fix: bound mixed-language incremental index work #494
No reviewers
Labels
No labels
code-review
correctness
dos
performance
security
severity/high
severity/low
severity/medium
tech-debt
Kind/Breaking
Kind/Bug
Kind/Documentation
Kind/Enhancement
Kind/Feature
Kind/Security
Kind/Testing
Priority
Critical
Priority
High
Priority
Low
Priority
Medium
Reviewed
Confirmed
Reviewed
Duplicate
Reviewed
Invalid
Reviewed
Won't Fix
Status
Abandoned
Status
Blocked
Status
Need More Info
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
h-dv/code-index!494
Loading…
Reference in a new issue
No description provided.
Delete branch "fix/scoped-resolve-cost"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Editing a large Rust file in a project with C# and Ruby packages could spend minutes invalidating namespace imports before resolution started, then multiply scoped constant lookup by every admitted contribution. Compute the namespace-invalidation language set once and probe constant-lookup admission after references only on scoped passes. Full passes retain their existing indexed admission plan. Capability, generation, scope and namespace predicates are preserved; no schema or dependency changes.
Validation: 704 selected indexer tests passed (four ignored), including the unchanged seven-corpus cold-work gate with the required-corpus setting, Ruby lookup, scoped/full equivalence, capability and generation isolation, and the admission registry; Clippy with warnings denied and formatting passed. Both new SQLite work-bound regressions fail when their fixes are reverted (about 15-fold work growth for fourfold input growth). On the real daemon with all 11 official packages, the large-file edit returned to ready in 87.9 seconds versus the published baseline still pending after 600 seconds. Every one of 486,008 active reference targets and deciding rules matched before and after; the new declaration was independently found, with no parse errors or stage-budget degradation.
The cold-work gate caught applying the probe ordering too broadly. Restricting it to scoped passes restores Rust full scans to about 514,005 and Ruby to 247,034, within the existing baselines. No thresholds or baselines were changed.
Refs #53. This fixes the two reproduced fan-outs; the broader resolver cost-model acceptance remains open.