refactor(ios): implement snapshot engine - #2211
Conversation
Size Report
npm unpacked components
Startup median (7 runs, lower is better):
Top changed chunks: no changes in the largest emitted chunks. Top changed packed files
|
6537862 to
490c83f
Compare
|
Re-reviewed exact head
The Android Integration failure ( |
|
Re-reviewed exact head BLOCKED only on readiness evidence: replace the stack boilerplate body with the #2191 slice-B scope, itemized move/growth rationale and why a materially smaller design was rejected, exact-head validation, and recorded iOS Simulator proof for the moved semantic-compaction route. Green iOS Smoke does not document that changed-path evidence. |
8917129 to
4c06c15
Compare
8917129 to
8e3a26e
Compare
|
Re-reviewed exact head BLOCKED on readiness evidence: the body is only stack boilerplate, so the #2191 slice-B scope, itemized growth and materially-smaller-design rationale, exact-head validation, and recorded iOS Simulator semantic-compaction proof are absent. This rebase also cancelled prior lanes; current exact-head CI has not completed. No readiness label. |
8e3a26e to
6fd7576
Compare
|
Re-reviewed exact head Still BLOCKED on readiness record: there is no recorded real iOS Simulator command/outcome for the moved semantic-compaction route, nor a precise blocker/residual-risk statement. The validation section is also stale because it says iOS Smoke is still in progress, and the >700-line growth rationale remains aggregate rather than itemizing the principal moved/new categories. Add the changed-path Simulator evidence, update the completed checks, and itemize justified growth before labeling. |
|
Re-reviewed new exact head Still BLOCKED on readiness. The body remains stamped at old |
|
Re-reviewed exact head |
|
Addressed in the PR body at exact head |
|
Reconciled unchanged exact head |
Summary
Slice B of #2191 (“refactor(ios): build and prove the converged snapshot engine”), stacked on #2206 at exact head
4ef845fe1fec3d0cd4ed9a17cfcad703a35bea45.@agent-device/capture-kit/ios-snapshot-engine: planning, viewport/cumulative clip folding, regular/raw projection, scope/depth, eligibility/actionability, hints, lineage, invariants, and publication.src/snapshot/snapshot-presentation/iosfaçade/noise owner.Scope and design rationale: 43 touched files, +2,047/-463 vs #2206. Itemized growth is the new capture-kit engine modules/tests and one explicit package export, offset by renames/deletions from the old semantic path and migrated consumers/tests. A materially smaller patch that kept the old façade/import surface was rejected because it would leave the superseded path as an effective second owner and make the one-engine invariant unenforceable; splitting the engine across existing packages would do the same. Bundle evidence is bounded: JS source +1.2 kB, npm unpacked +1.2 kB, tarball +210 B; Apple runner and other package components unchanged.
Validation
Exact reviewed head:
6fd7576902065aa38b70e7ed869385bd13f1dbecpnpm buildpnpm typecheckpnpm exec vitest run --project unit-core packages/capture-kit/src/ios-snapshot-engine/engine.test.ts(11/11)smoke:regular-visible-depth-frontier: public regular depth 1 kept the raw-deepvisible-depth-projected-childat presented depth 1 with wrapper reparenting, while raw full/depth-bounded output preserved the acquisition frontier;smoke:automation-input,smoke:form-input, andsmoke:capture-closealso passed.