miti99bot/plans/reports/fullstack-260415-1052-phase-04-retention-cron.md

Phase 04 — Retention Cron Implementation Report

Date: 2026-04-15 Status: DONE

Files Modified

• tests/fakes/fake-d1.js — enhanced with SQL-semantic SELECT/DELETE support (DISTINCT, ORDER BY ts DESC, LIMIT/OFFSET with bind params, DELETE WHERE id IN)
• src/modules/trading/retention.js — created; per-user + global trim handler, overridable caps
• src/modules/trading/index.js — imported trimTradesHandler, added crons array
• wrangler.toml — crons = ["0 17 * * *"]
• tests/modules/trading/retention.test.js — created; 9 tests covering all scenarios
• plans/260415-1010-d1-cron-infra/phase-04-retention-cron.md — Status → Complete, todos ticked
• plans/260415-1010-d1-cron-infra/plan.md — phase 04 → Complete

Tasks Completed

• retention.js with per-user + global trim (hybrid SELECT-then-DELETE approach)
• Caps exported as PER_USER_CAP=1000, GLOBAL_CAP=10000; overridable via optional arg for small-seed tests
• Cron entry wired in trading module (schedule: "0 17 * * *", name: "trim-trades")
• wrangler.toml schedule added
• 9 retention tests: per-user trim, small-user no-op, exact-cap no-op, multi-user, global trim, combined pass, idempotence, sql=null guard

Tests Status

• Unit tests: 169/169 passed (all files)
• Lint: clean (Biome + ESLint, 54 files)

Key Decision

fake-d1 is SQL-less — DELETE naively cleared entire table. Chose option (b): hybrid SELECT-IDs-then-DELETE-by-id-list in retention.js. Enhanced fake-d1 to support targeted DELETE WHERE id IN (?,...) and SELECT ... ORDER BY ts DESC LIMIT/OFFSET with bind param resolution (? tokens resolved against binds iterator). This keeps tests meaningful without adding better-sqlite3 dependency.

Concerns

None.