#!/usr/bin/env python3
"""
fix_i18n_w1_dspy.py — dspy 安全修复 W1（保守策略）
只处理纯字面量 "text": "中文"（match 后第一个非空白字符是 , 或 }），
跳过所有拼接（+）、多行隐式连接（后跟换行+字符串）、变量插值——那些记入 manual 清单。

替换: "text": "X"  →  "otext": "X", "text": "X", "i18n": True
引号风格跟随原文（' 或 "）。
用法: python3 scripts/fix_i18n_w1_dspy.py /tmp/i18n.json [--dry-run]
"""
import json, re, sys
from collections import defaultdict

TEXT_RE = re.compile(r'''(['"])text\1\s*:\s*(['"])((?:[^'"\\]|\\.)*)\2''')

def safe_literal(raw, m):
    """match 后第一个非空白字符必须是 , 或 }（排除拼接与多行隐式连接）"""
    j = m.end()
    while j < len(raw) and raw[j] in ' \t':
        j += 1
    return j < len(raw) and raw[j] in ',}'

def main():
    report = json.load(open(sys.argv[1], encoding='utf-8'))
    dry = '--dry-run' in sys.argv
    viols = [v for v in report['violations']
             if v['code'] == 'W1' and not v['git_ignored'] and v['file'].endswith('.dspy')]
    by_file = defaultdict(list)
    for v in viols:
        by_file[v['file']].append(v)

    fixed = manual = 0
    manual_list = []
    for fp, vs in sorted(by_file.items()):
        raw = open(fp, encoding='utf-8').read()
        # 从后往前
        changed = False
        for v in sorted(vs, key=lambda x: -x['pos']):
            pos = v['pos']
            m = TEXT_RE.match(raw, pos)
            if not m:
                # otext 无 i18n 的情形
                mo = re.compile(r'''(['"])otext\1\s*:\s*(['"])((?:[^'"\\]|\\.)*)\2''').match(raw, pos)
                if mo and not re.search(r"['\"]i18n['\"]\s*:\s*True", raw[mo.end():mo.end()+60]):
                    if safe_literal(raw, mo):
                        ins = f', {mo.group(1)}i18n{mo.group(1)}: True'
                        if not dry:
                            raw = raw[:mo.end()] + ins + raw[mo.end():]
                        fixed += 1; changed = True
                    else:
                        manual += 1; manual_list.append((fp, pos, 'otext动态', v['text'][:50]))
                else:
                    manual += 1; manual_list.append((fp, pos, '无法定位', v['text'][:50]))
                continue
            val = m.group(3)
            if not safe_literal(raw, m):
                manual += 1
                manual_list.append((fp, pos, '拼接/多行串', val[:50]))
                continue
            q1, q2 = m.group(1), m.group(2)
            seg = raw[m.start():m.end()]
            new_seg = f'{q1}otext{q1}: {q2}{val}{q2}, {q1}text{q1}: {q2}{val}{q2}, {q1}i18n{q1}: True'
            if not dry:
                raw = raw[:m.start()] + new_seg + raw[m.end():]
            fixed += 1; changed = True
        if changed and not dry:
            open(fp, 'w', encoding='utf-8').write(raw)

    print(f'{"DRY-RUN " if dry else ""}fixed={fixed} manual={manual} files={len(by_file)}')
    for mm in manual_list[:40]:
        print('  MANUAL', mm[0].split('/repos/')[-1], f'@{mm[1]}', mm[2], '|', mm[3])

if __name__ == '__main__':
    main()