From dfad37536b22b080c3a10fdcc581c9fe2020127c Mon Sep 17 00:00:00 2001
From: Hide_D <hided62@gmail.com>
Date: Sat, 27 Dec 2025 13:26:54 +0000
Subject: [PATCH] =?UTF-8?q?feat:=20=ED=85=8C=EC=8A=A4=ED=8A=B8=20=EC=A0=95?=
 =?UTF-8?q?=EC=B1=85=20=EB=AC=B8=EC=84=9C=20=EC=B6=94=EA=B0=80=20=EB=B0=8F?=
 =?UTF-8?q?=20=EA=B0=81=20=ED=85=8C=EC=8A=A4=ED=8A=B8=20=EB=A0=88=EC=9D=B4?=
 =?UTF-8?q?=EC=96=B4=EC=97=90=20=EB=8C=80=ED=95=9C=20=EC=A0=84=EB=9E=B5=20?=
 =?UTF-8?q?=EC=A0=95=EB=A6=AC?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 docs/architecture/overview.md |  1 +
 docs/testing-policy.md        | 55 +++++++++++++++++++++++++++++++++++
 2 files changed, 56 insertions(+)
 create mode 100644 docs/testing-policy.md

diff --git a/docs/architecture/overview.md b/docs/architecture/overview.md
index 22c8802..352435d 100644
--- a/docs/architecture/overview.md
+++ b/docs/architecture/overview.md
@@ -42,6 +42,7 @@ monorepo plan is prepared alongside it.
 - No ad-hoc randomness for gameplay; use deterministic RNG
 - Keep domain logic independent of endpoints or UI
 - Prefer clear Korean comments in core gameplay logic for maintainers
+- Test strategy and layering: `docs/testing-policy.md`
 
 ## Runtime Processing (Outline)
 
diff --git a/docs/testing-policy.md b/docs/testing-policy.md
new file mode 100644
index 0000000..e0f0715
--- /dev/null
+++ b/docs/testing-policy.md
@@ -0,0 +1,55 @@
+# Testing Policy (Draft)
+
+This document summarizes the testing strategy for the Sammo/HiDCHe repository.
+The key is to separate "DB behavior emulation" from "state/logic/flow validation"
+so each layer is tested with the right scope.
+
+## Core Principles
+- Prefer Repository/DB Port interfaces over ORM mocks; split implementations:
+  - InMemory Repository (Fake)
+  - Real DB Repository (e.g., Prisma)
+- Validate domain constraints in the logic layer where possible; DB constraints
+  act as a secondary safety net.
+- Use deterministic seeds for all gameplay-impacting RNG.
+
+## Test Layers
+
+### 1) Constraint Tests
+Goal: ensure constraints (unique/foreign key/check) behave consistently in
+both InMemory state and the DB.
+
+- Unit tests: validate domain constraints with InMemory repositories.
+- Integration tests: verify the same violations fail in the real DB.
+- Mock target: InMemory Repository (Fake).
+- Use real DB only in integration tests.
+
+### 2) Game Logic / Command Tests
+Goal: verify state input -> state output and that flush behaves as expected.
+
+- Unit tests: run logic against InMemory state and assert outputs.
+- Integration tests: execute the same command and confirm DB persistence.
+- Mock target: InMemory Repository (Fake).
+- "Send to DB" behavior is validated via real DB tests.
+
+### 3) Turn Flow Tests
+Goal: verify the end-to-end scheduler/turn processing flow.
+
+- Nature: integration/system tests.
+- Stack: Real DB + (test) Redis.
+- RNG uses fixed seeds for determinism.
+- Keep a small set of smoke/regression scenarios due to cost.
+
+### 4) Scenario Build Tests
+Goal: ensure scenario parsing and DB application are correct.
+
+- Unit tests: parse/validate scenario files (map size, ID collisions, ranges).
+- Integration tests: load scenarios into a test DB and verify key tables.
+
+## Result Composition Guidelines
+- Unit tests: fast, wide coverage.
+- Integration tests: focus on real DB/Redis behavior.
+- Smoke tests: minimal coverage of turn flow/build/scenario loading.
+
+## MockDB Conclusion
+- "InMemory Repository (Fake)" is more practical than an ORM mock.
+- DB-level behavior (constraints/transactions/locks) belongs to integration tests.