from __future__ import annotations

from PIL import Image

from ocr_studio.config import MAX_NEW_TOKENS, QUALITY_SHORT_CHARS, QUALITY_SPARSE_RATIO
from ocr_studio.spotting import TextSpan


def assess_quality(
    text: str,
    pages: list[Image.Image],
    spans_by_page: list[list[TextSpan]],
    truncated_pages: bool,
    deskew_angles: list[float],
) -> list[tuple[str, dict[str, object]]]:
    warnings: list[tuple[str, dict[str, object]]] = []
    chars = len((text or "").strip())
    if chars < QUALITY_SHORT_CHARS:
        warnings.append(("warn.short", {}))

    area = sum(max(1, image.width * image.height) for image in pages) or 1
    if chars / float(area) < QUALITY_SPARSE_RATIO and chars >= QUALITY_SHORT_CHARS:
        warnings.append(("warn.sparse", {}))

    if any(not page_spans for page_spans in spans_by_page) and any(pages):
        warnings.append(("warn.no_boxes", {}))

    if truncated_pages:
        warnings.append(("warn.truncated_pages", {"count": len(pages)}))

    notable = [angle for angle in deskew_angles if abs(angle) >= 0.8]
    if notable:
        warnings.append(("warn.deskew", {"angle": notable[0]}))

    if chars >= int(MAX_NEW_TOKENS * 2.2):
        warnings.append(("warn.tokens", {}))

    return warnings