This commit is contained in:
Steve Yegge
2025-11-24 01:17:35 -08:00
4 changed files with 315 additions and 76 deletions

File diff suppressed because one or more lines are too long

View File

@@ -192,6 +192,150 @@ Hash-based comparison (not mtime) prevents git pull false positives (issue bd-84
- **Memory overhead:** One goroutine + minimal channel buffers per command execution
- **Flush latency:** Debounce duration + JSONL write time (typically <100ms for incremental)
## Blocked Issues Cache (bd-5qim)
### Problem Statement
The `bd ready` command originally computed blocked issues using a recursive CTE on every query. On a 10K issue database, each query took ~752ms, making the command feel sluggish and impractical for large projects.
### Solution: Materialized Cache Table
The `blocked_issues_cache` table materializes the blocking computation, storing issue IDs for all currently blocked issues. Queries now use a simple `NOT EXISTS` check against this cache, completing in ~29ms (25x speedup).
### Architecture
```
┌─────────────────────────────────────────────────────────┐
│ GetReadyWork Query │
│ │
│ SELECT ... FROM issues WHERE status IN (...) │
│ AND NOT EXISTS ( │
│ SELECT 1 FROM blocked_issues_cache │
│ WHERE issue_id = issues.id │
│ ) │
│ │
│ Performance: 29ms (was 752ms with recursive CTE) │
└─────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────┐
│ Cache Invalidation Triggers │
│ │
│ 1. AddDependency (blocks/parent-child only) │
│ 2. RemoveDependency (blocks/parent-child only) │
│ 3. UpdateIssue (on any status change) │
│ 4. CloseIssue (changes status to closed) │
│ │
│ NOT triggered by: related, discovered-from deps │
└─────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────┐
│ Cache Rebuild Process │
│ │
│ 1. DELETE FROM blocked_issues_cache │
│ 2. INSERT INTO blocked_issues_cache │
│ WITH RECURSIVE CTE: │
│ - Find directly blocked issues (blocks deps) │
│ - Propagate to children (parent-child deps) │
│ 3. Happens in same transaction as triggering change │
│ │
│ Performance: <50ms full rebuild on 10K database │
└─────────────────────────────────────────────────────────┘
```
### Blocking Semantics
An issue is blocked if:
1. **Direct blocking**: Has a `blocks` dependency on an open/in_progress/blocked issue
2. **Transitive blocking**: Parent is blocked and issue is connected via `parent-child` dependency
Closed issues never block others. Related and discovered-from dependencies don't affect blocking.
### Cache Invalidation Strategy
**Full rebuild on every change**
Instead of incremental updates, the cache is completely rebuilt (DELETE + INSERT) on any triggering change. This approach is chosen because:
- Rebuild is fast (<50ms even on 10K issues) due to optimized CTE
- Simpler implementation with no risk of partial/stale updates
- Dependency changes are rare compared to reads
- Guarantees consistency - cache matches database state exactly
**Transaction safety**
All cache operations happen within the same transaction as the triggering change:
- Uses transaction if provided, otherwise direct db connection
- Cache can never be in an inconsistent state visible to queries
- Foreign key CASCADE ensures cache entries deleted when issues are deleted
**Selective invalidation**
Only `blocks` and `parent-child` dependencies trigger rebuilds since they affect blocking semantics. Related and discovered-from dependencies don't trigger invalidation, avoiding unnecessary work.
### Performance Characteristics
**Query performance (GetReadyWork):**
- Before cache: ~752ms (recursive CTE)
- With cache: ~29ms (NOT EXISTS)
- Speedup: 25x
**Write overhead:**
- Cache rebuild: <50ms
- Only triggered on dependency/status changes (rare operations)
- Trade-off: slower writes for much faster reads
### Edge Cases
1. **Parent-child transitive blocking**
- Children of blocked parents are automatically marked as blocked
- Propagates through arbitrary depth hierarchies (limited to depth 50 for safety)
2. **Multiple blockers**
- Issue blocked by multiple open issues stays blocked until all are closed
- DISTINCT in CTE ensures issue appears once in cache
3. **Status changes**
- Closing a blocker removes all blocked descendants from cache
- Reopening a blocker adds them back
4. **Dependency removal**
- Removing last blocker unblocks the issue
- Removing parent-child link unblocks orphaned subtree
5. **Foreign key cascades**
- Cache entries automatically deleted when issue is deleted
- No manual cleanup needed
### Testing
Comprehensive test coverage in `blocked_cache_test.go`:
- Cache invalidation on dependency add/remove
- Cache updates on status changes
- Multiple blockers
- Deep hierarchies
- Transitive blocking via parent-child
- Related dependencies (should NOT affect cache)
Run tests: `go test -v ./internal/storage/sqlite -run TestCache`
### Implementation Files
- `internal/storage/sqlite/blocked_cache.go` - Cache rebuild and invalidation
- `internal/storage/sqlite/ready.go` - Uses cache in GetReadyWork queries
- `internal/storage/sqlite/dependencies.go` - Invalidates on dep changes
- `internal/storage/sqlite/queries.go` - Invalidates on status changes
- `internal/storage/sqlite/migrations/015_blocked_issues_cache.go` - Schema and initial population
### Future Optimizations
If rebuild becomes a bottleneck in very large databases (>100K issues):
- Consider incremental updates for specific dependency types
- Add indexes to dependencies table for CTE performance
- Implement dirty tracking to avoid rebuilds when cache is unchanged
However, current performance is excellent for realistic workloads.
## Future Improvements
Potential enhancements for multi-agent scenarios:

View File

@@ -1,3 +1,89 @@
// Package sqlite provides the blocked_issues_cache optimization for GetReadyWork performance.
//
// # Performance Impact
//
// GetReadyWork originally used a recursive CTE to compute blocked issues on every query,
// taking ~752ms on a 10K issue database. With the cache, queries complete in ~29ms
// (25x speedup) by using a simple NOT EXISTS check against the materialized cache table.
//
// # Cache Architecture
//
// The blocked_issues_cache table stores issue_id values for all issues that are currently
// blocked. An issue is blocked if:
// - It has a 'blocks' dependency on an open/in_progress/blocked issue (direct blocking)
// - Its parent is blocked and it's connected via 'parent-child' dependency (transitive blocking)
//
// The cache is maintained automatically by invalidating and rebuilding whenever:
// - A 'blocks' or 'parent-child' dependency is added or removed
// - Any issue's status changes (affects whether it blocks others)
// - An issue is closed (closed issues don't block others)
//
// Related and discovered-from dependencies do NOT trigger cache invalidation since they
// don't affect blocking semantics.
//
// # Cache Invalidation Strategy
//
// On any triggering change, the entire cache is rebuilt from scratch (DELETE + INSERT).
// This full-rebuild approach is chosen because:
// - Rebuild is fast (<50ms even on 10K databases) due to optimized CTE logic
// - Simpler implementation than incremental updates
// - Dependency changes are rare compared to reads
// - Guarantees consistency - no risk of partial/stale updates
//
// The rebuild happens within the same transaction as the triggering change, ensuring
// atomicity and consistency. The cache can never be in an inconsistent state visible
// to queries.
//
// # Transaction Safety
//
// All cache operations support both transaction and direct database execution:
// - rebuildBlockedCache accepts optional *sql.Tx parameter
// - If tx != nil, uses transaction; otherwise uses direct db connection
// - Cache invalidation during CreateIssue/UpdateIssue/AddDependency happens in their tx
// - Ensures cache is always consistent with the database state
//
// # Performance Characteristics
//
// Query performance (GetReadyWork):
// - Before cache: ~752ms (recursive CTE on 10K issues)
// - With cache: ~29ms (NOT EXISTS check)
// - Speedup: 25x
//
// Write overhead:
// - Cache rebuild: <50ms (full DELETE + INSERT)
// - Only triggered on dependency/status changes (rare operations)
// - Trade-off: slower writes for much faster reads
//
// # Edge Cases Handled
//
// 1. Parent-child transitive blocking:
// - Children of blocked parents are automatically marked as blocked
// - Propagates through arbitrary depth hierarchies (limited to depth 50)
//
// 2. Multiple blockers:
// - Issue blocked by multiple open issues stays blocked until all are closed
// - DISTINCT in CTE ensures issue appears once in cache
//
// 3. Status changes:
// - Closing a blocker removes all blocked descendants from cache
// - Reopening a blocker adds them back
//
// 4. Dependency removal:
// - Removing last blocker unblocks the issue
// - Removing parent-child link unblocks orphaned subtree
//
// 5. Foreign key cascades:
// - Cache entries automatically deleted when issue is deleted (ON DELETE CASCADE)
// - No manual cleanup needed
//
// # Future Optimizations
//
// If rebuild becomes a bottleneck in very large databases (>100K issues):
// - Consider incremental updates for specific dependency types
// - Add indexes to dependencies table for CTE performance
// - Implement dirty tracking to avoid rebuilds when cache is unchanged
//
// However, current performance is excellent for realistic workloads.
package sqlite
import (

View File

@@ -82,7 +82,16 @@ func (s *SQLiteStorage) GetReadyWork(ctx context.Context, filter types.WorkFilte
orderBySQL := buildOrderByClause(sortPolicy)
// Use blocked_issues_cache for performance (bd-5qim)
// Cache is maintained by invalidateBlockedCache() called on dependency/status changes
// This optimization replaces the recursive CTE that computed blocked issues on every query.
// Performance improvement: 752ms → 29ms on 10K issues (25x speedup).
//
// The cache is automatically maintained by invalidateBlockedCache() which is called:
// - When adding/removing 'blocks' or 'parent-child' dependencies
// - When any issue status changes
// - When closing any issue
//
// Cache rebuild is fast (<50ms) and happens within the same transaction as the
// triggering change, ensuring consistency. See blocked_cache.go for full details.
// #nosec G201 - safe SQL with controlled formatting
query := fmt.Sprintf(`
SELECT i.id, i.content_hash, i.title, i.description, i.design, i.acceptance_criteria, i.notes,