test: 真实文件切片 fixture（#34 遗留项）

tests/test_reader.py

@@ -65,6 +65,55 @@ def test_missing_file_raises(self, tdx_reader):
             tdx_reader.read_daily_data(0, 'sz999999')
 
 
+@pytest.fixture
+def real_tdx_reader(tmp_path):
+    """真实文件切片（dd bs=32 count=3 自实际 vipdoc，2026-07-07）——
+    防合成 fixture 与解析器共享同一格式误解"""
+    for market, sub, src, dst in [
+        ('sz', 'lday', 'real_sz000001.day', 'sz000001.day'),
+        ('sh', 'lday', 'real_sh688001.day', 'sh688001.day'),
+        ('sz', 'fzline', 'real_sz000001.lc5', 'sz000001.lc5'),
+    ]:
+        d = tmp_path / 'vipdoc' / market / sub
+        d.mkdir(parents=True, exist_ok=True)
+        shutil.copy(FIXTURES / src, d / dst)
+    return TdxDataReader(tdx_path=str(tmp_path))
+
+
+class TestRealFileSlices:
+    """断言值为真实文件的观测值（历史数据永不变化，天然稳定）"""
+
+    def test_sz000001_1991_head(self, real_tdx_reader):
+        df = real_tdx_reader.read_daily_data(0, 'sz000001')
+        assert list(df.index.strftime('%Y-%m-%d')) == ['1991-04-03', '1991-04-04', '1991-04-05']
+        assert df.iloc[0]['close'] == pytest.approx(49.00, abs=1e-9)
+        assert df.iloc[1]['close'] == pytest.approx(48.76, abs=1e-9)
+        assert df.iloc[0]['amount'] == 5000.0
+        assert df.iloc[0]['volume'] == 1.0
+
+    def test_sh688001_2019_head_via_fallback(self, real_tdx_reader):
+        """688001 华兴源创（科创板第一股）上市首日起 3 条，走 raw 回退路径"""
+        df = real_tdx_reader.read_daily_data(1, 'sh688001')
+        assert list(df.index.strftime('%Y-%m-%d')) == ['2019-07-22', '2019-07-23', '2019-07-24']
+        first = df.iloc[0]
+        assert first['open'] == pytest.approx(55.40, abs=1e-9)
+        assert first['high'] == pytest.approx(72.02, abs=1e-9)
+        assert first['low'] == pytest.approx(39.59, abs=1e-9)
+        assert first['close'] == pytest.approx(55.50, abs=1e-9)
+        assert first['amount'] == pytest.approx(1.507398e9, rel=1e-6)   # float32 精度
+        assert first['volume'] == pytest.approx(290107.54, rel=1e-6)
+
+    def test_lc5_real_head(self, real_tdx_reader):
+        df = real_tdx_reader.read_5min_data(0, 'sz000001')
+        assert list(df['datetime'].dt.strftime('%Y-%m-%d %H:%M')) == [
+            '2025-01-21 09:35', '2025-01-21 09:40', '2025-01-21 09:45',
+        ]
+        first = df.iloc[0]
+        assert first['open'] == pytest.approx(11.45, abs=1e-5)          # float32
+        assert first['close'] == pytest.approx(11.38, abs=1e-5)
+        assert first['volume'] == 9444100
+
+
 class TestLc5Parsing:
     def test_fivemin_records(self, tdx_reader):
         df = tdx_reader.read_5min_data(0, 'sz000001')