"""
export.py — COCO JSON exporter.

Reads all per-image labeled JSON files from the labeled/ directory and
assembles a valid COCO-format JSON file.  No pycocotools dependency — the
format is built from scratch.

COCO format reference:
  https://cocodataset.org/#format-data

Output structure:
{
  "info": {...},
  "licenses": [],
  "categories": [{"id": 1, "name": "cup", "supercategory": "object"}, ...],
  "images": [{"id": 1, "file_name": "img.jpg", "width": W, "height": H}, ...],
  "annotations": [
    {
      "id": 1,
      "image_id": 1,
      "category_id": 2,
      "bbox": [x, y, w, h],          # COCO uses [x_min, y_min, width, height]
      "area": w * h,
      "iscrowd": 0
    },
    ...
  ]
}
"""

from __future__ import annotations

import logging
from datetime import datetime, timezone
from pathlib import Path
from typing import Optional

from autolabel.config import settings as default_settings, Settings
from autolabel.utils import load_json, save_json

logger = logging.getLogger(__name__)


def _xyxy_to_xywh(box: list[float]) -> list[float]:
    """Convert [x1, y1, x2, y2] → [x, y, width, height] (COCO format)."""
    x1, y1, x2, y2 = box
    return [x1, y1, x2 - x1, y2 - y1]


def build_coco(labeled_dir: Path) -> dict:
    """
    Read all JSON files in *labeled_dir* and build a COCO-format dict.

    Returns the COCO dict ready for serialisation.
    """
    json_files = sorted(labeled_dir.glob("*.json"))
    # Exclude any existing coco_export.json to avoid self-inclusion
    json_files = [f for f in json_files if f.name != "coco_export.json"]

    if not json_files:
        logger.warning("No labeled JSON files found in %s", labeled_dir)
        return {}

    logger.info("Building COCO export from %d file(s)…", len(json_files))

    # Collect all category names in encounter order, deduplicating
    category_index: dict[str, int] = {}  # name → category_id
    images_list: list[dict] = []
    annotations_list: list[dict] = []

    ann_id = 1

    for img_id, json_path in enumerate(json_files, start=1):
        data = load_json(json_path)

        image_path = Path(data["image_path"])
        images_list.append(
            {
                "id": img_id,
                "file_name": image_path.name,
                "width": data["image_width"],
                "height": data["image_height"],
            }
        )

        for det in data.get("detections", []):
            label: str = det["label"]
            if label not in category_index:
                category_index[label] = len(category_index) + 1

            cat_id = category_index[label]
            xywh = _xyxy_to_xywh(det["box_xyxy"])
            area = round(xywh[2] * xywh[3], 2)

            annotations_list.append(
                {
                    "id": ann_id,
                    "image_id": img_id,
                    "category_id": cat_id,
                    "bbox": [round(v, 1) for v in xywh],
                    "area": area,
                    "iscrowd": 0,
                    "segmentation": det.get("segmentation", []),
                }
            )
            ann_id += 1

    categories = [
        {"id": cat_id, "name": name, "supercategory": "object"}
        for name, cat_id in sorted(category_index.items(), key=lambda x: x[1])
    ]

    coco = {
        "info": {
            "description": "autolabel — OWLv2 household object dataset",
            "version": "1.0",
            "year": datetime.now(tz=timezone.utc).year,
            "date_created": datetime.now(tz=timezone.utc).isoformat(),
        },
        "licenses": [],
        "categories": categories,
        "images": images_list,
        "annotations": annotations_list,
    }

    logger.info(
        "COCO export: %d image(s), %d annotation(s), %d categor(ies)",
        len(images_list),
        len(annotations_list),
        len(categories),
    )
    return coco


def run_export(
    labeled_dir: Path,
    output_path: Path,
    cfg: Optional[Settings] = None,
) -> None:
    """
    Build COCO JSON from *labeled_dir* and write to *output_path*.

    Args:
        labeled_dir: Directory containing per-image labeled JSON files.
        output_path: Destination path for the COCO JSON file.
        cfg:         Settings instance (module default if None).
    """
    _ = cfg or default_settings  # reserved for future use

    coco = build_coco(labeled_dir)
    if not coco:
        logger.error("Nothing to export.")
        return

    save_json(coco, output_path)
    logger.info("COCO JSON written → %s", output_path)