fix(bench): reject duplicate raw quads in UPDATE comparisons

[jeswr]

🤖 SPARQ agent — I am @jeswr's agent for the sparq-org/sparq RDF/SPARQL engine. @jeswr runs multiple agents; this was written by the SPARQ agent, not the PSS agent (prod-solid-server).

The UPDATE differential comparator can report Same when two outputs have equal raw row counts but redistribute repeated identical quads. Blank-node canonicalization collapses those repetitions before comparison. The strict equality path now rejects a repeated raw complete-quad line, reporting its side, original line and occurrence count while preserving the existing unequal-count diagnostic.

This addresses #6483 within the private complete-quad snapshot and fully projected probe contract. It does not change engine results, canonicalization or the semantics of general SPARQL result bags. Duplicate-free blank-node relabelling and the same triple in distinct named graphs remain accepted. The borrowed set adds O(N log N) comparisons and O(N) references on the strict equality path; no measured performance improvement is claimed.

Validation on the exact committed source: all 26 actual module tests passed, including six new boundary/regression tests and both-engine default/named-graph controls. Removing only the new guard compiled successfully and made the redistribution regression fail by observing Same. Scoped actual-module Clippy with warnings denied, touched formatting and diff checks passed. These native tests used pinned recorded dependencies; they do not replace the normal Linux workspace and feature gates. The first run was interrupted by a resource-monitor race during normal temporary-directory cleanup; the unchanged compiled candidate passed after a controller-only correction. Both runs' evidence is preserved. Local author preflight encountered the established Bash 3 mapfile limitation, so Linux privacy checks remain required.

Implementation: GPT-6 Astra with extra-high reasoning. Actual independent Claude Opus 5 with extra-high reasoning approved commit 0b4554b924a80432cc1b572bd19f8e58cdbfb4e6 for normal CI, with no blocking findings. The issue stays open until the fix lands and relevant post-merge evidence is verified.

Benchmark

Local sparq-cli bench, operators suite (2,000 entities, 16k triples), release-fast binaries of main 4105e5489d and this PR. Best of 5 iterations per round, minimum over 5 interleaved rounds; row counts match on every query. Every query is within noise of main; the geomean ratio is 1.002.

query	main (µs)	PR (µs)	PR/main
q01_bgp	3.0	3.0	1.000
q02_star3	10.0	10.0	1.000
q03_chain	276.9	264.9	0.957
q04_triangle	4568.6	4696.3	1.028
q05_union	6.8	6.6	0.971
q06_optional	55.5	54.6	0.984
q07_optional_notbound	31.0	31.4	1.013
q08_minus	15.0	15.0	1.000
q09_filter_numeric	195.8	193.2	0.987
q10_filter_string	867.7	880.5	1.015
q11_filter_in	279.2	283.8	1.016
q12_filter_exists	494.8	496.2	1.003
q13_bind	893.7	895.1	1.002
q14_values	10.0	9.8	0.980
q15_agg_group_having	327.4	332.0	1.014
q16_distinct	9.8	9.5	0.969
q17_orderby_limit_offset	19.0	19.4	1.021
q18_path_plus	697.5	721.4	1.034
q19_path_star	1113.2	1117.5	1.004
q20_path_opt	7.2	7.3	1.014
q21_path_seq	9.3	9.7	1.043
q22_path_alt	6.3	6.4	1.016
q23_path_inverse	6.7	6.6	0.985
q24_path_negated_pset	6.6	6.5	0.985
q25_subquery	458.4	465.4	1.015
q26_ask	79.0	80.2	1.015
q27_construct	13.8	13.8	1.000
q28_describe	6.0	6.0	1.000
geomean			1.002

---

Generated by Claude Code

[Copilot]

🟢 Approval recommended

The change is narrowly scoped to the bench comparator, is well-defended by targeted tests, and does not alter engine semantics or canonicalization behavior.

Pull request overview

Tightens the sparq-bench UPDATE differential comparator to prevent false Verdict::Same outcomes when raw snapshot/probe output redistributes identical duplicate N-Quads lines that get collapsed by canonicalization.

Changes:

• Added strict detection/rejection of repeated raw N-Quads lines on the equality fast-path (after canonical forms match and raw totals match).
• Extended module documentation to clarify the comparator’s “unique emission” contract is specific to complete-quad snapshots / fully-projected probes (not general SPARQL result bags).
• Added focused regression and contract tests covering duplicate redistribution, self-comparison invalidity, probe scope/projection, and preservation of canonicalization error behavior.

File summaries

File	Description
crates/sparq-bench/src/update_fuzz.rs	Adds raw-duplicate rejection to the comparator’s Same fast-path and introduces regression/contract tests to prevent future false-equality.

Review details

• Files reviewed: 1/1 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[github-actions]

sparq engine

Details

Benchmark suite	Current: 1e800e1	Previous: 674f50b	Ratio
store_bytes_per_triple	88 bytes	88 bytes	1
dict_bytes_per_term	53 bytes	53 bytes	1
comp_store_bytes_per_triple	44 bytes	44 bytes	1
store_bytes_per_triple_small	88 bytes	88 bytes	1
wasm_bundle_bytes	1562916 bytes	1562916 bytes	1

This comment was automatically generated by workflow using github-action-benchmark.

[jeswr]

The gate check (ci-summary) timed out with 17 sibling checks still queued. The runner pool is saturated, with about 230 queued runs repo-wide. No test failed on this head. This is the repository-wide merge-path stall that the maintainer's roadmap work is replacing with a fast required check, so I'm not re-running gate into the same queue. The PR stays watched.

---

Generated by Claude Code

[jeswr]

🔎 Codex reviewer — gpt-6.1-sol

Automated review by the Codex reviewer (OpenAI gpt-6.1-sol via Codex CLI) of head 1e800e15db57. Scope: correctness, security, soundness and design; style nits omitted. A new head gets a fresh review.

Findings

No correctness, security, soundness or design problems found.

Verdict: The PR is safe to merge as is.

[jeswr]

🔎 Codex reviewer — gpt-6.1-sol

Automated review by the Codex reviewer (OpenAI gpt-6.1-sol via Codex CLI) of head 301bf5c2ae47. Scope: correctness, security, soundness and design; style nits omitted. A new head gets a fresh review.

Findings

No correctness, security, soundness or design problems found.

Verdict: The PR is safe to merge as is.