data.py

# -*- coding: utf-8 -*-
"""Data loading with a local cache, so results are reproducible and runs work offline."""
from __future__ import annotations

import os

import pandas as pd


def load_prices(symbols: list[str], cache_path: str = "data/prices.csv",
                refresh: bool = False) -> pd.DataFrame:
    """Adjusted daily closes for ``symbols``, cached to ``cache_path``.

    First run downloads from Yahoo Finance (via yfinance) and writes the cache; later runs
    read the cache, so results reproduce offline after the first download — and the numbers
    can't silently drift between runs. (The cache itself is gitignored, not shipped.)
    """
    if os.path.exists(cache_path) and not refresh:
        px = pd.read_csv(cache_path, index_col=0, parse_dates=True)
        missing = [s for s in symbols if s not in px.columns]
        if not missing:
            return px[symbols].dropna()
    import yfinance as yf  # imported lazily: tests never need it

    raw = yf.download(symbols, period="max", auto_adjust=True, progress=False)["Close"]
    if isinstance(raw, pd.Series):        # yfinance returns a 1-D Series for a single ticker
        raw = raw.to_frame(symbols[0])
    raw = raw[[c for c in symbols if c in raw.columns]].dropna()
    if raw.empty:
        raise RuntimeError("download returned no data")
    os.makedirs(os.path.dirname(cache_path) or ".", exist_ok=True)
    raw.to_csv(cache_path)
    return raw