fix: stabilize yfinance dividend TTM fixtures (#2214)

Author: ZhuLinsen

tests/test_yfinance_fundamental_adapter.py

@@ -8,20 +8,26 @@
 """
 from __future__ import annotations
 
-from datetime import datetime, timezone
 import unittest
-from datetime import datetime, timedelta
+from datetime import datetime, timezone
 from unittest.mock import patch, MagicMock
 
 import pandas as pd
-import pytz
 
 from data_provider.yfinance_fundamental_adapter import (
     YfinanceFundamentalAdapter,
     _convert_to_yf_symbol,
 )
 
 
+_DIVIDEND_EVENT_DATES = (
+    "2025-08-11",
+    "2025-11-10",
+    "2026-02-09",
+    "2026-05-11",
+)
+
+
 def _build_mock_ticker(
     info: dict,
     income_stmt: pd.DataFrame | None = None,
@@ -82,14 +88,6 @@ def test_populates_growth_earnings_dividend_boards_for_us_stock(self) -> None:
             "trailingAnnualDividendRate": 1.04,
             "dividendYield": 0.36,
         }
-        income_df = pd.DataFrame(
-            {
-                pd.Timestamp("2026-03-31"): {"Total Revenue": 1.11e11, "Net Income": 2.95e10},
-                pd.Timestamp("2025-12-31"): {"Total Revenue": 1.24e11, "Net Income": 3.62e10},
-                pd.Timestamp("2025-09-30"): {"Total Revenue": 9.49e10, "Net Income": 2.49e10},
-                pd.Timestamp("2025-06-30"): {"Total Revenue": 9.40e10, "Net Income": 2.34e10},
-            }
-        )
         # Need at least 5 columns to trigger statement-derived YoY.
         income_df_with_yoy = pd.DataFrame(
             {
@@ -106,18 +104,10 @@ def test_populates_growth_earnings_dividend_boards_for_us_stock(self) -> None:
                 pd.Timestamp("2025-12-31"): {"Operating Cash Flow": 3.5e10},
             }
         )
-        # Use dates relative to now so the 365-day TTM window always
-        # contains all 4 events regardless of when the test runs (#2204).
-        now_ny = datetime.now(pytz.timezone("America/New_York"))
         dividends = pd.Series(
             [0.26, 0.26, 0.26, 0.27],
             index=pd.DatetimeIndex(
-                [
-                    (now_ny - timedelta(days=330)).strftime("%Y-%m-%d"),
-                    (now_ny - timedelta(days=240)).strftime("%Y-%m-%d"),
-                    (now_ny - timedelta(days=150)).strftime("%Y-%m-%d"),
-                    (now_ny - timedelta(days=60)).strftime("%Y-%m-%d"),
-                ],
+                _DIVIDEND_EVENT_DATES,
                 tz="America/New_York",
             ),
             name="Dividends",
@@ -147,6 +137,7 @@ def test_populates_growth_earnings_dividend_boards_for_us_stock(self) -> None:
         # info.dividendYield (0.36) is intentionally ignored when TTM cash exists.
         self.assertAlmostEqual(div["ttm_dividend_yield_pct"], 0.3762, places=4)
         self.assertEqual(div["currency"], "USD")
+        self.assertEqual(div["events"][0]["ex_dividend_date"], "2026-05-11")
         self.assertEqual(
             bundle["belong_boards"],
             [
@@ -160,15 +151,8 @@ def test_dividends_parsed_from_single_column_dataframe(self) -> None:
         # Series. Without coercion, `.items()` yields (column_name, Series), every event
         # is dropped, and TTM silently falls back to the annual-rate estimate — the real
         # bug seen on live US/HK/JP/KR/TW reports (24.0 / "0 次" instead of the true sum).
-        # Use dates relative to now so the 365-day TTM window is always satisfied (#2204).
-        now_ny = datetime.now(pytz.timezone("America/New_York"))
         idx = pd.DatetimeIndex(
-            [
-                (now_ny - timedelta(days=330)).strftime("%Y-%m-%d"),
-                (now_ny - timedelta(days=240)).strftime("%Y-%m-%d"),
-                (now_ny - timedelta(days=150)).strftime("%Y-%m-%d"),
-                (now_ny - timedelta(days=60)).strftime("%Y-%m-%d"),
-            ],
+            _DIVIDEND_EVENT_DATES,
             tz="America/New_York",
         )
         dividends_df = pd.DataFrame({"Dividends": [0.26, 0.26, 0.26, 0.27]}, index=idx)
@@ -191,7 +175,7 @@ def test_dividends_parsed_from_single_column_dataframe(self) -> None:
 
     def test_ttm_dividend_window_uses_as_of_date_cutoff(self) -> None:
         idx = pd.DatetimeIndex(
-            ["2025-08-11", "2025-11-10", "2026-02-09", "2026-05-11"],
+            _DIVIDEND_EVENT_DATES,
             tz="America/New_York",
         )
         dividends = pd.Series([0.26, 0.26, 0.26, 0.27], index=idx, name="Dividends")