Sprint 151 — Programmers SQL Problem Entry: Editor Language Auto SQL Selection

• Status: Completed ✅
• Period: 2026-05-13 (single day)
• Trigger: User feedback
• start_commit: 2f68402 (Sprint 150 end)
• end_commit: 4313561 (Sprint 151 hotfix end)
• Merged PRs: 2 (squash merge)

1. Background / User Feedback

When users enter Programmers SQL category problems (MySQL/Oracle), the editor default language was hardcoded to 'python', requiring manual switching to SQL every time — creating friction. User requested removal of this friction.

2. Decisions

2.1 Schema Decision

• New category enum column on Problem entity ('ALGORITHM' | 'SQL', default 'ALGORITHM', NOT NULL)
• Enum name: ProblemCategory (PascalCase, matching existing Difficulty/ProblemStatus pattern)
• Enum values: UPPER_SNAKE_CASE (backend standard)
• Migration timestamp: 1709000016000 (sequential, after existing 1709000015000)
• Reason: Same enum ('algorithm'/'sql') already exists in Gateway sync stage (Sprint 108) → SSOT consistency. More explicit than tags marker

2.2 UX Decision

• useRef guard on Problem mount for single-time setLanguage('sql') application
• After user manual change within same mount: not forced again (ref true)
• New page mount → new ref → SQL again (intended UX, matches user feedback)
• ALGORITHM or category undefined (legacy): keep existing 'python'

2.3 PR Strategy

• Single feature PR + separated atomic commits (backend → gateway sync → frontend)
• Dependencies clear, so splitting would cause frontend fallback debt → bundled

2.4 SQL Language Identifier

• Use existing 'sql' single value (LANGUAGES constant, Sprint 108)
• MySQL/Oracle split does not exist → simplified

3. Implementation Flow

3.1 PR #229 — Main PR (998ec4e)

11 commits, 13 files, +449/-10

Wave	Agent	Changed Files	Verification
Wave 1 (Backend)	architect (sonnet)	problem.entity.ts (+ProblemCategory enum) / create-problem.dto.ts (@IsEnum) / 1709000016000-AddCategoryToProblems.ts (migration) / problem.service.ts (inject in create/update)	problem 170 PASS (+6)
Wave 2 (Gateway sync)	architect (sonnet)	programmers.service.ts (toProblemCategoryEnum helper export, lowercase→uppercase conversion)	gateway 783 PASS (+4)
Wave 3 (Frontend READ)	palette (opus)	types.ts (Problem.category) / page.tsx (useRef + useEffect ref-guarded setLanguage) / __tests__/sql-auto-language.test.tsx (4 scenarios 270 lines)	frontend 1356 PASS (+4)
Wave 3 R2 (Frontend WRITE)	palette (opus)	types.ts (CreateProblemData.category + UpdateProblemData.category) / AddProblemModal.tsx:675 (SQL branch category: 'SQL' as const) / AddProblemModal.test.tsx (+2 verifications)	AddProblemModal 7/7 PASS

3.2 PR #230 — Trivy Hotfix (4313561)

2 commits, 4 lockfiles, +74/-58

Package	Before → After	CVE	Affected Services
Next.js	15.5.15 → 15.5.18	CVE-2026-44578 SSRF / CVE-2026-44579 DoS / GHSA-8h8q-6873-q5fj etc. (total 9 HIGH)	frontend
fast-uri	3.1.0 → 3.1.2	CVE-2026-6321 path traversal	frontend, gateway, submission, problem
fast-xml-builder	1.1.4 → 1.2.0	CVE-2026-44665 XML Comment/CDATA Injection	gateway (transitive of fast-xml-parser 5.5.6)

npm audit final: HIGH/CRITICAL = 0 across all services, package.json caret unchanged (lockfile only updated)

4. Critic Rounds (Auto-Critic Queued 4 Times)

4.1 Wave 1+2 Auto-Critic R1

• session: 019e1eeb-1fdb-71a0-879c-fea828bb6efb
• P0/P1 0, P2 1: category not passed in gateway → problem-service write path → risk of silent ALGORITHM storage for SQL problems
• Resolution: Found in Wave 3 R2 at exact entry point (AddProblemModal) and handled in bulk

4.2 Wave 3 Auto-Critic R1

• session: 019e1ef2-96f4-7953-9e68-87e5071c09f9
• P0/P1 0, P2 1: Same essence — AddProblemModal.tsx:675 passes allowedLanguages: ['sql'] but category: 'SQL' not passed
• Codex identified exact entry point (concrete example of Wave 1+2 P2)
• Resolution path specified: CreateProblemData type + AddProblemModal payload + test strengthening

4.3 Wave 3 R2 Auto-Critic

• session: 019e1efa-8ae4-7df3-859b-b7e189b884d9
• ✅ Critical/High/Medium 0, mergeable
• Low 2 (optional): test name mismatch, UpdateProblemData.category proactive expansion (neither blocking)
• Codex original: "I did not find a discrete regression or blocking issue in the modified lines"

4.4 Hotfix Auto-Critic

• session: 019e1f14-e88e-75a3-a266-1e7abfe8c8ba
• ✅ Mergeable, 0 introduced defects
• Additional observations: gateway [email protected] new transitive (NaturalIntelligence MIT, no risk), Next.js patch 3-step SWC binary all-platform consistent

5. Verification

5.1 PR #229 (Main PR)

• problem 170 / gateway 783 / frontend 1356 tests PASS
• ESLint / tsc --noEmit / Next.js + NestJS build clean
• CI: 30 pass / 0 fail / 9 skip, mergeStateStatus CLEAN
• merged at 2026-05-13T01:47:30Z, mergeCommit 998ec4e

5.2 Main Push CI Failure + Hotfix

• run 25773141824 (main push @ 998ec4e): Trivy Scan frontend + gateway FAILED (HIGH 9 + 1)
• Cause: New CVEs exposed by Trivy DB update (unrelated to Sprint 151 changes)
• Sprint 150 lesson #1 directly reproduced: paths filter PR-stage TRIVY SKIPPING → main push + image build + fresh DB exposure

5.3 PR #230 (Hotfix)

• frontend 1356 / gateway 783 tests PASS (no regressions)
• CI: 30 pass / 0 fail / 9 skip, MERGEABLE/CLEAN
• merged at 2026-05-13T02:15:21Z, mergeCommit 4313561

5.4 Main Push CI (After Hotfix) — Core Verification

• run 25774076614 (main push @ 4313561):
  • state=completed | conclusion=success
  • 38 jobs pass / 0 fail
  • Trivy Scan 8 services all SUCCESS (frontend/gateway/submission/problem/identity/ai-analysis/blog/github-worker)
• Main green restored ✅

6. Regression Blocking Core 8 Layers (Sprint 145~150 Cumulative Pattern Inherited)

#	Layer	Location
1	Backend entity enum	problem.entity.ts ProblemCategory
2	Backend DTO validation	create-problem.dto.ts @IsEnum
3	Backend migration	1709000016000-AddCategoryToProblems.ts
4	Backend service injection	problem.service.ts create/update
5	Gateway sync conversion	programmers.service.ts toProblemCategoryEnum()
6	Frontend type	types.ts (Problem + CreateProblemData + UpdateProblemData)
7	Frontend WRITE payload	AddProblemModal.tsx SQL branch category delivery
8	Frontend READ auto-selection	page.tsx ref-guarded useEffect

8th cumulative dimension — Sprint 145 (metric) → 146 (label) → 147 (panel-title+variable) → 148 (rule-label+dashboard-structure) → 149 (regex-robustness) → 150 (submission service problem context fallback) → 151 (Problem schema + Frontend integration 6 layers + Gateway conversion + Hotfix dependency security)

7. New Patterns

1. Auto-Critic R1 → R2 Same Essence Refinement Pattern — Wave 1+2 AC identified general issue (write path), Wave 3 AC identified exact entry point (AddProblemModal). Integrating two perspectives allowed reaching exact change location with single R2 delegation
2. Frontend WRITE/READ Separated Diagnosis — User feedback was READ scenario, but prerequisite is WRITE flow (admin registration storing category 'SQL'). This PR closes both sides, achieving core satisfaction
3. palette Entry Point Autonomous Evaluation — create/edit page evaluated as low user scenario frequency → outside this PR scope (avoids overscope, subsequent optional)
4. Sprint 150 Lesson #1 Direct Reproduction → Immediate Hotfix Standardization — Standard response to debt from paths filter PR-stage SKIP / main-stage execution jobs (Trivy) exposed post-main-push. This cycle is reproduction case + lockfile-only hotfix as response becomes standard workflow
5. Auto-Critic Auto-Queuing 4 Times in Short Cycle — Auto-queued for all Wave splits + R2 + hotfix in same PR. R1/R2 results immediately reflected without human intervention. Single day / 2 PRs / Auto-Critic 4 times / all clean finalized

8. Lessons

1. User feedback essence closes both sides of data flow, not just surface — "Auto SQL on SQL entry" is READ surface, but prerequisite WRITE flow must be closed for operation. Simultaneous analysis of surface + prerequisite required when handling user feedback
2. Paths filter bypass debt exposed only at main push time — immediate hotfix is standard — Sprint 150 lesson #1 verified. PR CI green does not guarantee main green. Pattern of immediately responding with separate hotfix PR to main push CI failure is standard (no wait/defer)
3. Dependency CVE exposure from Trivy DB update can be handled lockfile-only — package.json caret range unchanged, lockfile only updated via npm install/npm update to auto-apply patch-level. Near zero breaking change risk. Standard hotfix pattern
4. Auto-Critic R1 different perspectives refined in R2 — Wave 1+2 AC vs Wave 3 AC caught the same essence P2 from different perspectives (general vs exact entry point). Integrating two perspectives allows reaching exact change location with single R2 delegation (variation of Sprint 146 R1 1 → R2 2 additional catches pattern)
5. Auto-Critic Codex cross-verification value reconfirmed — All 4 invocations made accurate P2 catches or clean verdicts. P0/P1 0 + accurate essential non-fulfillment P2 identification → effectively served as pre-merge safety net. This cycle strongly validates Auto-Critic auto-queuing policy (Sprint 117~) effectiveness

9. Sprint 152 Carryover Seeds

UAT User Direct (Outside Oracle's Scope)

• Seed #5: Programmers resubmission scoring pass confirmation (8 sprints accumulated, SQL auto-selection UAT naturally included with this PR)
• Seed #9: English environment + production Grafana CB dashboard ai-analysis visual consistency

Follow-up (Optional, Outside this PR Scope)

• create/edit page.tsx category UI addition (palette evaluation: low user scenario frequency → outside this PR scope)
• Programmers URL automatic category inference (admin input friction reduction via sourceUrl pattern matching)
• Existing SQL problem data backfill (manual ADMIN work or import script — new SQL problems auto-applied)
• Sprint 150 remaining 3 candidates (.claude-tools/ Oracle dispatch tool cleanup / CI paths filter bypass debt check automation / prom-client default metric stale check)

10. Metrics

Item	Value
Period	2026-05-13 (single day)
origin/main	2f68402 → 4313561
Merged PRs	2 (#229 main / #230 hotfix)
Main PR commits	11 (squash 1)
Hotfix commits	2 (squash 1)
Changed files (main PR)	13 (+449 / -10)
Changed files (hotfix)	4 lockfiles (+74 / -58)
Auto-Critic invocations	4 (R1×3 + R2 hotfix) — all P0/P1 0
Regression blocking layers	8 (Sprint 145~150 cumulative 8th dimension)
Branch discipline	17 consecutive sprints compliant (0 direct commits to main)
Tests	problem 170 / gateway 783 / frontend 1356 PASS
Main CI final	run 25774076614 SUCCESS, 38 pass / 0 fail, Trivy 8 services all SUCCESS ✅