#!/usr/bin/env python3
"""Improved Doxygen→Markdown extractor. Handles multi-line signatures."""
import re, os, glob

def read_until_braces(text, start):
    """Read from start position until we find a complete function/struct/enum declaration.
    Returns (full_signature, end_position)"""
    pos = start
    result = []
    paren_depth = 0
    angle_depth = 0
    found_open = False
    found_template = False
    
    while pos < len(text):
        ch = text[pos]
        result.append(ch)
        
        if ch == '(':
            paren_depth += 1
            found_open = True
        elif ch == ')':
            paren_depth -= 1
        elif ch == '<' and found_open:
            angle_depth += 1
        elif ch == '>' and found_open:
            angle_depth -= 1
        elif ch == '{' and found_open and paren_depth == 0 and angle_depth == 0:
            # Function body started
            break
        elif ch == ';' and found_open and paren_depth == 0 and angle_depth == 0:
            # Declaration ended
            break
        elif ch == '\n' and not found_open:
            # Too many newlines without finding a function
            if len(result) > 200:
                break
        
        pos += 1
    
    return ''.join(result).strip(), pos + 1

def extract_blocks(text):
    """Extract /** */ blocks and the function/struct/enum that follows."""
    results = []
    for m in re.finditer(r'/\*\*\s*\n(.*?)\*/', text, re.DOTALL):
        doc_raw = m.group(1)
        after_start = m.end()
        
        # Skip whitespace, includes, comments
        while after_start < len(text):
            ch = text[after_start]
            if ch in ' \t\n\r':
                after_start += 1
                continue
            line_start = after_start
            line_end = text.find('\n', line_start)
            if line_end < 0: line_end = len(text)
            line = text[line_start:line_end].strip()
            if line.startswith('#') or line.startswith('//') or line.startswith('using ') or line == '':
                after_start = line_end + 1
                continue
            break
        
        if after_start >= len(text):
            continue
            
        # Read full signature (multi-line)
        sig, end_pos = read_until_braces(text, after_start)
        
        # Clean doc
        doc_lines = []
        for line in doc_raw.split('\n'):
            line = line.strip()
            if line.startswith('*'):
                line = line[1:].strip()
            if line.startswith('@') or line:
                doc_lines.append(line)
        clean_doc = '\n'.join(doc_lines).strip()
        
        results.append((clean_doc, sig))
    
    return results

def to_md(doc, sig):
    """Convert doxygen doc + signature to markdown."""
    brief, params, ret, notes, sees, warns = '', [], '', [], [], []
    in_code = False
    
    for line in doc.split('\n'):
        line = line.strip()
        if line.startswith('@brief '):
            brief = line[7:]
        elif line.startswith('@param '):
            params.append(line[7:])
        elif line.startswith('@return ') or line.startswith('@retval '):
            ret = line.split(' ', 1)[1] if ' ' in line else ''
        elif line.startswith('@note '):
            notes.append(f'> {line[6:]}')
        elif line.startswith('@warning '):
            warns.append(f'> 🔴 {line[9:]}')
        elif line.startswith('@see '):
            sees.append(line[5:].strip())
        elif line.startswith('@code'):
            in_code = True
        elif line.startswith('@endcode'):
            in_code = False
        elif line.startswith('@ingroup') or line.startswith('@defgroup'):
            pass
        elif not in_code and line and not line.startswith('@') and not line.startswith('\\'):
            if not brief:
                brief = line
    
    if not brief:
        return ''
    
    md = f"#### {brief}\n\n"
    
    # Clean signature
    clean = sig.strip().rstrip(';')
    if clean and 'namespace' not in clean and clean != '};':
        # Truncate very long signatures
        if len(clean) > 200:
            # Find last complete parameter
            paren = clean.rfind(')')
            if paren > 50:
                clean = clean[:paren+1]
        md += f"```cpp\n{clean}\n```\n\n"
    
    if params:
        md += "| 参数 | 说明 |\n|------|------|\n"
        for p in params:
            sp = p.split(' ', 1)
            name = sp[0]
            desc = sp[1] if len(sp) > 1 else ''
            md += f"| `{name}` | {desc} |\n"
        md += "\n"
    
    if ret:
        md += f"**返回**: {ret}\n\n"
    
    for n in notes:
        md += f"{n}\n\n"
    for w in warns:
        md += f"{w}\n\n"
    
    if sees:
        md += f"**参见**: {', '.join(f'`{s}`' for s in sees)}\n\n"
    
    return md

# Main
all_headers = sorted(glob.glob('/ws/ViewDesignEngine/include/vde/**/*.h', recursive=True))
modules = {}
for h in all_headers:
    mod = h.split('/include/vde/')[1].split('/')[0]
    modules.setdefault(mod, []).append(h)

out = ['# ViewDesignEngine API 参考手册\n\n> 版本 3.2.0 | 76 个头文件 | 自动生成\n\n']

names = {
    'foundation':'基础模块','core':'核心几何','curves':'曲线曲面','mesh':'网格处理',
    'spatial':'空间索引','boolean':'布尔运算','collision':'碰撞检测','brep':'B-Rep建模',
    'sdf':'SDF隐式建模','sketch':'草图约束','capi':'C API'
}

for mod in ['foundation','core','curves','mesh','spatial','boolean','collision','brep','sdf','sketch','capi']:
    if mod not in modules: continue
    out.append(f'## {names[mod]} (`vde::{mod}`)\n\n')
    
    for h in sorted(modules[mod]):
        with open(h) as f: content = f.read()
        blocks = extract_blocks(content)
        if not blocks: continue
        
        fname = os.path.basename(h)
        out.append(f'### {fname}\n\n')
        
        for doc, sig in blocks:
            if len(doc) < 15 and not sig.strip(): continue
            md = to_md(doc, sig)
            if md.strip():
                out.append(md)
        out.append('\n')

result = '\n'.join(out)

# Remove empty ### headers
result = re.sub(r'### \n\n+', '', result)

with open('/ws/ViewDesignEngine/docs/API-REFERENCE.md', 'w') as f:
    f.write(result)

print(f'Written: {len(result)} chars, {result.count(chr(10))} lines')