feat(#64): 配置项 CRUD 存储引擎（模型超参/RAG/布局三类，文件系统版本化 JSON+校验）

2026-08-05 05:29:53 +08:00
parent 94c2782e75
commit b0ae378150
2 changed files with 482 additions and 0 deletions
@@ -0,0 +1,191 @@
+# -*- coding: utf-8 -*-
+"""配置项 CRUD 存储引擎测试（issue #64）。
+
+覆盖：
+1. 三类配置 CRUD（list/get/upsert/delete）；
+2. 原子写 + 持久化（重开 store 仍在）；
+3. 校验规则（model_param/rag_config/layout，非法值拒绝）；
+4. 快照 snapshot/restore（为 #66 提供基础）；
+5. 可解释字段（meaning/reason/updated_by/updated_at 落盘）。
+"""
+import json
+import os
+import sys
+import tempfile
+import unittest
+
+sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+import _bootstrap  # noqa: F401
+
+from template_console.config_store import (  # noqa: E402
+    ALLOWED_WIDGET_TYPES,
+    ConfigItem,
+    ConfigKind,
+    ConfigStore,
+    ValidationResult,
+    validate_item,
+)
+
+
+class _TmpStore:
+    def __init__(self):
+        self._tmp = tempfile.mkdtemp()
+        self.store = ConfigStore(self._tmp)
+
+    def cleanup(self):
+        import shutil
+        shutil.rmtree(self._tmp, ignore_errors=True)
+
+
+class ValidationTest(unittest.TestCase):
+    """校验规则。"""
+
+    def test_model_param_scalar_ok(self):
+        self.assertTrue(validate_item(ConfigKind.MODEL_PARAM, "learning_rate", 0.001))
+
+    def test_model_param_learning_rate_range(self):
+        vr = validate_item(ConfigKind.MODEL_PARAM, "learning_rate", 1.5)
+        self.assertFalse(vr)
+        self.assertTrue(any("learning_rate" in e for e in vr.errors))
+
+    def test_model_param_bad_key(self):
+        vr = validate_item(ConfigKind.MODEL_PARAM, "Bad Key!", 1)
+        self.assertFalse(vr)
+
+    def test_rag_top_k_bounds(self):
+        self.assertFalse(validate_item(ConfigKind.RAG_CONFIG, "top_k", 0))
+        self.assertFalse(validate_item(ConfigKind.RAG_CONFIG, "top_k", 51))
+        self.assertTrue(validate_item(ConfigKind.RAG_CONFIG, "top_k", 10))
+
+    def test_rag_similarity_threshold(self):
+        self.assertTrue(validate_item(ConfigKind.RAG_CONFIG, "similarity_threshold", 0.5))
+        self.assertFalse(validate_item(ConfigKind.RAG_CONFIG, "similarity_threshold", 1.5))
+
+    def test_rag_sources_must_be_nonempty_list(self):
+        self.assertFalse(validate_item(ConfigKind.RAG_CONFIG, "sources", []))
+        self.assertFalse(validate_item(ConfigKind.RAG_CONFIG, "sources", ["", "x"]))
+        self.assertTrue(validate_item(ConfigKind.RAG_CONFIG, "sources", ["sop", "gb"]))
+
+    def test_layout_widget_type(self):
+        bad = [{"type": "unknown", "x": 0, "y": 0, "w": 1, "h": 1}]
+        self.assertFalse(validate_item(ConfigKind.LAYOUT, "dashboard", bad))
+        good = [{"type": "trend", "x": 0, "y": 0, "w": 6, "h": 2}]
+        self.assertTrue(validate_item(ConfigKind.LAYOUT, "dashboard", good))
+
+    def test_layout_widget_coords_nonneg_int(self):
+        bad = [{"type": "trend", "x": -1, "y": 0, "w": 1, "h": 1}]
+        vr = validate_item(ConfigKind.LAYOUT, "dashboard", bad)
+        self.assertFalse(vr)
+
+
+class CrudTest(unittest.TestCase):
+    """CRUD + 持久化。"""
+
+    def setUp(self):
+        self.ctx = _TmpStore()
+        self.store = self.ctx.store
+
+    def tearDown(self):
+        self.ctx.cleanup()
+
+    def test_upsert_and_get(self):
+        self.store.upsert(ConfigKind.MODEL_PARAM, "iterations", 100,
+                          meaning="迭代数", updated_by="li", reason="标定")
+        it = self.store.get(ConfigKind.MODEL_PARAM, "iterations")
+        self.assertIsNotNone(it)
+        self.assertEqual(it.value, 100)
+        self.assertEqual(it.updated_by, "li")
+        self.assertEqual(it.reason, "标定")
+        self.assertTrue(it.updated_at)  # 时间戳已写
+
+    def test_upsert_rejects_invalid(self):
+        with self.assertRaises(ValueError):
+            self.store.upsert(ConfigKind.RAG_CONFIG, "top_k", 999)
+
+    def test_upsert_overwrites(self):
+        self.store.upsert(ConfigKind.MODEL_PARAM, "lr", 0.1)
+        self.store.upsert(ConfigKind.MODEL_PARAM, "lr", 0.01, reason="调小")
+        it = self.store.get(ConfigKind.MODEL_PARAM, "lr")
+        self.assertEqual(it.value, 0.01)
+        self.assertEqual(it.reason, "调小")
+
+    def test_list_and_delete(self):
+        self.store.upsert(ConfigKind.RAG_CONFIG, "top_k", 5)
+        self.store.upsert(ConfigKind.RAG_CONFIG, "similarity_threshold", 0.6)
+        self.assertEqual(len(self.store.list(ConfigKind.RAG_CONFIG)), 2)
+        self.assertTrue(self.store.delete(ConfigKind.RAG_CONFIG, "top_k"))
+        self.assertIsNone(self.store.get(ConfigKind.RAG_CONFIG, "top_k"))
+        self.assertFalse(self.store.delete(ConfigKind.RAG_CONFIG, "nope"))
+
+    def test_persistence_across_reopen(self):
+        self.store.upsert(ConfigKind.MODEL_PARAM, "lr", 0.001)
+        # 重开一个指向同一目录的 store
+        store2 = ConfigStore(self.ctx._tmp)
+        it = store2.get(ConfigKind.MODEL_PARAM, "lr")
+        self.assertIsNotNone(it)
+        self.assertEqual(it.value, 0.001)
+
+    def test_json_file_is_human_readable(self):
+        self.store.upsert(ConfigKind.MODEL_PARAM, "lr", 0.001, meaning="学习率")
+        path = os.path.join(self.ctx._tmp, "model_params.json")
+        with open(path, encoding="utf-8") as fh:
+            blob = json.load(fh)
+        self.assertEqual(blob["schema_version"], 1)
+        self.assertEqual(blob["kind"], "model_param")
+        self.assertEqual(blob["items"][0]["meaning"], "学习率")
+
+
+class SnapshotTest(unittest.TestCase):
+    """快照与恢复（#66 基础）。"""
+
+    def setUp(self):
+        self.ctx = _TmpStore()
+        self.store = self.ctx.store
+
+    def tearDown(self):
+        self.ctx.cleanup()
+
+    def test_snapshot_captures_all_kinds(self):
+        self.store.upsert(ConfigKind.MODEL_PARAM, "lr", 0.001)
+        self.store.upsert(ConfigKind.RAG_CONFIG, "top_k", 8)
+        snap = self.store.snapshot()
+        self.assertIn("captured_at", snap)
+        self.assertEqual(set(snap["kinds"].keys()),
+                         {"model_param", "rag_config", "layout"})
+        self.assertEqual(len(snap["kinds"]["model_param"]), 1)
+
+    def test_restore_replicates_state(self):
+        self.store.upsert(ConfigKind.MODEL_PARAM, "lr", 0.001)
+        self.store.upsert(ConfigKind.RAG_CONFIG, "top_k", 8)
+        snap = self.store.snapshot()
+        # 清空再恢复
+        self.store.delete(ConfigKind.MODEL_PARAM, "lr")
+        self.store.delete(ConfigKind.RAG_CONFIG, "top_k")
+        self.store.restore(snap)
+        self.assertEqual(self.store.get(ConfigKind.MODEL_PARAM, "lr").value, 0.001)
+        self.assertEqual(self.store.get(ConfigKind.RAG_CONFIG, "top_k").value, 8)
+
+    def test_item_counts(self):
+        self.store.upsert(ConfigKind.LAYOUT, "dashboard",
+                          [{"type": "trend", "x": 0, "y": 0, "w": 6, "h": 2}])
+        counts = self.store.item_counts()
+        self.assertEqual(counts["layout"], 1)
+        self.assertEqual(counts["model_param"], 0)
+
+
+class ConfigItemSerializationTest(unittest.TestCase):
+    """ConfigItem 序列化往返。"""
+
+    def test_roundtrip(self):
+        it = ConfigItem(key="lr", value=0.1, kind=ConfigKind.MODEL_PARAM,
+                        meaning="学习率", updated_by="li", reason="init",
+                        updated_at="2026-01-01T00:00:00Z")
+        d = it.to_dict()
+        self.assertEqual(d["kind"], "model_param")
+        it2 = ConfigItem.from_dict(d)
+        self.assertEqual(it2.value, 0.1)
+        self.assertEqual(it2.kind, ConfigKind.MODEL_PARAM)
+
+
+if __name__ == "__main__":
+    unittest.main()