#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""DDL 幂等修正器：把 tenant_id 提升为每张表的物理首列（QC 第三轮 #1 配套）。

背景
----
data-model.md 的「tenant_id 强制打头」有两层含义：
  (a) 键前缀契约 —— 每个复合 UNIQUE KEY / KEY 的首列必须是 tenant_id（影响分区裁剪与查询计划，**必须成立**）；
  (b) 物理列序 —— CREATE TABLE 的第一个列定义是 tenant_id（可读性/评审口径）。
核验器 gate_64 以 (a) 为判定项、(b) 为 INFO。本脚本把 (b) 也补齐，使两种口径同时成立，
且**幂等**：已经是首列的表不做任何改动，重复执行结果一致。

用法
----
    python3 apps/pbls/scripts/patch_ddl_tenant_first.py \
        --file apps/pbls/scripts/ddl/pbls_tables.sql [--dry-run]

安全约束
--------
* 只做「列定义整行搬移」，不改列类型、不改索引、不改表选项、不改注释；
* 解析失败（找不到 tenant_id 列 / 括号不配对）时跳过该表并打印 SKIP，绝不写坏文件；
* --dry-run 只报告将变更的表与行号，不落盘。
"""

import argparse
import io
import os
import re
import sys

CREATE_RX = re.compile(r"CREATE\s+TABLE\s+(?:IF\s+NOT\s+EXISTS\s+)?[`\"]?(\w+)[`\"]?\s*\(", re.I)
TENANT_COL_RX = re.compile(r"^\s*[`\"]?tenant_id[`\"]?\s+", re.I)


def match_paren(text, open_idx):
    depth = 0
    for i in range(open_idx, len(text)):
        if text[i] == "(":
            depth += 1
        elif text[i] == ")":
            depth -= 1
            if depth == 0:
                return i
    return -1


def line_of(text, offset):
    return text[:offset].count("\n") + 1


def patch(sql):
    """返回 (new_sql, changes, skips)。changes/skips 元素为 (表名, 行号, 说明)。"""
    changes, skips = [], []
    out = sql
    offset_shift = 0
    for m in CREATE_RX.finditer(sql):
        tbl = m.group(1)
        open_idx = m.end() - 1
        close_idx = match_paren(sql, open_idx)
        if close_idx < 0:
            skips.append((tbl, line_of(sql, m.start()), "括号不配对，跳过"))
            continue
        body = sql[open_idx + 1:close_idx]
        lines = body.split("\n")
        tenant_i = None
        for i, ln in enumerate(lines):
            if TENANT_COL_RX.match(ln):
                tenant_i = i
                break
        if tenant_i is None:
            skips.append((tbl, line_of(sql, m.start()), "未找到 tenant_id 列定义，跳过"))
            continue
        first_meaningful = None
        for i, ln in enumerate(lines):
            if ln.strip():
                first_meaningful = i
                break
        if first_meaningful is None or tenant_i == first_meaningful:
            continue  # 已是首列，幂等无操作
        moved = lines.pop(tenant_i)
        lines.insert(first_meaningful, moved)
        new_body = "\n".join(lines)
        start = open_idx + 1 + offset_shift
        end = close_idx + offset_shift
        out = out[:start] + new_body + out[end:]
        offset_shift += len(new_body) - len(body)
        changes.append((tbl, line_of(sql, m.start()),
                        "tenant_id 由第 %d 个列定义提升为物理首列" % (tenant_i - first_meaningful + 1)))
    return out, changes, skips


def main():
    ap = argparse.ArgumentParser(description="DDL tenant_id 物理首列幂等修正器")
    ap.add_argument("--file", required=True, help="DDL 文件路径")
    ap.add_argument("--dry-run", action="store_true", help="只报告不落盘")
    args = ap.parse_args()

    if not os.path.isfile(args.file):
        sys.stderr.write("[patch_ddl] 文件不存在：%s\n" % args.file)
        return 2
    with io.open(args.file, "r", encoding="utf-8") as fh:
        sql = fh.read()

    new_sql, changes, skips = patch(sql)
    for tbl, ln, msg in changes:
        print("[PATCH] %-34s CREATE @L%-5d %s" % (tbl, ln, msg))
    for tbl, ln, msg in skips:
        print("[SKIP ] %-34s CREATE @L%-5d %s" % (tbl, ln, msg))
    print("[SUMMARY] 变更 %d 张表，跳过 %d 张表，幂等=%s"
          % (len(changes), len(skips), "是（重复执行不再变更）" if not changes else "本次有变更"))

    if changes and not args.dry_run:
        with io.open(args.file, "w", encoding="utf-8") as fh:
            fh.write(new_sql)
        print("[WRITE] 已落盘：%s（%d 字节）" % (args.file, os.path.getsize(args.file)))
    elif args.dry_run:
        print("[DRY-RUN] 未落盘")
    return 0


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