#!/usr/bin/env python3
"""重新生成界面字体（字体二进制不入库）。

用法：
    uv run --with fonttools --with brotli tools/build-fonts.py

流程：从 lxgw-wenkai-webfont 的 CDN 下载 Regular / Bold 的全部分包 woff2 子集，
合并成单个字体后裁剪到 GB2312 + Latin + 中文标点，写入 frontend/public/fonts/。
生成的字体仅供本地构建与打包使用，不提交到仓库。
"""

from __future__ import annotations

import io
import re
import sys
import tempfile
import urllib.request
from pathlib import Path

try:
    from fontTools.merge import Merger
    from fontTools.subset import Options, Subsetter
    from fontTools.ttLib import TTFont
except ImportError:  # pragma: no cover - 由 uv --with 保证
    sys.exit("缺少依赖：uv run --with fonttools --with brotli tools/build-fonts.py")

CDN = "https://cdn.staticfile.net/lxgw-wenkai-webfont/1.7.0/"
WEIGHTS = {
    "Regular": "lxgwwenkai-regular.css",
    "Bold": "lxgwwenkai-bold.css",
}
OUT_DIR = Path(__file__).resolve().parent.parent / "frontend" / "public" / "fonts"


def fetch(url: str) -> bytes:
    request = urllib.request.Request(
        url, headers={"User-Agent": "taskpilot-font-builder"}
    )
    with urllib.request.urlopen(request, timeout=60) as response:
        return response.read()


def subset_urls(css_name: str) -> list[str]:
    css = fetch(CDN + css_name).decode("utf-8")
    # 形如：src: url('./files/lxgwwenkai-regular-subset-4.woff2')
    matches = re.findall(r"url\(\s*['\"]?([^'\")]+\.woff2)['\"]?\s*\)", css)
    urls = sorted({match.lstrip("./") for match in matches})
    if not urls:
        raise RuntimeError(f"未能从 {css_name} 解析出字体子集，CDN 格式可能已变更")
    return urls


def build_codepoints() -> set[int]:
    points: set[int] = set()
    for start, end in (
        (0x20, 0x7F),  # ASCII
        (0xA0, 0x100),  # Latin-1
        (0x2000, 0x2070),  # 通用标点
        (0x25A0, 0x2600),  # 几何图形
        (0x3000, 0x3040),  # 中文标点
        (0xFF00, 0xFFF0),  # 全角字符
    ):
        points.update(range(start, end))
    points.update(
        [0x00D7, 0x00F7, 0x2190, 0x2191, 0x2192, 0x2193, 0x2713, 0x2717, 0x23F3]
    )
    # GB2312 汉字与符号行
    for high in range(0xA1, 0xF8):
        for low in range(0xA1, 0xFF):
            try:
                points.add(ord(bytes([high, low]).decode("gb2312")))
            except UnicodeDecodeError:
                continue
    return points


def build(weight: str, css_name: str, points: set[int], workdir: Path) -> None:
    urls = subset_urls(css_name)
    print(f"[{weight}] 下载 {len(urls)} 个子集…")
    paths: list[str] = []
    for url in urls:
        data = fetch(CDN + url)
        font = TTFont(io.BytesIO(data))
        target = workdir / Path(url).name.replace(".woff2", ".ttf")
        font.save(target)
        font.close()
        paths.append(str(target))

    print(f"[{weight}] 合并 {len(paths)} 个字体…")
    merged = Merger().merge(paths)

    options = Options()
    options.flavor = "woff2"
    options.hinting = False
    options.layout_features = ["*"]
    options.name_IDs = ["*"]
    subsetter = Subsetter(options=options)
    subsetter.populate(unicodes=points)
    subsetter.subset(merged)

    output = OUT_DIR / f"LXGWWenKai-{weight}.woff2"
    merged.save(output)
    print(f"[{weight}] 写出 {output}  {output.stat().st_size / 1024 / 1024:.2f} MB")


def main() -> int:
    OUT_DIR.mkdir(parents=True, exist_ok=True)
    points = build_codepoints()
    print(f"目标字符集：{len(points)} 个码位")
    with tempfile.TemporaryDirectory() as directory:
        workdir = Path(directory)
        for weight, css_name in WEIGHTS.items():
            build(weight, css_name, points, workdir)
    return 0


if __name__ == "__main__":
    sys.exit(main())