from __future__ import annotations

import json
import re
from typing import Any

import bleach
import markdown

TABLE_RE = re.compile(r"<table\b[\s\S]*?</table>", re.IGNORECASE)
JSON_OBJECT_RE = re.compile(
    r'\{[^{}]*?"bbox"\s*:\s*\[[^\]]+\][^{}]*?\}',
    re.IGNORECASE,
)
BBOX_TAG_RE = re.compile(
    r"""<(?:bbox|box)(?:\s+label=["']?([^"'>]+)["']?)?[^>]*>
        \s*\[?\s*([\d.]+)\s*,\s*([\d.]+)\s*,\s*([\d.]+)\s*,\s*([\d.]+)\s*\]?
        \s*</(?:bbox|box)>""",
    re.IGNORECASE | re.VERBOSE,
)

TABLE_TAGS = [
    "table",
    "caption",
    "thead",
    "tbody",
    "tfoot",
    "tr",
    "th",
    "td",
    "br",
    "strong",
    "em",
]
TABLE_ATTRIBUTES = {
    "th": ["colspan", "rowspan", "scope"],
    "td": ["colspan", "rowspan"],
}
TEXT_TAGS = [
    "p",
    "h1",
    "h2",
    "h3",
    "h4",
    "h5",
    "ul",
    "ol",
    "li",
    "strong",
    "em",
    "code",
    "pre",
    "blockquote",
    "hr",
    "br",
]
TABLE_TEXT_MARKER = "`<table>...</table>`"


def _normalize_box(candidate: dict[str, Any], index: int) -> dict | None:
    bbox = candidate.get("bbox")
    if not isinstance(bbox, list) or len(bbox) != 4:
        return None

    try:
        coords = [float(value) for value in bbox]
    except (TypeError, ValueError):
        return None

    maximum = max(coords)
    scale = 1 if maximum <= 1 else (1000 if maximum <= 1000 else maximum)
    x1, y1, x2, y2 = [min(1, max(0, value / scale)) for value in coords]
    if x2 <= x1 or y2 <= y1:
        return None

    label = str(candidate.get("label") or candidate.get("type") or "element")
    return {
        "id": f"{index}-{'-'.join(str(value) for value in coords)}",
        "label": label[:48],
        "bbox": [x1, y1, x2, y2],
    }


def extract_bounding_boxes(raw_text: str) -> list[dict]:
    candidates: list[dict] = []
    for match in JSON_OBJECT_RE.finditer(raw_text):
        try:
            candidate = json.loads(match.group(0))
        except json.JSONDecodeError:
            continue
        if isinstance(candidate, dict):
            candidates.append(candidate)

    for match in BBOX_TAG_RE.finditer(raw_text):
        candidates.append(
            {
                "label": match.group(1) or "element",
                "bbox": [float(match.group(index)) for index in range(2, 6)],
            }
        )

    boxes = []
    for index, candidate in enumerate(candidates):
        normalized = _normalize_box(candidate, index)
        if normalized:
            boxes.append(normalized)
    return boxes[:250]


def extract_tables(raw_text: str) -> str:
    tables = TABLE_RE.findall(raw_text)
    if not tables:
        return ""
    return bleach.clean(
        "\n".join(tables),
        tags=TABLE_TAGS,
        attributes=TABLE_ATTRIBUTES,
        protocols=[],
        strip=True,
    )


def render_text(raw_text: str) -> str:
    text_with_table_markers = TABLE_RE.sub(
        f"\n\n{TABLE_TEXT_MARKER}\n\n",
        raw_text,
    )
    rendered = markdown.markdown(
        text_with_table_markers,
        extensions=["extra", "sane_lists"],
        output_format="html",
    )
    return bleach.clean(
        rendered,
        tags=TEXT_TAGS,
        attributes={},
        protocols=[],
        strip=True,
    )


def shape_parser_output(raw_text: str) -> dict:
    output = raw_text if isinstance(raw_text, str) else ""
    return {
        "raw_output": output,
        "output_chars": len(output),
        "boxes": extract_bounding_boxes(output),
        "tables_html": extract_tables(output),
        "text_html": render_text(output),
    }


def _structured_box(
    payload: dict[str, Any],
    *,
    label: str,
    index: int,
) -> dict | None:
    bbox = payload.get("bounding_box_normalized")
    if not isinstance(bbox, dict):
        return None
    return _normalize_box(
        {
            "label": label,
            "bbox": [
                bbox.get("top_left_x"),
                bbox.get("top_left_y"),
                bbox.get("bottom_right_x"),
                bbox.get("bottom_right_y"),
            ],
        },
        index,
    )


def _shape_blocks_page(page: dict[str, Any]) -> tuple[str, list[dict]]:
    blocks = page.get("blocks")
    if not isinstance(blocks, list):
        raise ValueError("blocks page is missing blocks")

    content: list[str] = []
    boxes: list[dict] = []
    for block in blocks:
        if not isinstance(block, dict):
            raise ValueError("invalid parse block")

        block_type = block.get("type")
        if block_type == "text":
            text = block.get("text")
            if not isinstance(text, dict) or not isinstance(text.get("content"), str):
                raise ValueError("invalid text block")
            content.append(text["content"])
            continue

        if block_type == "table":
            table = block.get("table")
            if not isinstance(table, dict) or not isinstance(table.get("html"), str):
                raise ValueError("invalid table block")
            content.append(table["html"])
            box = _structured_box(
                table,
                label=str(table.get("title") or "table"),
                index=len(boxes),
            )
        elif block_type == "image":
            image = block.get("image")
            if not isinstance(image, dict):
                raise ValueError("invalid image block")
            description = str(image.get("description") or "Image")
            content.append(f"*Image: {description}*")
            box = _structured_box(
                image,
                label=str(image.get("category") or image.get("id") or "image"),
                index=len(boxes),
            )
        else:
            raise ValueError("unsupported parse block")

        if box:
            boxes.append(box)

    return "\n\n".join(content), boxes


def _shape_markdown_page(page: dict[str, Any]) -> tuple[str, list[dict]]:
    markdown_page = page.get("markdown")
    if not isinstance(markdown_page, dict) or not isinstance(
        markdown_page.get("content"), str
    ):
        raise ValueError("invalid markdown page")

    boxes = []
    for image in markdown_page.get("images") or []:
        if not isinstance(image, dict):
            raise ValueError("invalid markdown image")
        box = _structured_box(
            image,
            label=str(image.get("category") or image.get("id") or "image"),
            index=len(boxes),
        )
        if box:
            boxes.append(box)
    return markdown_page["content"], boxes


def shape_parse_response(body: dict[str, Any]) -> dict:
    if not isinstance(body, dict):
        raise TypeError("parse response must be an object")
    pages = body.get("pages")
    if not isinstance(pages, list) or not pages:
        raise ValueError("parse response is missing pages")

    page_content: list[str] = []
    boxes: list[dict] = []
    for page in pages:
        if not isinstance(page, dict):
            raise ValueError("invalid parse page")
        page_type = page.get("type")
        if page_type == "blocks":
            content, page_boxes = _shape_blocks_page(page)
        elif page_type == "markdown":
            content, page_boxes = _shape_markdown_page(page)
        elif page_type == "raw_generation" and isinstance(
            page.get("raw_generation"), str
        ):
            content, page_boxes = page["raw_generation"], []
        else:
            raise ValueError("unsupported parse page")
        page_content.append(content)
        boxes.extend(page_boxes)

    text_output = "\n\n".join(page_content)
    shaped = shape_parser_output(text_output)
    return {
        **shaped,
        "raw_output": json.dumps(body, indent=2, ensure_ascii=False),
        "output_chars": len(text_output),
        "boxes": boxes[:250],
        "text_output": text_output,
        "response": body,
    }