opsxsql/backend/app/services/risk_evaluator.py

"""工单风险等级自动评估。

根据 SQL 类型 + 操作对象 + 审核结果，自动计算 risk_level:
  - low:      SELECT / 简单 DML (有 WHERE + LIMIT)
  - medium:   普通 DML / 低风险 DDL (CREATE TABLE)
  - high:     大表 ALTER / 无 WHERE 的 UPDATE/DELETE / 多表 JOIN
  - critical: DROP TABLE / TRUNCATE / FLUSHDB / FLUSHALL / dropDatabase

评估时机：工单创建时自动计算，写入 ticket.risk_level。
"""

import re
from typing import Optional


def evaluate_risk_level(
    sql_content: str,
    db_type: str = "mysql",
    sql_type: str = "",
    audit_findings: Optional[list] = None,
    statement_count: int = 1,
    table_sizes: Optional[dict] = None,
) -> str:
    """评估 SQL 的风险等级。

    Args:
        sql_content: SQL 文本（可能包含多条语句）
        db_type: 数据库类型 mysql/postgres/redis/mongodb
        sql_type: 前端传入的 SQL 类型标注 DDL/DML/DQL 等
        audit_findings: 审核引擎返回的 findings 列表
        statement_count: SQL 语句条数
        table_sizes: EXPLAIN 预检返回的表大小 {table_name: row_count}

    Returns:
        risk_level: low / medium / high / critical
    """
    sql_upper = sql_content.strip().upper()

    # ── Redis 特殊处理 ──
    if db_type == "redis":
        return _evaluate_redis_risk(sql_upper)

    # ── MongoDB 特殊处理 ──
    if db_type == "mongodb":
        return _evaluate_mongo_risk(sql_upper)

    # ── MySQL / PostgreSQL ──

    # 1. critical 级别 — 不可逆危险操作
    critical_patterns = [
        r"\bDROP\s+TABLE\b",
        r"\bDROP\s+DATABASE\b",
        r"\bDROP\s+SCHEMA\b",
        r"\bTRUNCATE\s",
    ]
    for pattern in critical_patterns:
        if re.search(pattern, sql_upper):
            return "critical"

    # 2. high 级别 — 高风险操作
    high_indicators = 0

    # ALTER 大操作（但不包括 ADD INDEX 这类较小操作）
    if re.search(r"\bALTER\s+TABLE\b", sql_upper):
        alter_high = [
            r"\bDROP\s+COLUMN\b",
            r"\bMODIFY\s+COLUMN\b",
            r"\bCHANGE\s+COLUMN\b",
            r"\bDROP\s+INDEX\b",
            r"\bDROP\s+PRIMARY\s+KEY\b",
            r"\bCONVERT\s+TO\s+CHARACTER\s+SET\b",
            r"\bRENAME\s+(TO|AS)\b",
        ]
        for p in alter_high:
            if re.search(p, sql_upper):
                high_indicators += 1
                break

    # 无 WHERE 的 UPDATE / DELETE
    if re.search(r"\bUPDATE\s", sql_upper) and not re.search(r"\bWHERE\b", sql_upper):
        high_indicators += 1
    if re.search(r"\bDELETE\s+FROM\b", sql_upper) and not re.search(r"\bWHERE\b", sql_upper):
        high_indicators += 1

    # 大量语句（>5条）
    if statement_count > 5:
        high_indicators += 1

    # 审核结果有 ERROR 级别
    if audit_findings:
        errors = [f for f in audit_findings if f.get("level") == "error"]
        if len(errors) >= 2:
            high_indicators += 1

    if high_indicators > 0:
        return "high"

    # 2.5 大表检测 — EXPLAIN 预检返回的 table_sizes
    # ALTER/UPDATE/DELETE 涉及大表(>10万行)时升级为 high
    if table_sizes:
        max_rows = max(table_sizes.values()) if table_sizes else 0
        if max_rows > 100_000 and re.search(r"\b(ALTER\s+TABLE|UPDATE|DELETE\s+FROM)\b", sql_upper):
            return "high"

    # 3. medium 级别 — 常规变更
    medium_types = {"DDL", "DML", "SET", "HSET", "HMSET"}
    if sql_type in medium_types:
        return "medium"

    # 有 WHERE 但没有 LIMIT 的 DELETE
    if re.search(r"\bDELETE\s+FROM\b", sql_upper):
        if not re.search(r"\bLIMIT\b", sql_upper):
            return "medium"

    # 任何 DDL
    ddl_keywords = ["CREATE ", "ALTER ", "RENAME ", "GRANT ", "REVOKE ", "CREATE INDEX", "CREATE UNIQUE"]
    for kw in ddl_keywords:
        if kw in sql_upper:
            return "medium"

    # 4. low 级别 — 只读 / 安全操作
    return "low"


def _evaluate_redis_risk(sql_upper: str) -> str:
    """Redis 命令风险评估。"""
    critical_cmds = ["FLUSHDB", "FLUSHALL"]
    for cmd in critical_cmds:
        if sql_upper.startswith(cmd):
            return "critical"

    # 危险写操作
    high_cmds = ["DEL ", "UNLINK ", "RENAME ", "FLUSHDB", "FLUSHALL"]
    for cmd in high_cmds:
        if sql_upper.startswith(cmd):
            return "high"

    # 一般写操作
    write_cmds = ["SET ", "HSET ", "HMSET ", "LPUSH ", "RPUSH ", "SADD ", "ZADD ",
                  "INCR", "DECR", "EXPIRE ", "PEXPIRE "]
    for cmd in write_cmds:
        if sql_upper.startswith(cmd):
            return "medium"

    # 只读命令
    return "low"


def _evaluate_mongo_risk(sql_upper: str) -> str:
    """MongoDB 操作风险评估。"""
    sql_lower = sql_upper.lower()
    critical_ops = ["dropdatabase", "drop database"]
    for op in critical_ops:
        if op in sql_lower:
            return "critical"

    high_ops = ["drop(", "dropcollection", "remove(", "deletemany"]
    for op in high_ops:
        if op in sql_lower:
            return "high"

    write_ops = ["insertone", "insertmany", "updateone", "updatemany", "deleteone",
                 "createindex", "create_collection"]
    for op in write_ops:
        if op in sql_lower:
            return "medium"

    return "low"


# ── 风险等级配置（用于前端展示 / 审批流匹配）──

RISK_LEVEL_CONFIG = {
    "low": {
        "label": "低风险",
        "color": "#52c41a",
        "description": "只读查询 / 安全操作",
        "min_approvers": 1,
    },
    "medium": {
        "label": "中风险",
        "color": "#faad14",
        "description": "常规数据变更 / DDL 操作",
        "min_approvers": 1,
    },
    "high": {
        "label": "高风险",
        "color": "#fa8c16",
        "description": "无 WHERE 删除/更新 / 大表 ALTER / 批量操作",
        "min_approvers": 2,
    },
    "critical": {
        "label": "极高风险",
        "color": "#f5222d",
        "description": "DROP TABLE / TRUNCATE / FLUSHDB 等不可逆操作",
        "min_approvers": 2,
    },
}