mitai-jinkendo/backend/evaluation_helper.py

"""
Training Type Profiles - Helper Functions
Utilities for loading parameters, profiles, and running evaluations.

Issue: #15
Date: 2026-03-23
"""
from typing import Dict, Optional, List
from decimal import Decimal
import logging

from db import get_cursor
from profile_evaluator import TrainingProfileEvaluator

logger = logging.getLogger(__name__)


def convert_decimals(obj):
    """
    Recursively converts Decimal objects to float for JSON serialization.

    PostgreSQL returns numeric values as Decimal, but psycopg2.Json() can't serialize them.
    """
    if isinstance(obj, Decimal):
        return float(obj)
    elif isinstance(obj, dict):
        return {k: convert_decimals(v) for k, v in obj.items()}
    elif isinstance(obj, list):
        return [convert_decimals(item) for item in obj]
    return obj


def load_parameters_registry(cur) -> Dict[str, Dict]:
    """
    Loads training parameters registry from database.

    Returns:
        Dict mapping parameter_key -> config
    """
    cur.execute("""
        SELECT key, name_de, name_en, category, data_type, unit,
               description_de, source_field, validation_rules
        FROM training_parameters
        WHERE is_active = true
    """)

    registry = {}
    for row in cur.fetchall():
        registry[row['key']] = dict(row)

    return registry


def load_training_type_profile(cur, training_type_id: int) -> Optional[Dict]:
    """
    Loads training type profile for a given type ID.

    Returns:
        Profile JSONB or None if not configured
    """
    cur.execute(
        "SELECT profile FROM training_types WHERE id = %s",
        (training_type_id,)
    )
    row = cur.fetchone()

    if row and row['profile']:
        return row['profile']

    return None


def load_evaluation_context(
    cur,
    profile_id: str,
    activity_date: str,
    lookback_days: int = 30
) -> Dict:
    """
    Loads context data for evaluation (user profile + recent activities).

    Args:
        cur: Database cursor
        profile_id: User profile ID
        activity_date: Date of activity being evaluated
        lookback_days: How many days of history to load

    Returns:
        {
            "user_profile": {...},
            "recent_activities": [...],
            "historical_activities": [...]
        }
    """
    # Load user profile
    cur.execute(
        "SELECT hf_max, sleep_goal_minutes FROM profiles WHERE id = %s",
        (profile_id,)
    )
    user_row = cur.fetchone()
    user_profile = dict(user_row) if user_row else {}

    # Load recent activities (last N days)
    cur.execute("""
        SELECT id, date, training_type_id, duration_min, hr_avg, hr_max,
               distance_km, kcal_active, rpe
        FROM activity_log
        WHERE profile_id = %s
          AND date >= %s::date - INTERVAL '%s days'
          AND date < %s::date
        ORDER BY date DESC
        LIMIT 50
    """, (profile_id, activity_date, lookback_days, activity_date))

    recent_activities = [dict(r) for r in cur.fetchall()]

    # Historical activities (same for MVP)
    historical_activities = recent_activities

    return {
        "user_profile": user_profile,
        "recent_activities": recent_activities,
        "historical_activities": historical_activities
    }


def evaluate_and_save_activity(
    cur,
    activity_id: str,
    activity_data: Dict,
    training_type_id: int,
    profile_id: str
) -> Optional[Dict]:
    """
    Evaluates an activity and saves the result to the database.

    Args:
        cur: Database cursor
        activity_id: Activity ID
        activity_data: Activity data dict
        training_type_id: Training type ID
        profile_id: User profile ID

    Returns:
        Evaluation result or None if no profile configured
    """
    # Load profile
    profile = load_training_type_profile(cur, training_type_id)
    if not profile:
        logger.info(f"[EVALUATION] No profile for training_type {training_type_id}, skipping")
        return None

    # Load parameters registry
    parameters = load_parameters_registry(cur)

    # Load context
    context = load_evaluation_context(
        cur,
        profile_id,
        activity_data.get("date"),
        lookback_days=30
    )

    # Convert Decimal values in activity_data and context
    activity_data_clean = convert_decimals(activity_data)
    context_clean = convert_decimals(context)

    # Evaluate
    evaluator = TrainingProfileEvaluator(parameters)
    evaluation_result = evaluator.evaluate_activity(
        activity_data_clean,
        profile,
        context_clean
    )

    # Save to database
    from psycopg2.extras import Json

    # Convert Decimal to float for JSON serialization
    evaluation_result_clean = convert_decimals(evaluation_result)

    cur.execute("""
        UPDATE activity_log
        SET evaluation = %s,
            quality_label = %s,
            overall_score = %s
        WHERE id = %s
    """, (
        Json(evaluation_result_clean),
        evaluation_result_clean.get("quality_label"),
        evaluation_result_clean.get("overall_score"),
        activity_id
    ))

    logger.info(
        f"[EVALUATION] Activity {activity_id}: "
        f"{evaluation_result.get('quality_label')} "
        f"(score: {evaluation_result.get('overall_score')})"
    )

    return evaluation_result


def batch_evaluate_activities(
    cur,
    profile_id: str,
    limit: Optional[int] = None
) -> Dict:
    """
    Re-evaluates all activities for a user.

    Useful for:
    - Initial setup after profiles are configured
    - Re-evaluation after profile changes

    Args:
        cur: Database cursor
        profile_id: User profile ID
        limit: Optional limit for testing

    Returns:
        {
            "total": int,
            "evaluated": int,
            "skipped": int,
            "errors": int
        }
    """
    # Load all activities
    query = """
        SELECT id, profile_id, date, training_type_id, duration_min,
               hr_avg, hr_max, distance_km, kcal_active, kcal_resting,
               rpe, pace_min_per_km, cadence, elevation_gain
        FROM activity_log
        WHERE profile_id = %s
        ORDER BY date DESC
    """
    params = [profile_id]

    if limit:
        query += " LIMIT %s"
        params.append(limit)

    cur.execute(query, params)
    activities = cur.fetchall()

    stats = {
        "total": len(activities),
        "evaluated": 0,
        "skipped": 0,
        "errors": 0
    }

    # Track error details
    error_details = []

    for activity in activities:
        activity_dict = dict(activity)
        try:
            result = evaluate_and_save_activity(
                cur,
                activity_dict["id"],
                activity_dict,
                activity_dict["training_type_id"],
                profile_id
            )

            if result:
                stats["evaluated"] += 1
            else:
                stats["skipped"] += 1

        except Exception as e:
            logger.error(f"[BATCH-EVAL] Error evaluating {activity_dict['id']}: {e}")
            error_details.append({
                "activity_id": activity_dict['id'],
                "training_type_id": activity_dict.get('training_type_id'),
                "error": str(e)
            })
            stats["errors"] += 1

    # Add error details to stats (limit to first 10)
    if error_details:
        stats["error_details"] = error_details[:10]

    logger.info(f"[BATCH-EVAL] Completed: {stats}")
    return stats