: I pushed a second commit that broadens this PR from STEP cache reuse to the shared The new cache reuses one in-flight/completed load per URL, returns independent clones so callers can set transforms without mutating the cached source object, and evicts failed loads so a transient failure can be retried. Verification run locally:
|
f5e267d to
dfebc9d
Compare
|
One maintainer-facing verification note for the bounty claim: this PR is non-visual loader/cache work rather than a UI rendering change, so I verified it with focused regression tests plus the package checks/build instead of a demo video. The current checks are green, and the patch covers the duplicate in-flight/completed model load path while preserving independent clones and retry behavior after failed loads. Happy to add a short capture or narrow the scope further if that would make review easier. |
Summary
load3DModelcache so repeated STL/OBJ/WRL/GLTF/GLB URLs reuse one in-flight/completed load instead of fetching and parsing duplicates.Verification
Closes #93
/claim #93