Add comprehensive documentation for the blocked_issues_cache optimization
that improved GetReadyWork performance from 752ms to 29ms (25x speedup).
Documentation locations:
- blocked_cache.go: Detailed package comment covering architecture,
invalidation strategy, transaction safety, edge cases, and future
optimizations
- ready.go: Enhanced comment at query site explaining the optimization
and maintenance triggers
- ARCHITECTURE.md: New section with diagrams, blocking semantics,
performance characteristics, and testing instructions
Closes bd-1w6i
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Introduces a materialized cache table to store blocked issue IDs,
replacing the expensive recursive CTE computation that was causing
~752ms query times on 10K databases (bd-5qim).
The cache is maintained via invalidation on dependency and status
changes, reducing GetReadyWork from O(n²) recursive traversal to
O(1) cache lookup.
Technical details:
- New blocked_issues_cache table with single issue_id column
- ON DELETE CASCADE ensures automatic cleanup
- Migration populates cache using existing recursive CTE logic
- rebuildBlockedCache() fully rebuilds cache on invalidation
- execer interface allows both *sql.DB and *sql.Tx usage
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
