instrument(parser): establish strategy/work receipt foundation (#9444) by EffortlessSteven · Pull Request #7332 · EffortlessMetrics/perl-lsp-swarm

EffortlessSteven · 2026-08-12T06:00:04Z

Train position

Parser implementation train: #8700 — work-receipt W01
Controlling issue: #9444
Parent receipt controller: #7072

Restack ruling

This draft currently sits on the historical #6982 stored-checkpoint branch. Preserve its useful receipt work, but restack the W01 semantic slice off that dependency after #7292 and the live parser-work authority #8757/#8786 are current.

The restacked PR must not include #6982 checkpoint implementation merely because it is in this branch’s ancestry. Checkpoint-specific work moves to #7758. This PR is the shared receipt/invocation foundation needed by conservative fresh_full_parse as well as later incremental strategies.

Claim

Canonical parser operations carry one versioned per-strategy work receipt that records real parser invocations and authoritative live parser-work fields, with every other strategy-specific plane explicitly not_applicable, partial, or not yet instrumented rather than fabricated as zero.

Current correction

Review of the first candidate found false precision:

final AST population was used as nodes_constructed;
clone/patch/comparison fields were reported as zero without observing those operations;
checkpoint invalidation/decision fields could default to zero without a complete denominator;
unchanged-result observation/clone work was not represented.

The restack must remove those claims. #7757 completes AST work; #7758 completes checkpoint/no-op/observation work.

Intended changes

stable shared strategy IDs including fresh full parse, unchanged, full fallback, checkpoint-to-EOF, future exact sync/leaf patch, complete artifact reuse, and comparison-only analysis;
per-strategy applicability/completeness for work planes;
real full/recovery/validation parser invocation counts at actual entry points;
refactor(parser): charge core token, node, diagnostic, and skipped-input work through the live tracker #8786 core parser-work counters reused rather than duplicated;
typed instrument partial/failure/overflow state;
impossible-combination validation;
no unknown work represented as observed zero;
propagation seam that contract(parser): publish one canonical generation-bound ParseSnapshot #7001 can bind into ParseSnapshot without another source/config identity.

Boundary

This PR does not implement:

stored checkpoints or checkpoint decision accounting (feat(parser-incremental): persist generation-bound full lexer restart checkpoints #6982/instrument(parser-incremental): account checkpoint decisions and no-op observation work #7758);
AST construction/clone/patch/retention/comparison completion (instrument(parser): count AST construction, cloning, patching, and comparison at operation sites #7757; construction itself comes from refactor(parser): charge core token, node, diagnostic, and skipped-input work through the live tracker #8786);
suffix reuse (feat(parser-incremental): require exact state and token-window proof before suffix reuse #6986);
a no-full-parser fast path (feat(parser-incremental): admit a no-full-parser same-width integer-literal edit strategy #7092);
complete instrument(parser-incremental): emit truthful strategy and parser-work receipts #7072 propagation/convergence;
snapshot/configuration semantics (contract(parser): publish one canonical generation-bound ParseSnapshot #7001/api(parser): remove ignored parse configuration or make every field load-bearing #7080).

A fresh_full_parse receipt may be complete while checkpoint fields are non-applicable. That is required so parser reliability does not wait on incremental checkpoint work.

Validation

Use current equivalents of:

cargo fmt --all -- --check
cargo test --locked -p perl-parser-core --all-targets
cargo test --locked -p perl-parser --features incremental --all-targets
python3 scripts/ci/run_parser_integration.py
cargo clippy --locked -p perl-parser-core -p perl-parser --all-targets -- -D warnings
git diff --check

Closes #9444
Refs #7072
Refs #7757
Refs #7758
Refs #8171
Refs #7045

gemini-code-assist · 2026-08-12T06:00:08Z

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

coderabbitai · 2026-08-12T06:00:14Z

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 244673ec-e438-4a48-a8cd-a5219c795324

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

🔍 Trigger review

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

EffortlessSteven

Review verdict: not ready as work/benchmark authority. The receipt vocabulary separates the right categories, but several populated values are inferred constants or final-state sizes rather than observations of performed work. That is more dangerous than leaving them unknown because #7099 and claim cleanup will treat the receipt as ground truth. Instrument the actual construction, cloning, checkpoint-selection, and validation sites—or make unsupported fields explicitly unavailable—before publication.

EffortlessSteven · 2026-08-12T07:19:07Z

+    pub(crate) fn refresh_parse_output(&mut self) -> ParserInvocationReceipt {
        let mut parser = Parser::new(self.source());
        let parse_output = parser.parse_with_recovery();
+        let nodes_constructed = parse_output.ast.count_nodes();


[P1] ast.count_nodes() measures the size of the final retained tree, not how many nodes the parser constructed. Recovery/backtracking may construct and discard additional nodes, so this value cannot support a performed-work claim. The next line also clones the full AST into the compatibility mirror, while the receipt later reports nodes_cloned = 0. Instrument construction and publication cloning at their actual sites, or rename these to final-tree size and make clone work explicit/unknown.

EffortlessSteven · 2026-08-12T07:19:07Z

+            nodes_patched: 0,
+            nodes_compared_only: 0,
+            checkpoints_restored,
+            checkpoints_invalidated: 0,


[P1] This field is documented as rejected or invalidated checkpoint candidates but is hard-coded to zero. The selection loop can reject multiple stored checkpoints for source identity, timeout sensitivity, edit overlap, or restore validity before choosing an earlier one or falling back. Count those decisions with stable reason tokens and carry them into the receipt; otherwise the fallback/eligibility envelope is systematically understated.

EffortlessSteven · 2026-08-12T07:19:07Z

+        0,
+        state.source().len(),
+        state.tokens().len(),
+        state.parse_output().ast.count_nodes(),


[P1] The Unchanged path walks the entire AST to compute final_node_count, and then clones ParseOutput for the returned result, yet the receipt says zero analysis/validation work and zero cloned nodes. A no-op edit should either reuse stored counts and return an identity-backed snapshot, or account for this traversal and clone. As written, the supposedly zero-work path can be O(AST size) while reporting no work.

EffortlessSteven · 2026-08-21T15:41:19Z

Read-only stale-PR preparation (live base fdfe5f1): disposition PREREQUISITE_BOUND. The receipt concept remains relevant, but required parser authorities and consumers are not all landed; the current head also retains challenged zero/inferred measurements. Reconcile prerequisites and split the receipt ownership before rebuilding this slice. No mutation was made.

EffortlessSteven · 2026-08-21T21:24:56Z

Disposition (campaign #11869 stale-queue sweep, 2026-08-21): REPAIR_EXISTING_CANDIDATE

source basis: PR head vs current main 0fa96eedb; read-only sweep, no proof re-run
claim checked: parser strategy/work receipt foundation (instrument(parser-incremental): establish the strategy/work receipt and real invocation counters #9444)
main interaction: none superseding; restack expected once work-counter authority (refactor(parser): make the primary parser own one live operation context #8757 → refactor(parser): charge core token, node, diagnostic, and skipped-input work through the live tracker #8786) lands — explicit stack prerequisite movement
unique value if closed: truthful strategy/work receipt substrate absent from main
defect to repair in place: refresh candidate onto current receipt-model seams before further review; keep claims within the foundation slice
rationale: sole owner, no competing retained implementation found on main or open candidates
follow-up owner: instrument(parser-incremental): establish the strategy/work receipt and real invocation counters #9444
resume when: refactor(parser): make the primary parser own one live operation context #8757/refactor(parser): charge core token, node, diagnostic, and skipped-input work through the live tracker #8786 land (restack), or earlier for in-place seam repair

Idle >3h → reclaimable per umbrella #11869 queue policy.

EffortlessSteven · 2026-08-21T21:30:11Z

Disposition: REPAIR (campaign #11869 stale-PR sweep, 2026-08-21). Stacked on #7324 (itself conflicting with main); advancing standalone would build on a stale stack. Wake event: #7324 rebased and green; then rebase this slice.

EffortlessSteven · 2026-08-22T21:18:26Z

Disposition (campaign #11869 stale-queue sweep, 2026-08-22): REPAIR_EXISTING_CANDIDATE (stack-gated) — reconciles the two 2026-08-21 packets; dependency states re-verified.

source basis: head 0dc07e77 (+442/−126 across 7 files) vs current main
claim checked: parser strategy/work receipt foundation (#9444, open)
current disposition: REPAIR_EXISTING_CANDIDATE - repair in place via the recorded restack ruling once its gates open
current-main interaction: none superseding - no merged implementation owns strategy/work receipts (searched); restack is explicit stack-prerequisite movement
unique value: versioned per-strategy receipt vocabulary with per-plane applicability/completeness, real invocation counting at actual entry points, not_applicable/partial/not-yet-instrumented honesty rules, #8786 counter reuse instead of duplication
concrete defects/blockers: mergeable CONFLICTING on historical #6982 ancestry it must shed; false-precision fields recorded in body must be removed in the W01 slice; gate authorities #8757 and #8786 both still open (verified 2026-08-22), as are accounting leaves #7757/#7758
proof run: none this sweep (read-only)
review currentness: standing review current; restack refreshes affected dimensions only
base-reconciliation reason: deferred until gates land - then restack off checkpoint ancestry per the recorded ruling rather than rebasing the contaminated stack
merge/close rationale: none - truthful receipt substrate absent from main; issue open
follow-up: resume when #7324 is reconciled AND #8757/#8786 land; then deliver the narrowed W01 slice without checkpoint implementation

instrument(parser): emit truthful incremental work receipts

0dc07e7

EffortlessSteven commented Aug 12, 2026

View reviewed changes

EffortlessSteven changed the title ~~instrument(parser): add incremental strategy and work receipts (#7072)~~ instrument(parser): establish strategy/work receipt foundation (#9444) Aug 14, 2026

EffortlessSteven mentioned this pull request Aug 14, 2026

epic(parser): execute the production-baseline and incremental-admission PR train #8700

Open

14 tasks

claude Bot mentioned this pull request Aug 16, 2026

ops(pr): 36 of 150 open PRs predate the main re-root and cannot be merged, rebased, or updated #10205

Open

EffortlessSteven mentioned this pull request Aug 21, 2026

feat(parser): persist generation-bound lexer restart checkpoints (#6982) #7324

Draft

EffortlessSteven mentioned this pull request Aug 21, 2026

campaign: continuously drain actionable P0/P1 issues through clean reviewed PRs #11869

Open

Navigation Menu

Uh oh!

instrument(parser): establish strategy/work receipt foundation (#9444) - #7332

instrument(parser): establish strategy/work receipt foundation (#9444)#7332
EffortlessSteven wants to merge 1 commit into
agent/6982-stored-lexer-checkpointsfrom
agent/7072-truthful-incremental-work-receipts

EffortlessSteven commented Aug 12, 2026 •

edited

Loading

Uh oh!

gemini-code-assist Bot commented Aug 12, 2026

Uh oh!

coderabbitai Bot commented Aug 12, 2026

Review skipped

Uh oh!

EffortlessSteven left a comment

Uh oh!

EffortlessSteven Aug 12, 2026

Uh oh!

EffortlessSteven Aug 12, 2026

Uh oh!

EffortlessSteven Aug 12, 2026

Uh oh!

EffortlessSteven commented Aug 21, 2026

Uh oh!

EffortlessSteven commented Aug 21, 2026

Uh oh!

EffortlessSteven commented Aug 21, 2026

Uh oh!

EffortlessSteven commented Aug 22, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

EffortlessSteven commented Aug 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Train position

Restack ruling

Claim

Current correction

Intended changes

Boundary

Validation

Uh oh!

gemini-code-assist Bot commented Aug 12, 2026

Uh oh!

coderabbitai Bot commented Aug 12, 2026

Review skipped

Uh oh!

EffortlessSteven left a comment

Choose a reason for hiding this comment

Uh oh!

EffortlessSteven Aug 12, 2026

Choose a reason for hiding this comment

Uh oh!

EffortlessSteven Aug 12, 2026

Choose a reason for hiding this comment

Uh oh!

EffortlessSteven Aug 12, 2026

Choose a reason for hiding this comment

Uh oh!

EffortlessSteven commented Aug 21, 2026

Uh oh!

EffortlessSteven commented Aug 21, 2026

Uh oh!

EffortlessSteven commented Aug 21, 2026

Uh oh!

EffortlessSteven commented Aug 22, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

EffortlessSteven commented Aug 12, 2026 •

edited

Loading