ContextQMD
Back to Plate

Slate v2 ClawSweeper Next Execution Plan

docs/plans/2026-05-08-slate-v2-clawsweeper-next-execution-plan.md

53.0.638.7 KB
Original Source

Slate v2 ClawSweeper Next Execution Plan

Current Verdict

Do not create another broad slate-ralplan for this batch.

The architecture planning lanes and this full Ralph execution plan are closed:

  • Mobile/IME: done, with no exact fixed/improved Mobile/IME claims promoted.
  • DOM selection/focus bridge: done.
  • React/decorations: done.
  • Performance/scalability: ralplan done; benchmark repair, proof slices, compare unblocker, and claim accounting are done.

Completed work:

  1. Repair benchmark API drift.
  2. Run performance proof slices behind the repaired benchmark surface.
  3. Run ClawSweeper claim sync after each slice.
  4. Keep Mobile/IME leftovers in the right proof buckets instead of pretending desktop Chromium rows close raw Android/iOS/Safari reports.

Source Of Truth

Read first:

  • .tmp/completion-checks/slate-v2-performance-scalability-slate-issues-ralplan.md
  • docs/plans/2026-05-08-slate-v2-performance-scalability-slate-issues-ralplan.md
  • docs/plans/2026-05-07-slate-v2-mobile-ime-input-runtime-ralplan.md
  • docs/slate-v2/ledgers/issue-coverage-matrix.md
  • docs/slate-v2/ledgers/fork-issue-dossier.md
  • docs/slate-v2/references/pr-description.md
  • docs/slate-issues/gitcrawl-live-open-ledger.md
  • docs/slate-issues/gitcrawl-clusters.md
  • live files under .tmp/slate-v2

Current completion state:

  • umbrella Ralph execution: done
  • completed: Item 1 benchmark API drift and Slices 2A through 2H
  • current owner: none
  • root completion check: expected green for this file-backed execution plan

Non-Goals

  • No one-patch "fix all performance" work.
  • No new Fixes #... claim from headless benchmark output alone.
  • No raw-device Mobile/IME closure from desktop Chromium or desktop WebKit.
  • No table/product/plugin policy closure in raw Slate unless the raw primitive actually changes.
  • No manual live-ledger classification columns.

Item 1 - Benchmark API Drift

Goal

Make the existing current benchmark runners use the current Slate v2 write API. The slice is infrastructure repair, not behavior work.

Known Drift Owners

Stale patterns found in .tmp/slate-v2:

  • Editor.withTransaction(...)
  • editor.insertText(...)
  • editor.select(...)
  • old normalization assertions that no longer match current normalization
  • possible stale helper wrappers inside benchmark-only files

Primary files:

  • .tmp/slate-v2/scripts/benchmarks/core/current/transaction-execution.mjs
  • .tmp/slate-v2/scripts/benchmarks/core/current/text-selection.mjs
  • .tmp/slate-v2/scripts/benchmarks/core/current/normalization.mjs
  • .tmp/slate-v2/scripts/benchmarks/core/current/refs-projection.mjs
  • .tmp/slate-v2/scripts/benchmarks/core/current/query-ref-observation.mjs
  • .tmp/slate-v2/scripts/benchmarks/core/current/node-transforms.mjs

Reference pattern:

  • .tmp/slate-v2/scripts/benchmarks/core/current/editor-store.mjs

Use current transaction calls:

ts
editor.update((tx) => {
  tx.selection.set(target);
  tx.text.insert(text, options);
});

Use current snapshot/assertion APIs:

ts
Editor.replace(editor, snapshot);
Editor.getSnapshot(editor);

Execution Slices

Slice 1A - Transaction Benchmark API

Owner:

  • .tmp/slate-v2/scripts/benchmarks/core/current/transaction-execution.mjs

Work:

  • Replace Editor.withTransaction(...) with current editor.update(...) or the current batch/apply API.
  • Preserve the benchmark's comparison intent: manual multi-op write versus applyOperations(...).
  • Add shared percentile artifact output if missing.

Driver gates:

bash
cd /Users/zbeyens/git/slate-v2
bun run bench:core:transaction:local
bun run bench:slate:6038:local

Claim rule:

  • Keep #6038 as Improves unless accepted op-family thresholds and artifact rows exist.

Slice 1B - Text/Selection Benchmark API

Owner:

  • .tmp/slate-v2/scripts/benchmarks/core/current/text-selection.mjs

Work:

  • Replace stale editor.select and editor.insertText helper bodies with transaction-owned tx.selection.set(...) and tx.text.insert(...).
  • Keep selection assertions behavior-facing.
  • Emit p75/p95/p99 through shared summary helpers.

Driver gate:

bash
cd /Users/zbeyens/git/slate-v2
bun run bench:core:text-selection:local

Claim rule:

  • No issue promotion from API repair.

Slice 1C - Ref And Observation Benchmarks

Owner:

  • .tmp/slate-v2/scripts/benchmarks/core/current/refs-projection.mjs
  • .tmp/slate-v2/scripts/benchmarks/core/current/query-ref-observation.mjs

Work:

  • Convert stale write helpers to current transaction writes.
  • Preserve ref/projection behavior assertions.
  • Keep this focused on benchmark health, not ref behavior redesign.

Driver gates:

bash
cd /Users/zbeyens/git/slate-v2
bun run bench:core:refs-projection:local
bun run bench:core:query-ref-observation:local

Slice 1D - Normalization Benchmark API

Owner:

  • .tmp/slate-v2/scripts/benchmarks/core/current/normalization.mjs

Work:

  • Convert stale text writes.
  • Recheck normalization assertions against current Slate v2 normalization law.
  • If an assertion is stale, update the assertion only when current package tests already prove that behavior.

Driver gate:

bash
cd /Users/zbeyens/git/slate-v2
bun run bench:core:normalization:local

Slice 1E - Node Transforms Benchmark API

Owner:

  • .tmp/slate-v2/scripts/benchmarks/core/current/node-transforms.mjs

Work:

  • Replace stale editor.select(...) calls with current selection writes.
  • Keep node-transform scenario shape intact.

Driver gate:

bash
cd /Users/zbeyens/git/slate-v2
bun run bench:core:node-transforms:local

Item 1 Closure

Run:

bash
cd /Users/zbeyens/git/slate-v2
bun run bench:core:transaction:local
bun run bench:core:text-selection:local
bun run bench:core:refs-projection:local
bun run bench:core:query-ref-observation:local
bun run bench:core:normalization:local
bun run bench:core:node-transforms:local
bun check

Then in plate-2:

bash
pnpm lint:fix
bun run completion-check -- --file .tmp/completion-checks/slate-v2-performance-scalability-slate-issues-ralplan.md

Completion rule:

  • Current benchmark runners are green.
  • Percentile artifacts exist where relevant.
  • No issue claim is promoted by this item.
  • Completion state names the next performance proof owner.

Item 2 - Performance Proof Slices

These run only after Item 1 is green. Each slice gets its own ralph execution entry and its own ClawSweeper sync decision.

Slice 2A - Core Transaction Thresholds

Issue pressure:

  • #6038
  • related guards: #3656, #4141, #3430, #2051

Work:

  • Add op-family tags to transaction benchmark output if absent.
  • Capture p75/p95/p99 for manual transaction and batch apply paths.
  • Define review thresholds as calibrated gates, not release failure gates.
  • Compare repeated runs before hardening any threshold.

Driver gates:

bash
cd /Users/zbeyens/git/slate-v2
bun run bench:core:transaction:local
bun run bench:slate:6038:local
bun check

Claim rule:

  • Keep #6038 as Improves until threshold repeatability is accepted.
  • Do not claim Fixes #6038 from one local benchmark run.

Slice 2B - React Rerender Breadth

Issue pressure:

  • #5131
  • related performance rows: #5433, #5420, #5274, #4317, #3354

Work:

  • Re-run selector/depth/leaf/subscriber benchmark rows.
  • Confirm broad hooks remain broad by contract.
  • Record selector-locality proof without pretending useSlate becomes narrow.

Driver gates:

bash
cd /Users/zbeyens/git/slate-v2
bun run bench:react:rerender-breadth:local
bun test ./packages/slate-react/test/provider-hooks-contract.tsx ./packages/slate-react/test/projections-and-selection-contract.tsx
bun check

Claim rule:

  • #5131 stays Not claimed while useSlate / useEditor remain broad by contract.
  • Selector APIs can be documented as the intended scalable path.

Slice 2C - Projection And Decoration Churn

Issue pressure:

  • overlay/decorations scalability rows already covered by the React/decorations plan
  • future-proof performance rows from the matrix

Work:

  • Use source-scoped invalidation counters from existing React benchmark output.
  • Confirm unrelated decoration source recompute counts stay at 0.
  • Do not change public decoration APIs unless a benchmark exposes a real gap.

Driver gates:

bash
cd /Users/zbeyens/git/slate-v2
bun run bench:react:huge-document-overlays:local
bun test ./packages/slate-react/test/projections-and-selection-contract.tsx -t source
bun check

Claim rule:

  • No fixed issue claim unless the original repro is replayed.

Slice 2D - Rendering Strategy Behavior

Issue pressure:

  • large-document rendering strategy
  • shell/native behavior rows
  • future static/virtual rendering pressure

Work:

  • Keep virtualized/staged rendering experimental unless default behavior is proven safe.
  • Add or update browser rows only for native behavior that users actually hit: selection, copy, find, IME, deletion, drag, and caret repair.
  • Keep rendering metrics separate from browser correctness.

Driver gates:

bash
cd /Users/zbeyens/git/slate-v2
bun run bench:react:huge-document-overlays:local
PLAYWRIGHT_RETRIES=0 PLAYWRIGHT_WORKERS=1 bun run playwright playwright/integration/examples/dom-coverage-boundaries.test.ts --project=chromium
bun check

Claim rule:

  • No large-document superiority claim unless compared against the same legacy cohort with legacy chunking enabled.

Slice 2E - Clipboard Large Payload

Issue pressure:

  • #5945
  • #5992
  • #4056
  • proof-route backlog: #2733, #2669, #790, #5216, #5592, #4202, #4210, #3748, #5349, #4025

Work:

  • Extend existing clipboard benchmark families instead of adding one-off benchmark files.
  • Add browser stress rows only when they map to an issue-size path.
  • Keep core benchmark and browser stress claims separate.

Driver gates:

bash
cd /Users/zbeyens/git/slate-v2
bun run bench:core:clipboard-large-payload:local
bun run bench:slate:5945:issue
PLAYWRIGHT_RETRIES=0 PLAYWRIGHT_WORKERS=1 bun run playwright playwright/stress/generated-editing.test.ts --project=chromium
bun check

Note:

  • bench:core:clipboard-large-payload:local is not currently listed in package.json; if it stays absent, add a package script only if this becomes a repeated driver gate. Otherwise run the benchmark file directly.

Claim rule:

  • No Fixes #5945, Fixes #5992, or Fixes #4056 from headless benchmark proof alone.
  • Promotion needs browser stress proof or explicit maintainer equivalence.

Slice 2F - Safari Backward Selection

Issue pressure:

  • Safari/WebKit selection latency and backward-selection rows
  • do not merge with Mobile/IME Safari rows unless composition is involved

Work:

  • Add the narrowest WebKit browser row that proves backward selection behavior.
  • Keep it out of raw mobile/device language.

Driver gate:

bash
cd /Users/zbeyens/git/slate-v2
PLAYWRIGHT_RETRIES=0 PLAYWRIGHT_WORKERS=1 bun run playwright <target-test> --project=webkit --grep "<backward selection row>"

Claim rule:

  • WebKit desktop proof can support browser selection claims.
  • It does not close iOS keyboard/autocorrect/composition issues.

Slice 2G - History Retained Memory

Issue pressure:

  • #3752
  • retained payload and history memory pressure

Work:

  • Add current-only retained-memory benchmark.
  • Include retained payload tags, operation count, history entry count, and heap evidence when available.
  • Keep memory benchmark separate from latency benchmark.

Driver gate:

bash
cd /Users/zbeyens/git/slate-v2
bun <new-or-existing-history-memory-benchmark>
bun check

Claim rule:

  • No Fixes #3752 without retained-memory evidence.

Slice 2H - Legacy Compare Unblocker

Work:

  • Repair compare runners only after current-only benchmarks are healthy.
  • The earlier failure was legacy /Users/zbeyens/git/slate build related; do not mix that failure with current Slate v2 correctness.

Driver gates:

bash
cd /Users/zbeyens/git/slate-v2
bun run bench:core:normalization:compare:local
bun run bench:core:huge-document:compare:local

Claim rule:

  • Compare rows inform PR language only after both sides build and run.

Item 3 - ClawSweeper Sync After Each Slice

Run this loop after every Item 1 or Item 2 slice.

Sync Decision Matrix

Slice resultRequired sync
No behavior, proof, or claim changeUpdate active plan/completion state only.
New benchmark artifact onlyUpdate plan/completion; update matrix only if proof status changes.
New browser proofUpdate plan/completion, matrix row, and fork dossier row for related issues.
New Improves claimUpdate matrix, fork dossier, PR reference, fixed/improved counts if needed.
New Fixes claimUpdate matrix, fork dossier, PR reference fixed claims/counts, and relevant ledger notes.
Explicit non-claim decisionRecord in plan and matrix/dossier only when it prevents future overclaiming.

Files To Check

  • docs/slate-v2/ledgers/issue-coverage-matrix.md
  • docs/slate-v2/ledgers/fork-issue-dossier.md
  • docs/slate-v2/references/pr-description.md
  • docs/slate-issues/gitcrawl-live-open-ledger.md
  • docs/slate-issues/gitcrawl-clusters.md
  • active plan under docs/plans/
  • active completion file under .tmp/completion-checks/
  • active goal state

Rules

  • Do not add Fixes #... without exact repro proof.
  • Do not add "maybe fixed" prose to the PR description.
  • Do not add live-ledger manual classification columns.
  • If the live ledger needs a note, add a dated processing note that preserves live gitcrawl fields.
  • Keep PR prose short: exact fixed claims, exact improved claims, and links to the full matrix. Nothing else.
  • If a slice only repairs benchmark infrastructure, say "issue claims unchanged" in the completion state.

Sync Gates

For docs-only sync in plate-2:

bash
pnpm lint:fix
bun run completion-check

For any slice touching .tmp/slate-v2:

bash
cd /Users/zbeyens/git/slate-v2
bun check

Then return to plate-2:

bash
pnpm lint:fix
bun run completion-check

Item 4 - Mobile/IME Leftovers

The Mobile/IME macro lane is done for autonomous desktop proof. The remaining rows are not "more harvesting." They need the right proof owner.

Bucket 4A - Raw Device / Manual Exact Closure

Issues:

  • #6051, #6022, #5983, #5989, #5883, #4400, #5680, #5493, #5130, #5836, #5805, #5666, #5291, #5371, #5167, #4959, #4861, #4602, #4354, #4372, #5183, #5391

Plan:

  • Keep as related/non-claim until matching raw environment proof exists.
  • Use bun test:mobile-device-proof:raw only on a real Android/iOS/Appium lane that can produce raw device artifacts.
  • Do not let Playwright mobile viewport, Chromium CDP IME, or semantic mobile labels satisfy raw-device claims.

Driver gate when device lane exists:

bash
cd /Users/zbeyens/git/slate-v2
SLATE_BROWSER_RAW_MOBILE_REQUIRED=1 bun test:mobile-device-proof:raw

Claim rule:

  • Exact closure needs issue-matching device, keyboard/browser, event trace, and model/DOM assertion artifacts.

Bucket 4B - Safari/WebKit Composition Or Autocorrect

Issues:

  • #4030, #3943, #4640, #4543, #4085

Plan:

  • Keep desktop WebKit direct-input proof separate from honest composition proof.
  • Add WebKit rows only when they actually exercise the Safari/WebKit behavior under claim.
  • Do not close iOS/Safari autocorrect reports from Chromium composition rows.

Driver gate:

bash
cd /Users/zbeyens/git/slate-v2
PLAYWRIGHT_RETRIES=0 PLAYWRIGHT_WORKERS=1 bun run playwright <target-test> --project=webkit --grep "<exact Safari/WebKit row>"

Claim rule:

  • Desktop WebKit proof may support a browser-specific Related or Improves row. Raw iOS closure still needs iOS proof.

Bucket 4C - Android ReadOnly / OnChange / Manager Integration

Issues:

  • #5078, #5034, #5026, #4994

Plan:

  • Create integration tests only if they can observe operation freshness, onChange timing, readOnly transitions, and Android/input-manager routing.
  • Unit-level model-input rows are not enough for exact closure.
  • Keep these out of Mobile/IME fixed/improved PR language until the integration path proves the reported lifecycle.

Candidate owners:

  • .tmp/slate-v2/packages/slate-react/test/model-input-strategy-contract.test.ts
  • .tmp/slate-v2/packages/slate-react/test/selection-reconciler-contract.ts
  • a new focused React integration test only if existing package tests cannot observe the required lifecycle

Driver gates:

bash
cd /Users/zbeyens/git/slate-v2
bun --filter slate-react test:vitest -- model-input-strategy-contract.test.ts
bun test ./packages/slate-react/test/selection-reconciler-contract.ts
bun --filter slate-react typecheck

Claim rule:

  • No exact closure until the test observes the lifecycle named by the issue.

Bucket 4D - Product / Plugin / Policy Rows

Issues:

  • #5928, #4532, #4223, #3573, #4621, #3222, #3177

Plan:

  • Keep as product/plugin/policy pressure.
  • Raw Slate can expose primitives, event policy, and substrate behavior.
  • Do not implement product policy in raw Slate just to close an issue.
  • Move to Plate/example/plugin owner only when a current Slate v2 product surface exists.

Claim rule:

  • Default status stays Related or Not claimed.

Global Execution Order

  1. Item 1A: transaction benchmark API drift.
  2. Item 1B: text/selection benchmark API drift.
  3. Item 1C: refs/projection and query/ref benchmark API drift.
  4. Item 1D: normalization benchmark API drift.
  5. Item 1E: node transforms benchmark API drift.
  6. Item 3 sync: record that Item 1 changed no issue claims.
  7. Item 2A: core transaction thresholds.
  8. Item 3 sync: #6038 decision.
  9. Item 2B: React rerender breadth.
  10. Item 3 sync: #5131 and selector-path decision.
  11. Item 2C: projection/decorations churn.
  12. Item 3 sync: overlay/decorations proof status.
  13. Item 2D: rendering strategy behavior.
  14. Item 3 sync: large-document/native behavior decisions.
  15. Item 2E: clipboard large payload.
  16. Item 3 sync: #5945/#5992/#4056 decision.
  17. Item 2F: Safari backward selection.
  18. Item 3 sync: WebKit proof status.
  19. Item 2G: history retained memory.
  20. Item 3 sync: #3752 decision.
  21. Item 2H: legacy compare unblocker, only if current-only proof is already healthy.
  22. Item 4: keep Mobile/IME leftovers in their proof buckets. Do not run more desktop proof unless one of the buckets gains a concrete owner.

Completion Criteria

This umbrella plan is complete when:

  • all current benchmark runners named in Item 1 are green or intentionally removed from package scripts with a reason;
  • each performance proof slice has an artifact, browser row, retained-memory row, or explicit non-claim decision;
  • every changed issue claim is reflected in the matrix, fork dossier, and PR reference;
  • Mobile/IME exact-closure leftovers are bucketed by proof owner and no desktop proof is mislabeled as raw-device closure;
  • .tmp/slate-v2 bun check passes after any source/test/docs change there;
  • plate-2 pnpm lint:fix and bun run completion-check pass after plan, ledger, or completion-state edits here.

Stop Rules

  • Stop after each bounded Ralph slice if the next slice would touch a different owner family.
  • Keep going inside one slice only while the failures belong to that slice.
  • If the same benchmark runner fails three times, record attempts and pivot to the smallest different runner that proves the same helper.
  • Mark blocked only when the remaining proof requires unavailable raw Android/iOS/Safari/device infrastructure or a maintainer decision.
  • Otherwise keep status pending and name the next runnable owner.

Ralph Execution Ledger

Slice 1A - Transaction Benchmark API Drift - 2026-05-08

Status: complete.

Owner:

  • .tmp/slate-v2/scripts/benchmarks/core/current/transaction-execution.mjs

Work completed:

  • Replaced removed Editor.withTransaction(...) with current editor.update((tx) => tx.operations.replay(...)).
  • Replaced removed editor.applyOperations(...) with the current comparison shape: separate per-operation updates versus one batched transaction replay.
  • Preserved legacy artifact fields withTransactionMeanMs and applyBatchMeanMs as compatibility aliases while adding explicit separateUpdateMeanMs, updateReplayMeanMs, and percentile lane summaries.
  • Switched artifact writing to the shared writeBenchmarkArtifact(...) helper.

Evidence:

  • From .tmp/slate-v2, bun run bench:core:transaction:local passed.
  • From .tmp/slate-v2, bun run bench:slate:6038:local passed.
  • Artifact .tmp/slate-v2/tmp/bench-slate-6038.json includes p75, p95, and p99 for separateUpdateMs and updateReplayMs.
  • From .tmp/slate-v2, bun lint:fix passed.
  • From .tmp/slate-v2, bun check passed.

Issue claims:

  • No issue claim changed.
  • #6038 remains Improves, not Fixes.

Reference docs:

  • No change. This slice repaired benchmark API drift and artifact shape only.

Next owner:

  • Slice 1B: .tmp/slate-v2/scripts/benchmarks/core/current/text-selection.mjs.

Slice 1B - Text/Selection Benchmark API Drift - 2026-05-08

Status: complete.

Owner:

  • .tmp/slate-v2/scripts/benchmarks/core/current/text-selection.mjs

Work completed:

  • Replaced removed editor.insertText(...), editor.select(...), editor.delete(...), editor.setSelection(...), editor.setPoint(...), editor.move(...), and editor.collapse(...) benchmark calls with current transaction APIs under tx.text.* and tx.selection.*.
  • Preserved existing benchmark lanes and behavior assertions.

Evidence:

  • From .tmp/slate-v2, bun run bench:core:text-selection:local passed.
  • Artifact .tmp/slate-v2/tmp/slate-text-selection-benchmark.json includes p75, p95, and p99 in each lane through the shared summary helper.

Issue claims:

  • No issue claim changed.

Reference docs:

  • No change. This slice repaired benchmark API drift only.

Next owner:

  • Slice 1C: .tmp/slate-v2/scripts/benchmarks/core/current/refs-projection.mjs and .tmp/slate-v2/scripts/benchmarks/core/current/query-ref-observation.mjs.

Slice 1C - Ref And Observation Benchmark API Drift - 2026-05-08

Status: complete.

Owner:

  • .tmp/slate-v2/scripts/benchmarks/core/current/refs-projection.mjs
  • .tmp/slate-v2/scripts/benchmarks/core/current/query-ref-observation.mjs

Work completed:

  • Replaced removed editor.insertText(...) helpers with current tx.text.insert(...).
  • Replaced removed editor.moveNodes(...) benchmark calls with tx.nodes.move(...).
  • Replaced stale Editor.nodes(...) observation read with current editor.read((state) => state.nodes.match(...)).

Evidence:

  • From .tmp/slate-v2, bun run bench:core:refs-projection:local passed.
  • From .tmp/slate-v2, bun run bench:core:query-ref-observation:local passed.
  • Artifacts .tmp/slate-v2/tmp/slate-refs-projection-benchmark.json and .tmp/slate-v2/tmp/slate-query-ref-observation-benchmark.json include percentile lane fields through the shared summary helper.

Issue claims:

  • No issue claim changed.

Reference docs:

  • No change. This slice repaired benchmark API drift only.

Next owner:

  • Slice 1D: .tmp/slate-v2/scripts/benchmarks/core/current/normalization.mjs.

Slice 1D - Normalization Benchmark API Drift - 2026-05-08

Status: complete.

Owner:

  • .tmp/slate-v2/scripts/benchmarks/core/current/normalization.mjs

Work completed:

  • Replaced removed editor.insertText(...) helper usage with current tx.text.insert(...).
  • Replaced stale inline schema mutation with current extension-owned element specs: editor.extend({ elements: [{ type: 'inline', inline: true }] }).
  • Kept the inline-flattening assertion because current package schema tests prove inline behavior through extension specs and the benchmark assertion passes under that contract.

Evidence:

  • From .tmp/slate-v2, bun run bench:core:normalization:local passed.
  • Artifact .tmp/slate-v2/tmp/slate-normalization-benchmark.json includes percentile lane fields through the shared summary helper.

Issue claims:

  • No issue claim changed.

Reference docs:

  • No change. This slice repaired benchmark API drift only.

Next owner:

  • Slice 1E: .tmp/slate-v2/scripts/benchmarks/core/current/node-transforms.mjs.

Slice 1E - Node Transforms Benchmark API Drift - 2026-05-08

Status: complete.

Owner:

  • .tmp/slate-v2/scripts/benchmarks/core/current/node-transforms.mjs

Work completed:

  • Replaced removed editor transform calls with current transaction-owned APIs: tx.selection.set(...), tx.fragment.insert(...), and tx.nodes.*.
  • Preserved the node-transform scenario shape and existing benchmark assertions.

Evidence:

  • From .tmp/slate-v2, bun run bench:core:node-transforms:local passed.
  • Artifact .tmp/slate-v2/tmp/slate-node-transform-benchmark.json includes percentile lane fields through the shared summary helper.

Issue claims:

  • No issue claim changed.

Reference docs:

  • No change. This slice repaired benchmark API drift only.

Next owner:

  • Item 3 sync for Item 1, then Item 2A core transaction thresholds.

Item 1 Closure - Benchmark API Drift - 2026-05-08

Status: complete.

Work completed:

  • Repaired all current benchmark runners named by Item 1 against current Slate v2 transaction/write APIs.
  • Confirmed the repaired benchmark family as a group.

Evidence:

  • From .tmp/slate-v2, these passed:
    • bun run bench:core:transaction:local
    • bun run bench:core:text-selection:local
    • bun run bench:core:refs-projection:local
    • bun run bench:core:query-ref-observation:local
    • bun run bench:core:normalization:local
    • bun run bench:core:node-transforms:local
    • bun check

Issue claims:

  • No issue claim changed.
  • #6038 remains Improves, not Fixes.

Reference docs:

  • No matrix, dossier, or PR-reference update required. Item 1 changed benchmark infrastructure only.

Next owner:

  • Slice 2A: core transaction thresholds.

Slice 2A - Core Transaction Thresholds - 2026-05-08

Status: complete.

Owner:

  • .tmp/slate-v2/scripts/benchmarks/core/current/transaction-execution.mjs

Work completed:

  • Extended the existing #6038 transaction benchmark artifact with op-family rows for mixed structural/text batches, text insert/remove, node-property, node-insert, node-split, and node-move operations.
  • Kept compatibility fields for existing consumers: withTransactionMeanMs, applyBatchMeanMs, separateUpdateMeanMs, and updateReplayMeanMs.
  • Added calibration-only threshold metadata instead of inventing a release failure gate from one local machine.

Evidence:

  • From .tmp/slate-v2, bun run bench:core:transaction:local passed.
  • From .tmp/slate-v2, bun run bench:slate:6038:local passed.
  • Artifact .tmp/slate-v2/tmp/bench-slate-6038.json has opFamilyLanes.*.separateUpdateMs.p95/p99 and opFamilyLanes.*.updateReplayMs.p95/p99.

Issue claims:

  • No issue claim changed.
  • #6038 stays Improves; no Fixes #6038 claim until thresholds are accepted after repeat variance.

Reference docs:

  • No matrix, dossier, or PR-reference update required. The existing #6038 Improves decision already captures the non-promotion rule.

Next owner:

  • Slice 2B: React rerender breadth.

Slice 2B - React Rerender Breadth - 2026-05-09

Status: complete.

Owner:

  • .tmp/slate-v2/scripts/benchmarks/browser/react/rerender-breadth.tsx

Work completed:

  • Re-ran selector/depth/leaf/subscriber benchmark rows.
  • Confirmed the artifact still proves selector-locality rows without claiming broad editor hooks become narrow.
  • Confirmed broad hook non-claim policy remains correct for #5131.

Evidence:

  • From .tmp/slate-v2, bun run bench:react:rerender-breadth:local passed.
  • From .tmp/slate-v2, bun test ./packages/slate-react/test/provider-hooks-contract.tsx ./packages/slate-react/test/projections-and-selection-contract.tsx passed.
  • Artifact .tmp/slate-v2/packages/slate-react/tmp/slate-react-rerender-breadth-benchmark.json has percentile rows for sibling leaf breadth, deep ancestor breadth, broad selection renders, and source-scoped invalidation.

Issue claims:

  • No issue claim changed.
  • #5131 stays Not claimed while useSlate / useEditor remain broad by contract.
  • Selector APIs remain the intended scalable path.

Reference docs:

  • No matrix, dossier, or PR-reference update required. Existing rows already record #5131 as not claimed and point at the rerender breadth benchmark.

Next owner:

  • Slice 2C: projection and decoration churn.

Slice 2C - Projection And Decoration Churn - 2026-05-09

Status: complete.

Owner:

  • .tmp/slate-v2/scripts/benchmarks/browser/react/huge-document-overlays.tsx
  • .tmp/slate-v2/packages/slate-react/test/projections-and-selection-contract.tsx

Work completed:

  • Re-ran the huge-document overlays benchmark.
  • Confirmed source-scoped projection/decorations churn stays localized: unrelated active-edit and shell-promotion lanes kept projection recompute count at 0, while the overlay toggle lane recomputed exactly once.
  • Re-ran the source-focused projection contract tests.

Evidence:

  • From .tmp/slate-v2, bun run bench:react:huge-document-overlays:local passed.
  • From .tmp/slate-v2, bun test ./packages/slate-react/test/projections-and-selection-contract.tsx -t source passed.
  • Artifact .tmp/slate-v2/packages/slate-react/tmp/slate-react-huge-document-overlays-benchmark.json has percentile rows and the expected projection recompute counts: activeEditAfterOverlay.projectionRecomputeCount.mean = 0, overlayToggle.projectionRecomputeCount.mean = 1, and shellPromotion.projectionRecomputeCount.mean = 0.

Issue claims:

  • No issue claim changed.
  • No fixed issue claim from this local benchmark-only proof.

Reference docs:

  • No matrix, dossier, or PR-reference update required. This slice confirmed the planned projection/decorations proof without changing issue claims or public API shape.

Next owner:

  • Slice 2D: rendering strategy behavior.

Slice 2D - Rendering Strategy Behavior - 2026-05-09

Status: complete.

Owner:

  • .tmp/slate-v2/scripts/benchmarks/browser/react/huge-document-overlays.tsx
  • .tmp/slate-v2/playwright/integration/examples/dom-coverage-boundaries.test.ts

Work completed:

  • Re-ran the huge-document overlays benchmark as the rendering metrics proof.
  • Re-ran the DOM coverage boundaries browser row under Chromium for native behavior proof across find, placeholders, copy, select-all, drag selection, hidden model updates, and IME composition.
  • Kept rendering metrics separate from browser correctness.

Evidence:

  • From .tmp/slate-v2, bun run bench:react:huge-document-overlays:local passed.
  • From .tmp/slate-v2, PLAYWRIGHT_RETRIES=0 PLAYWRIGHT_WORKERS=1 bun run playwright playwright/integration/examples/dom-coverage-boundaries.test.ts --project=chromium passed with 7 Chromium rows and 1 mobile-touch row skipped.
  • From .tmp/slate-v2, bun check passed.

Issue claims:

  • No issue claim changed.
  • No large-document superiority claim. This slice did not compare against a matched legacy cohort with legacy chunking enabled.
  • No raw mobile/device claim. The skipped mobile-touch row is not raw-device proof.

Reference docs:

  • No matrix, dossier, or PR-reference update required. This slice confirmed the planned rendering/browser proof without changing issue claims or public API shape.

Next owner:

  • Slice 2E: clipboard large payload.

Slice 2E - Clipboard Large Payload - 2026-05-09

Status: complete.

Owner:

  • .tmp/slate-v2/scripts/benchmarks/core/current/clipboard-large-payload.mjs
  • .tmp/slate-v2/playwright/stress/generated-editing.test.ts
  • .tmp/slate-v2/packages/slate-react/src/editable/mutation-controller.ts

Work completed:

  • Re-ran the clipboard large payload core benchmark directly because bench:core:clipboard-large-payload:local is not listed in package.json.
  • Re-ran the #5945 issue benchmark wrapper.
  • Re-ran generated editing browser stress under Chromium.
  • Fixed the browser stress regression found during proof: full-block delete fast path now removes every fully selected sibling block instead of treating a multi-block start-to-start selection as a one-block deletion.

Evidence:

  • From .tmp/slate-v2, bun ./scripts/benchmarks/core/current/clipboard-large-payload.mjs passed.
  • From .tmp/slate-v2, bun run bench:slate:5945:issue passed with issue thresholds green: cutTwoBlocksEditMsP50 = 2.02ms, cutTwoBlocksMsP50 = 2.17ms, and operationCount = 1.
  • From .tmp/slate-v2, STRESS_FAMILIES=huge-document-cut PLAYWRIGHT_RETRIES=0 PLAYWRIGHT_WORKERS=1 bun run playwright playwright/stress/generated-editing.test.ts --project=chromium passed.
  • From .tmp/slate-v2, PLAYWRIGHT_RETRIES=0 PLAYWRIGHT_WORKERS=1 bun run playwright playwright/stress/generated-editing.test.ts --project=chromium passed with 24 Chromium rows.
  • From .tmp/slate-v2, bun check passed.
  • Benchmark artifact: .tmp/slate-v2/tmp/slate-clipboard-large-payload-benchmark.json.

Issue claims:

  • No Fixes #5945, Fixes #5992, or Fixes #4056 claim promoted from this slice.
  • Browser stress proof covers the huge-document cut behavior used by this slice, but the plan still requires explicit maintainer equivalence before closing those issue claims.

Reference docs:

  • No matrix, dossier, or PR-reference update required. This slice fixed runtime behavior and proved the planned benchmark/browser route without changing public API shape or issue-claim status.
  • Added solution note: docs/solutions/logic-errors/2026-05-09-slate-v2-full-block-delete-fast-path-must-stop-at-selection-boundaries.md.

Next owner:

  • Slice 2F: Safari backward selection.

Slice 2F - Safari Backward Selection - 2026-05-09

Status: complete.

Owner:

  • .tmp/slate-v2/playwright/stress/generated-editing.test.ts

Work completed:

  • Added a focused browser stress row for desktop WebKit backward selection.
  • The row starts from a real DOM caret in the richtext example, presses Shift+ArrowLeft, and asserts both DOM and model selection direction: anchor remains at offset 4, focus moves to offset 3, and selected text is s.
  • Kept the proof scoped to desktop WebKit selection behavior. No mobile, iOS keyboard, autocorrect, or IME claim is made.

Evidence:

  • From .tmp/slate-v2, STRESS_FAMILIES=webkit-backward-selection PLAYWRIGHT_RETRIES=0 PLAYWRIGHT_WORKERS=1 bun run playwright playwright/stress/generated-editing.test.ts --project=webkit passed.
  • From .tmp/slate-v2, bun check passed.

Issue claims:

  • No raw mobile/device claim.
  • No IME or composition claim.
  • Desktop WebKit backward-selection behavior now has a focused regression row.

Reference docs:

  • No matrix, dossier, or PR-reference update required. This slice added proof coverage without changing issue-claim status or public API shape.

Next owner:

  • Slice 2G: history retained memory.

Slice 2G - History Retained Memory - 2026-05-09

Status: complete.

Owner:

  • .tmp/slate-v2/scripts/benchmarks/core/current/history-retained-memory.mjs
  • .tmp/slate-v2/package.json
  • .tmp/slate-v2/scripts/benchmarks/README.md
  • docs/slate-v2/ledgers/issue-coverage-matrix.md
  • docs/slate-v2/ledgers/fork-issue-dossier.md

Work completed:

  • Added a current-only retained-memory benchmark for history payload pressure.
  • Added stable driver command: bun run bench:core:history-retained-memory:local.
  • The benchmark records history entry count, redo entry count, operation count, retained batch operation count, retained operation types, retained payload tags, retained JSON byte counts, and process.memoryUsage().heapUsed samples when available.
  • Synced #3752 from Related to Improves in the matrix/dossier while keeping exact memory-leak closure out of scope.

Evidence:

  • From .tmp/slate-v2, bun run bench:core:history-retained-memory:local passed.
  • Artifact: .tmp/slate-v2/tmp/slate-history-retained-memory.json.
  • fullDocumentReplaceChildren: one history entry, one retained replace_children operation, historyJsonBytes = 1585394.
  • rangeDeleteReplaceChildren: one history entry, one retained replace_children operation, historyJsonBytes = 775230.
  • Heap samples were captured from process.memoryUsage().heapUsed; local GC was unavailable, so heap samples remain calibration evidence, not thresholds.
  • From .tmp/slate-v2, bun check passed after bun lint:fix formatted the new benchmark file.

Issue claims:

  • #3752 is now Improves, not Fixes.
  • No exact memory-leak closure. Exact closure still needs an accepted leak threshold or detached DOM/node retention proof that matches the original report.

Reference docs:

  • Updated docs/slate-v2/ledgers/issue-coverage-matrix.md.
  • Updated docs/slate-v2/ledgers/fork-issue-dossier.md.

Next owner:

  • Slice 2H: legacy compare unblocker.

Slice 2H - Legacy Compare Unblocker - 2026-05-09

Action:

  • Repaired the normalization and huge-document compare runners so the current Slate v2 side uses source imports for full internal Editor runtime APIs, while legacy Slate keeps package imports.
  • Updated compare runner adapters for v2 transaction writes: tx.text.insert, tx.selection.set, and tx.fragment.insert.
  • Kept legacy /Users/zbeyens/git/slate setup/build separate from current Slate v2 correctness.

Files changed:

  • .tmp/slate-v2/scripts/benchmarks/core/compare/normalization.mjs
  • .tmp/slate-v2/scripts/benchmarks/core/compare/huge-document.mjs

Commands:

  • cd /Users/zbeyens/git/slate-v2 && bun run bench:core:normalization:compare:local
  • cd /Users/zbeyens/git/slate-v2 && bun run bench:core:huge-document:compare:local
  • cd /Users/zbeyens/git/slate-v2 && bun lint:fix
  • cd /Users/zbeyens/git/slate-v2 && bun check

Artifacts:

  • .tmp/slate-v2/tmp/slate-normalization-compare-benchmark.json
  • .tmp/slate-v2/tmp/slate-core-huge-document-benchmark.json

Compare evidence:

  • Normalization compare passed. Current v2 was faster on explicit adjacent-text normalization and explicit inline-flatten normalization; legacy was faster on insert-text-with-read-after-each.
  • Huge-document compare passed. Current v2 was faster on full-document text replacement and full-document fragment insertion; legacy was faster on simple start/middle typing and select-all.

Issue claims:

  • Unchanged. This slice repairs benchmark infrastructure and records compare evidence only.
  • No new Fixes or Improves claim.

Reference docs:

  • No ledger or PR-reference claim sync needed because issue claims did not change.

Next owner:

  • None for this umbrella plan. Mobile/IME leftovers remain bucketed by proof owner and require raw device/manual proof before exact closure.