This repository was archived by the owner on Sep 25, 2026. It is now read-only.
Repository navigation
Improve ranked phrase pruning and support benchmark regex queries - #201
Merged
Merged
Conversation
Add bounded phrase score-floor sampling and selective conjunction seeks while preserving exact rankings and existing index formats. Same-index Cascade Lake measurements improve RGB low-phrase top-10 from 8,706 to 22,205 QPS and medium-phrase top-100 from 8,112 to 11,685 QPS; retain measured regressions and ARM controls. Include validation and reproducible evidence, update the post comparison table, and remove infrastructure identifiers from public documentation and evidence.
Share bounded term-pattern execution with wildcard filters, expose RegexQuery and explicit regex syntax, and preserve punctuation/escaped operators through field analysis. The real HTTP capability probe accepts all 826 expressions, up from 801, with all previously accepted responses unchanged. Validate 2,054 search-stack tests, async-only regex/wildcard regressions and 41 WASM tests. Full-corpus analyzer, sloppy-phrase and expansion-limit parity remains separate work; existing performance tables still cover 15 queries.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Ranked RGB phrase queries can discover competitive scores too late to prune expensive position checks. Add a bounded pilot that proves a score floor from real matches, tighter exact-phrase frequency bounds, and selective conjunction seeks. Stable-ID ties, exhaustive counts, cancellation, eligibility filtering and existing index formats are preserved.
Also support the benchmark's 13 regex expressions and 12 rejected punctuation/escaped-literal expressions. RegexQuery and explicit regex syntax share wildcard's bounded dictionary matching and existing constant-score union scorer. The real HTTP capability probe now accepts 826/826 expressions on the unchanged four-document fixture, with all 801 previously accepted responses unchanged. Full-corpus analyzer, sloppy-phrase and expansion-limit parity remains separate work.
Same-index Cascade Lake measurements improve RGB low-frequency phrase top-10 from 8,706 to 22,205 QPS (+155%) and medium-frequency phrase top-100 from 8,112 to 11,685 QPS (+44%). Ordinary conjunction top-10 improves 3.1% but still trails Luxir. The evidence includes regressions, ARM controls, memory, CPU costs and exact result audits. These timings precede the query-syntax additions and cover 15/826 queries; no throughput claim is made for the newly supported families.
Includes reproducible evidence, a post-ready comparison table, and removal of infrastructure identifiers from public docs and bundled evidence. No schema change or rebuild is required for the query improvements.
Validation: search harness passes all five stages, 2,054 tests (25 ignored), strict Clippy and feature checks. All eight focused regex/wildcard tests pass with async-only native features. WASM release build and 41 JavaScript tests pass. Documentation links, archive/member checksums, unchanged benchmark analysis after evidence sanitization, and 826 HTTP capability responses are verified. Full lifecycle/RPC checks were not repeated because those paths are unchanged.