grader.py

from __future__ import annotations

from typing import Iterable

try:
    from models import AppObservation
except ImportError:
    from app.models import AppObservation


def _clamp_score(value):
    return max(0.0, min(1.0, float(value)))

def _safe_ratio(numerator, denominator):
    if denominator <= 0:
        return 0.0
    return _clamp_score(numerator / denominator)


def _attempt_quality(rewards):
    rewards = list(rewards)
    if not rewards:
        return 0.0

    positive_attempts = sum(1 for reward in rewards if reward > 0)
    return _safe_ratio(positive_attempts, len(rewards))


def feedback(score, category):
    if score >= 0.9:
        return f"{category} performance is excellent and highly reliable."
    if score >= 0.7:
        return f"{category} performance is solid with minor room for improvement."
    if score >= 0.4:
        return f"{category} performance is partial and needs more consistency."
    return f"{category} performance is weak and needs significant improvement."


def grade_segmentation(app_obs: AppObservation):
    total_objects = len(app_obs.objectsFound) + len(app_obs.objectsLeft)
    progress_score = _safe_ratio(len(app_obs.objectsFound), total_objects)
    quality_score = _attempt_quality(app_obs.rewardListSegment)
    score = _clamp_score((progress_score * 0.7) + (quality_score * 0.3))
    return score, feedback(score, "Segmentation")


def grade_placement(app_obs: AppObservation):
    total_objects = len(app_obs.objectsFound) + len(app_obs.objectsLeft)
    progress_score = _safe_ratio(app_obs.numberPlaced, total_objects)
    quality_score = _attempt_quality(app_obs.rewardListPlace)
    score = _clamp_score((progress_score * 0.7) + (quality_score * 0.3))
    return score, feedback(score, "Placement")


def grade_adjustment(app_obs: AppObservation):
    rewards = list(app_obs.rewardListAdjust)
    if not rewards:
        return 0.0, "Adjustment was not attempted."

    score = _attempt_quality(rewards)
    return score, feedback(score, "Adjustment")