fix(data): include the requested end date in yfinance fetches

yfinance treats end as exclusive, so get_YFin_data_online dropped the requested
end_date row and load_ohlcv dropped the current day. Request one day past the
end so the range is inclusive (look-ahead is still prevented by the curr_date
filter; the header still shows the requested range). Also correct the load_ohlcv
docstring to the 5-year window it actually downloads.
This commit is contained in:
Yijia-Xiao
2026-06-13 21:30:11 +00:00
parent 65608831f8
commit dab07688fb
3 changed files with 73 additions and 6 deletions

View File

@@ -0,0 +1,61 @@
"""yfinance treats ``end`` as exclusive; we must request one extra day so the
requested end_date (and the current day) is actually included.
Regressions for #986 (current-day OHLCV excluded) and #987 (requested end_date
row omitted).
"""
import pandas as pd
import pytest
import tradingagents.dataflows.stockstats_utils as su
import tradingagents.dataflows.y_finance as yfin
from tradingagents.dataflows.config import set_config
@pytest.mark.unit
def test_get_yfin_requests_inclusive_end(monkeypatch):
captured = {}
class FakeTicker:
def __init__(self, symbol):
pass
def history(self, start, end):
captured["start"] = start
captured["end"] = end
idx = pd.to_datetime(["2025-05-08", "2025-05-09"])
return pd.DataFrame(
{"Open": [1.0, 2.0], "High": [1.0, 2.0], "Low": [1.0, 2.0],
"Close": [1.0, 2.0], "Volume": [1, 2]},
index=idx,
)
monkeypatch.setattr(yfin.yf, "Ticker", FakeTicker)
out = yfin.get_YFin_data_online("AAPL", "2025-05-01", "2025-05-09")
# end is requested one day past end_date so 2025-05-09 is included (#987).
assert captured["end"] == "2025-05-10"
# Header still reflects the requested range, not the internal +1 day.
assert "to 2025-05-09" in out
@pytest.mark.unit
def test_load_ohlcv_requests_inclusive_end(monkeypatch, tmp_path):
set_config({"data_cache_dir": str(tmp_path)})
captured = {}
def fake_download(symbol, start, end, **kwargs):
captured["end"] = end
idx = pd.to_datetime([pd.Timestamp.today().normalize()])
return pd.DataFrame(
{"Open": [100.0], "High": [100.0], "Low": [100.0],
"Close": [100.0], "Volume": [1]},
index=idx,
)
monkeypatch.setattr(su.yf, "download", fake_download)
today = pd.Timestamp.today().strftime("%Y-%m-%d")
su.load_ohlcv("AAPL", today)
expected_end = (pd.Timestamp.today() + pd.Timedelta(days=1)).strftime("%Y-%m-%d")
assert captured["end"] == expected_end # tomorrow -> today's row included (#986)

View File

@@ -65,7 +65,7 @@ def _clean_dataframe(data: pd.DataFrame) -> pd.DataFrame:
def load_ohlcv(symbol: str, curr_date: str) -> pd.DataFrame: def load_ohlcv(symbol: str, curr_date: str) -> pd.DataFrame:
"""Fetch OHLCV data with caching, filtered to prevent look-ahead bias. """Fetch OHLCV data with caching, filtered to prevent look-ahead bias.
Downloads 15 years of data up to today and caches per symbol. On Downloads 5 years of data up to today and caches per symbol. On
subsequent calls the cache is reused. Rows after curr_date are subsequent calls the cache is reused. Rows after curr_date are
filtered out so backtests never see future prices. filtered out so backtests never see future prices.
""" """
@@ -78,11 +78,14 @@ def load_ohlcv(symbol: str, curr_date: str) -> pd.DataFrame:
config = get_config() config = get_config()
curr_date_dt = pd.to_datetime(curr_date) curr_date_dt = pd.to_datetime(curr_date)
# Cache uses a fixed window (15y to today) so one file per symbol # Cache uses a fixed window (5y to today) so one file per symbol.
today_date = pd.Timestamp.today() today_date = pd.Timestamp.today()
start_date = today_date - pd.DateOffset(years=5) start_date = today_date - pd.DateOffset(years=5)
start_str = start_date.strftime("%Y-%m-%d") start_str = start_date.strftime("%Y-%m-%d")
end_str = today_date.strftime("%Y-%m-%d") # yfinance ``end`` is EXCLUSIVE; request tomorrow so today's row is included
# when curr_date is the current day (#986). Look-ahead is still prevented by
# the curr_date filter below.
end_str = (today_date + pd.Timedelta(days=1)).strftime("%Y-%m-%d")
os.makedirs(config["data_cache_dir"], exist_ok=True) os.makedirs(config["data_cache_dir"], exist_ok=True)
data_file = os.path.join( data_file = os.path.join(

View File

@@ -14,14 +14,17 @@ def get_YFin_data_online(
): ):
datetime.strptime(start_date, "%Y-%m-%d") datetime.strptime(start_date, "%Y-%m-%d")
datetime.strptime(end_date, "%Y-%m-%d") end_dt = datetime.strptime(end_date, "%Y-%m-%d")
# Resolve broker/forex symbols to Yahoo's convention (XAUUSD+ -> GC=F). # Resolve broker/forex symbols to Yahoo's convention (XAUUSD+ -> GC=F).
canonical = normalize_symbol(symbol) canonical = normalize_symbol(symbol)
ticker = yf.Ticker(canonical) ticker = yf.Ticker(canonical)
# Fetch historical data for the specified date range # yfinance treats ``end`` as EXCLUSIVE, so it would drop the requested
data = yf_retry(lambda: ticker.history(start=start_date, end=end_date)) # end_date row (and the current day when end_date is today). Request one day
# past end_date so the requested range is actually inclusive (#986/#987).
end_inclusive = (end_dt + relativedelta(days=1)).strftime("%Y-%m-%d")
data = yf_retry(lambda: ticker.history(start=start_date, end=end_inclusive))
# Empty result means the symbol is unknown/delisted. Raise a typed error # Empty result means the symbol is unknown/delisted. Raise a typed error
# instead of returning prose: the routing layer turns it into a single # instead of returning prose: the routing layer turns it into a single