"""Intraday volume profile features from 5-minute bars.

Computes institutional conviction signals from reaction-day intraday data:
  - first_half_volume_pct: fraction of volume in 9:30-12:00
  - volume_front_loading_ratio: first_half / second_half volume
  - vwap_premium_pct: (close - vwap) / vwap
  - institutional_conviction_score: composite [0, 1]
"""
from __future__ import annotations

from typing import Any

from libs.common.logging import get_logger

logger = get_logger(__name__)

# NYSE session: 9:30 ET - 16:00 ET; midpoint at 12:00 ET
_MIDPOINT_HOUR = 12
_MIDPOINT_MINUTE = 0


def compute_intraday_features(bars: list[dict[str, Any]]) -> dict[str, Any] | None:
    """Compute intraday volume profile features from 5-minute bars.

    Args:
        bars: list of dicts with keys: timestamp, open, high, low, close, volume.
              timestamp format: "2024-01-15T09:30:00-05:00" or similar.

    Returns:
        Feature dict or None if insufficient data.
    """
    if not bars or len(bars) < 10:
        return None

    first_half_vol = 0
    second_half_vol = 0
    total_volume = 0
    vwap_numerator = 0.0

    for bar in bars:
        vol = int(bar.get("volume", 0))
        if vol <= 0:
            continue

        ts = bar.get("timestamp", "")
        hour, minute = _extract_hour_minute(ts)
        if hour is None:
            continue

        total_volume += vol
        typical_price = (float(bar["high"]) + float(bar["low"]) + float(bar["close"])) / 3.0
        vwap_numerator += typical_price * vol

        if hour < _MIDPOINT_HOUR or (hour == _MIDPOINT_HOUR and minute == 0):
            first_half_vol += vol
        else:
            second_half_vol += vol

    if total_volume < 100:
        return None

    first_half_pct = first_half_vol / total_volume if total_volume > 0 else 0.0
    front_loading = first_half_vol / max(1, second_half_vol)
    vwap = vwap_numerator / total_volume if total_volume > 0 else 0.0

    last_close = float(bars[-1].get("close", 0.0))
    vwap_premium = (last_close - vwap) / vwap if vwap > 0 else 0.0

    conviction = _compute_conviction_score(first_half_pct, front_loading, vwap_premium)

    return {
        "first_half_volume_pct": round(first_half_pct, 4),
        "volume_front_loading_ratio": round(front_loading, 4),
        "vwap_premium_pct": round(vwap_premium, 6),
        "institutional_conviction_score": round(conviction, 4),
    }


def _compute_conviction_score(
    first_half_pct: float,
    front_loading: float,
    vwap_premium: float,
) -> float:
    """Composite conviction score [0, 1].

    High conviction = volume front-loaded (institutions acting early)
    + closing above VWAP (sustained buying pressure).

    Components (equal weight):
      - Front-loading: 1.4-2.0x first/second half ratio -> 0.5-1.0
      - VWAP premium: 0%-2%+ close above VWAP -> 0.5-1.0
      - Volume concentration: 55-70% in first half -> 0.5-1.0
    """
    # Front-loading ratio score
    if front_loading < 1.0:
        fl_score = 0.2
    elif front_loading < 1.4:
        fl_score = 0.2 + (front_loading - 1.0) / 0.4 * 0.3
    elif front_loading <= 2.0:
        fl_score = 0.5 + (front_loading - 1.4) / 0.6 * 0.5
    else:
        fl_score = 1.0

    # VWAP premium score
    if vwap_premium < 0:
        vwap_score = max(0.0, 0.3 + vwap_premium * 10)
    elif vwap_premium < 0.01:
        vwap_score = 0.3 + vwap_premium / 0.01 * 0.4
    elif vwap_premium <= 0.02:
        vwap_score = 0.7 + (vwap_premium - 0.01) / 0.01 * 0.3
    else:
        vwap_score = 1.0

    # Volume concentration score
    if first_half_pct < 0.45:
        conc_score = 0.1
    elif first_half_pct < 0.55:
        conc_score = 0.1 + (first_half_pct - 0.45) / 0.10 * 0.4
    elif first_half_pct <= 0.70:
        conc_score = 0.5 + (first_half_pct - 0.55) / 0.15 * 0.5
    else:
        conc_score = 1.0

    return max(0.0, min(1.0, (fl_score + vwap_score + conc_score) / 3.0))


def _extract_hour_minute(timestamp_str: str) -> tuple[int | None, int | None]:
    """Extract hour and minute from an ISO timestamp string."""
    try:
        # Handle formats like "2024-01-15T09:30:00-05:00" or "2024-01-15 09:30:00"
        time_part = timestamp_str.split("T")[-1] if "T" in timestamp_str else timestamp_str.split(" ")[-1]
        parts = time_part.split(":")
        return int(parts[0]), int(parts[1])
    except (IndexError, ValueError):
        return None, None