#!/usr/bin/env python3
"""
strip_page.py - HTML 脱脂脚本 v2
修复：全面处理所有格式的 base64 图片（无引号/双引号/单引号/url()/style属性）
"""

import re
import os
import sys
from pathlib import Path

STRIP_THRESHOLD_STYLE_CHARS = 50000
BASE64_PLACEHOLDER = 'data:image/png;base64,iVBORw0KGgo='


def extract_css_variables(style_text):
    vars_found = []
    for m in re.finditer(r'(--[\w-]+)\s*:\s*([^;{]+?)(?:\s*!important)?;', style_text):
        vars_found.append(f'  {m.group(1)}: {m.group(2).strip()};')
    return vars_found


def strip_html(input_path, output_dir=None):
    input_path = Path(input_path)
    if output_dir is None:
        output_dir = input_path.parent
    output_dir = Path(output_dir)
    output_dir.mkdir(parents=True, exist_ok=True)

    stem = input_path.stem
    html_out = output_dir / f'{stem}.stripped.html'
    css_out = output_dir / f'{stem}.styles.css'
    vars_out = output_dir / f'{stem}.variables.css'

    with open(input_path, 'r', encoding='utf-8', errors='replace') as f:
        html = f.read()

    original_size = len(html)
    stats = {
        'base64_replaced': 0,
        'base64_saved_chars': 0,
        'styles_extracted': 0,
        'styles_saved_chars': 0,
        'css_vars_extracted': 0,
        'scripts_removed': 0,
        'data_hrefs_fixed': 0,
    }

    # ── 0. 扫描所有 CSS 变量（先于任何替换）────────────
    all_vars = []
    for m in re.finditer(r'<style[^>]*>(.*?)</style>', html, flags=re.DOTALL):
        all_vars.extend(extract_css_variables(m.group(1)))
    # 也从 style 属性中提取
    for m in re.finditer(r'style="([^"]+)"', html):
        all_vars.extend(extract_css_variables(m.group(1)))

    # ── 1. 全面替换所有 base64 图片 ─────────────────
    # 格式1: url(data:image/...)  无引号
    def repl_url_nq(m):
        stats['base64_replaced'] += 1
        stats['base64_saved_chars'] += len(m.group(1))
        return f'url("{BASE64_PLACEHOLDER}")'
    n = len(re.findall(r'url\((data:image[^\s\)]+)\)', html))
    html = re.sub(r'url\((data:image[^\s\)]+)\)', repl_url_nq, html)

    # 格式2: url("data:image/...")  双引号
    def repl_url_dq(m):
        stats['base64_replaced'] += 1
        stats['base64_saved_chars'] += len(m.group(1))
        return f'url("{BASE64_PLACEHOLDER}")'
    html = re.sub(r'url\("(data:image[^"]+)"\)', repl_url_dq, html)

    # 格式3: url('data:image/...')  单引号
    def repl_url_sq(m):
        stats['base64_replaced'] += 1
        stats['base64_saved_chars'] += len(m.group(1))
        return f"url('{BASE64_PLACEHOLDER}')"
    html = re.sub(r"url\('(data:image[^']+)'\)", repl_url_sq, html)

    # 格式4: style="background-image:url(data:...)"  style属性中无引号
    def repl_style_nq(m):
        prefix = m.group(1)
        data = m.group(2)
        stats['base64_replaced'] += 1
        stats['base64_saved_chars'] += len(data)
        return f'{prefix}url("{BASE64_PLACEHOLDER}")'
    html = re.sub(r'(background-image\s*:\s*)url\((data:image[^\s\)]+)\)', repl_style_nq, html, flags=re.IGNORECASE)

    # 格式5: style="background-image:url("data:...")"  style属性中有引号
    def repl_style_dq(m):
        prefix = m.group(1)
        data = m.group(2)
        stats['base64_replaced'] += 1
        stats['base64_saved_chars'] += len(data)
        return f'{prefix}url("{BASE64_PLACEHOLDER}")'
    html = re.sub(r'(background-image\s*:\s*)url\("(data:image[^"]+)"\)', repl_style_dq, html, flags=re.IGNORECASE)

    # 格式6: <img src="data:image/...">
    def repl_img(m):
        attrs = m.group(1)
        stats['base64_replaced'] += 1
        stats['base64_saved_chars'] += len(m.group(2))
        return f'<img{attrs}src="{BASE64_PLACEHOLDER}"'
    html = re.sub(r'<img([^>]*?)src="(data:image/[^"]+)"', repl_img, html, flags=re.IGNORECASE)

    # ── 2. 处理极大的内联 style ──────────────────────
    extracted_styles = []

    def process_style(m):
        tag = m.group(1)
        attrs = m.group(2)
        content = m.group(3)

        if len(content) < STRIP_THRESHOLD_STYLE_CHARS:
            return m.group(0)

        stats['styles_extracted'] += 1
        stats['styles_saved_chars'] += len(content)
        all_vars.extend(extract_css_variables(content))

        style_id = f'extracted-style-{stats["styles_extracted"]}'
        html_tag = f'<link rel="stylesheet" id="{style_id}" href="{stem}.styles.css">'
        extracted_styles.append((style_id, content))
        return f'<!-- {style_id} moved to external CSS -->'

    html = re.sub(r'(<style)([^>]*?)>(.*?)</style>', process_style, html, flags=re.DOTALL)

    # ── 3. 写外置 CSS ────────────────────────────────
    if extracted_styles:
        css_parts = ['/* === Extracted inline styles === */', '']
        for style_id, content in extracted_styles:
            css_parts.append(f'/* --- {style_id} --- */')
            css_parts.append(content)
            css_parts.append('')
        css_out.write_text('\n'.join(css_parts), encoding='utf-8')

    # ── 4. 写变量文件 ────────────────────────────────
    if all_vars:
        seen = set()
        unique_vars = []
        for v in all_vars:
            key = v.strip().split(':')[0]
            if key not in seen and key.startswith('--'):
                seen.add(key)
                unique_vars.append(v)
        vars_out.write_text(
            '/* === Extracted CSS Variables === */\n:root {\n' +
            '\n'.join(unique_vars) +
            '\n}\n',
            encoding='utf-8'
        )
        stats['css_vars_extracted'] = len(unique_vars)

    # ── 5. 移除 inline script ────────────────────────
    stats['scripts_removed'] = len(re.findall(r'<script[^>]*>.*?</script>', html, flags=re.DOTALL))
    html = re.sub(r'<script[^>]*>.*?</script>', '', html, flags=re.DOTALL)
    html = re.sub(r'<script[^>]*/>\s*', '', html)

    # ── 6. 清理 data: href ──────────────────────────
    stats['data_hrefs_fixed'] = len(re.findall(r'href="data:[^"]+"', html))
    html = re.sub(r'href="(data:[^"]+)"', 'href="#data-link-removed"', html)

    # ── 7. 写脱脂 HTML ──────────────────────────────
    html = re.sub(r'<!--\s*-->', '', html)

    if extracted_styles and '<link rel="stylesheet"' not in html:
        html = html.replace('<head>', f'<head>\n  <link rel="stylesheet" href="{stem}.styles.css">', 1)

    html_out.write_text(html, encoding='utf-8')

    new_html_size = len(html)
    css_size = css_out.stat().st_size if css_out.exists() else 0
    vars_size = vars_out.stat().st_size if vars_out.exists() else 0
    saved = original_size - new_html_size - css_size

    print(f'''
✅ 脱脂完成！
{'='*54}
📄 输入: {input_path.name}
   原始大小: {original_size/1024:.1f} KB ({original_size:,} 字符)
{'='*54}
📊 处理结果:
   base64 图片替换:  {stats["base64_replaced"]} 个
   base64 节省:      {stats["base64_saved_chars"]/1024:.1f} KB
   style 块提取:    {stats["styles_extracted"]} 个
   style 节省:       {stats["styles_saved_chars"]/1024:.1f} KB
   CSS 变量提取:    {stats["css_vars_extracted"]} 个
   script 移除:      {stats["scripts_removed"]} 个
   data href 修复:  {stats["data_hrefs_fixed"]} 个
{'='*54}
📦 输出文件:
   HTML (脱脂):     {html_out.name}    ({new_html_size/1024:.1f} KB)
   CSS (提取样式):  {css_out.name}     ({css_size/1024:.1f} KB)
   CSS (变量):      {vars_out.name}    ({vars_size/1024:.1f} KB)
{'='*54}
💾 HTML 体积压缩:  {new_html_size/1024:.1f} KB (原始 {original_size/1024:.1f} KB)
   节省比例:        {(saved)/original_size*100:.1f}%
''')


if __name__ == '__main__':
    if len(sys.argv) < 2:
        print('用法: python3 strip_page.py <input.html> [output_dir]')
        sys.exit(1)
    strip_html(sys.argv[1], sys.argv[2] if len(sys.argv) > 2 else None)