"""Reproducible syntax experiments, distinct from PDF extraction benchmarks.
HTML is stored as .txt so it can pass the existing public evidence allowlist.
"""
import argparse
import importlib.metadata
import json
import platform
import subprocess
from datetime import datetime, timezone
from html.parser import HTMLParser
from pathlib import Path
import markdown
from markdown_it import MarkdownIt

SAMPLES = {
    'diagnostic': 'alpha\nbeta\n\nalpha\\\nbeta\n\nalpha  \nbeta\n',
    'soft-break': 'alpha\nbeta\n',
    'two-spaces': 'alpha  \nbeta\n',
    'backslash-break': 'alpha\\\nbeta\n',
    'blank-line': 'alpha\n\nbeta\n',
    'ordered-list': '3. alpha\n4. beta\n\n8. restart\n',
    'bullet-list': '- alpha\n- beta\n',
    'table': '| a | b |\n| --- | --- |\n| 1 | 2 |\n',
    'code-block': '```python\nprint("alpha")\n```\n',
    'link': '[alpha](https://example.com)\n',
    'checkbox': '- [x] alpha\n- [ ] beta\n',
    'frontmatter': '---\ntitle: Alpha\n---\n\n# Alpha\n',
}


class Structure(HTMLParser):
    def __init__(self):
        super().__init__()
        self.counts = {}

    def handle_starttag(self, tag, attrs):
        self.counts[tag] = self.counts.get(tag, 0) + 1


def run(lab):
    lab = Path(lab)
    out = lab / 'renderers'
    out.mkdir(parents=True, exist_ok=True)
    version = subprocess.check_output(['pandoc', '--version'], text=True).splitlines()[0]
    versions = {'markdown-it-commonmark': importlib.metadata.version('markdown-it-py'),
                'markdown-it-breaks': importlib.metadata.version('markdown-it-py'),
                'python-markdown': importlib.metadata.version('Markdown'),
                'pandoc-gfm': version}
    results = []
    for name, source in SAMPLES.items():
        (out / f'{name}.md').write_text(source)
        rendered = {
            'markdown-it-commonmark': MarkdownIt('commonmark').render(source),
            'markdown-it-breaks': MarkdownIt('commonmark', {'breaks': True}).render(source),
            'python-markdown': markdown.markdown(source, extensions=['tables', 'fenced_code']),
            'pandoc-gfm': subprocess.run(['pandoc', '-f', 'gfm', '-t', 'html'], input=source,
                                       text=True, capture_output=True, check=True).stdout,
        }
        for renderer, html in rendered.items():
            path = out / f'{name}--{renderer}.txt'
            path.write_text(html)
            parser = Structure()
            parser.feed(html)
            results.append({'sample': name, 'renderer': renderer, 'tags': parser.counts,
                            'source': f'lab/renderers/{name}.md',
                            'output': f'lab/renderers/{path.name}'})
    report = {'kind': 'markdown-rendering', 'date': datetime.now(timezone.utc).isoformat(),
              'python': platform.python_version(), 'versions': versions,
              'scope': 'Local implementations; not GitHub.com or VS Code UI. Tag counts describe structure, not visual fidelity.',
              'command': '.venv/bin/python runs/2026-10-18-markovo-markdown-line-break/lab/reproduce.py --lab runs/2026-10-18-markovo-markdown-line-break/lab',
              'results': results}
    (out / 'results.json').write_text(json.dumps(report, indent=2) + '\n')
    lines = ['# Markdown renderer experiment', '', report['scope'], '',
             '| Sample | Renderer | br | p | table | li |', '|---|---|---:|---:|---:|---:|']
    for row in results:
        c = row['tags']
        lines.append(f"| {row['sample']} | {row['renderer']} | {c.get('br', 0)} | {c.get('p', 0)} | {c.get('table', 0)} | {c.get('li', 0)} |")
    (out / 'results.md').write_text('\n'.join(lines) + '\n')
    return report


if __name__ == '__main__':
    ap = argparse.ArgumentParser()
    ap.add_argument('--lab', type=Path, required=True)
    report = run(ap.parse_args().lab)
    print(json.dumps({'kind': report['kind'], 'cases': len(report['results']), 'versions': report['versions']}))