feat/db-rework #3

Merged

jd merged 2 commits from feat/db-rework into feat/indexer

2026-01-20 01:43:49 +00:00

jd commented

2026-01-20 01:43:42 +00:00

Owner

No description provided.

jd added 2 commits

2026-01-20 01:43:42 +00:00

refactor(db): migrate from UUID to int64 primary keys across entire database layer 51422cefc0

This is a comprehensive refactoring that replaces UUID identifiers with auto-incrementing
`bigserial` primary keys across all database tables and associated Go code. The migration
includes schema changes, generated sqlc code updates, and corresponding application code
modifications.

Key changes:
- Schema: Changed all UUID PRIMARY KEY columns to BIGSERIAL PRIMARY KEY NOT NULL
- Schema: Added urlHash UUID column to Pages table for unique URL identification
- Models: Updated all struct types to use int64 instead of uuid.UUID for IDs
- New Types: Added dbtypes.NilInt64 for handling nullable bigserial references
- Generated Code: Added SQL query documentation comments via sqlc emit_sql_as_comment
- Application: Updated all database access functions to use int64 identifiers
- Cache Layer: Replaced UUID bytes with int64 binary encoding for groupcache keys
- Build: Removed github.com/google/uuid dependency from affected packages

refactor(db): modernize database schema with hash-based lookups and identity columns c3e07d892b

This comprehensive refactor introduces several significant database and application changes:

- Migrated from BIGSERIAL to BIGINT GENERATED ALWAYS AS IDENTITY for modern PostgreSQL identity columns
- Added domainHash UUID column for efficient O(1) domain lookups replacing string-based searches
- Replaced separate state/status columns with single PageStates lookup table containing 12 predefined state combinations
- Added computed score column for simplified page ranking queries
- Updated all SQL queries and Go models to use hash-based lookups
- Added uuidhash package using Blake2b for consistent 16-byte hash generation
- Changed sqlc configuration to emit pointer structs for better null handling
- Added manual InsertPageFast implementation with upsert logic using temporary tables

BREAKING CHANGE: Database schema requires migration. Domain lookups now use domainHash UUID instead of domain string. Page state tracking changed from separate state/status columns to single PageStatus reference. All client code must update to use new hash-based APIs.