PayrollManagementSystem/backend/tests/unit/test_audit_serialization.py

"""操作日志入库前处理单元测试。

守的是入库前的**截断与脱敏**：

- 截断：各列超长在 MySQL 严格模式下抛 1406，而「审计失败不影响业务」的约定
  会把该异常吞掉 —— 整条日志静默消失；request_data 尤甚，中文一个字符
  占 3 字节，必须按 **UTF-8 字节边界**切，否则会切出乱码残字
- 脱敏：密码是最容易从日志泄漏的凭据，request_data 入库前必经 scrub
"""

import json
from datetime import datetime

from app.models.operation_log import (
    ERROR_MESSAGE_MAX_LEN,
    REQUEST_DATA_MAX_BYTES,
    SUMMARY_MAX_LEN,
)
from app.repositories.operation_log_repository import _clip, _serialize_request_data


class TestClip:
    """按字符数截断（长度常量与数据库列宽均按字符定义）。"""

    def test_超长截断至上限(self) -> None:
        result = _clip("x" * (SUMMARY_MAX_LEN + 50), SUMMARY_MAX_LEN)
        assert result is not None
        assert len(result) == SUMMARY_MAX_LEN

    def test_未超长原样返回(self) -> None:
        assert _clip("正常长度", SUMMARY_MAX_LEN) == "正常长度"

    def test_空值返回None(self) -> None:
        assert _clip(None, ERROR_MESSAGE_MAX_LEN) is None
        assert _clip("", ERROR_MESSAGE_MAX_LEN) is None

    def test_中文按字符数截断不切字(self) -> None:
        """str 切片天然按字符，不会切出半个汉字。"""
        result = _clip("中" * 100, 10)
        assert result == "中" * 10


class TestSerializeRequestData:
    """序列化：JSON 化 → 脱敏 → 字节边界截断。"""

    def test_常规参数原样序列化(self) -> None:
        text = _serialize_request_data({"status": 0, "roles": ["user"]})
        assert text is not None
        assert json.loads(text) == {"status": 0, "roles": ["user"]}

    def test_空值返回None(self) -> None:
        assert _serialize_request_data(None) is None
        assert _serialize_request_data({}) is None

    def test_密码值被抹除而键名保留(self) -> None:
        """键名保留便于审计阅读"提交过密码字段"，值必须抹掉。"""
        for payload in ({"password": "S3cret@2026"}, {"new_password": "S3cret@2026"}):
            text = _serialize_request_data(payload) or ""
            assert "S3cret@2026" not in text
            assert "***" in text

    def test_手机号被脱敏(self) -> None:
        text = _serialize_request_data({"phone": "13812345678"}) or ""
        assert "13812345678" not in text
        # phone 键已加入结构化抹除清单（整值 ***）—— 带空格/横线分隔的号码
        # （如 138 0013 8000）会穿过分值文本正则，键名整值抹除是更强的防线
        assert '"phone": "***"' in text

    def test_带分隔符的手机号被脱敏(self) -> None:
        """修复回归：'138 0013 8000' 类分隔符写法曾同时穿过分值正则与键名清单。"""
        for payload in (
            {"phone": "138 0013 8000"},
            {"phone": "138-0013-8000"},
            {"mobile": "13800138000"},
        ):
            text = _serialize_request_data(payload) or ""
            assert "13800138000" not in text.replace(" ", "").replace("-", "")
            assert "***" in text

    def test_超长中文按字节边界截断且无残字(self) -> None:
        """75000 字节（25000 个汉字）必然超限：
        若按字符截断则字节数三倍超标；若按字节硬切可能切出半个字符。
        正确结果应是原始 JSON 串的**干净前缀**。"""
        payload = {"remark": "中" * 25000}
        text = _serialize_request_data(payload)
        assert text is not None
        assert len(text.encode("utf-8")) <= REQUEST_DATA_MAX_BYTES
        assert "<EFBFBD>" not in text  # 无替换字符
        raw = json.dumps(payload, ensure_ascii=False)
        assert raw.startswith(text)

    def test_未超长的中文完整保留(self) -> None:
        payload = {"remark": "中" * 100}
        text = _serialize_request_data(payload)
        assert text is not None
        assert json.loads(text) == payload

    def test_不可JSON序列化的值降级为字符串(self) -> None:
        text = _serialize_request_data({"at": datetime(2026, 10, 10, 12, 0, 0)})
        assert text is not None
        assert "2026-10-10" in text


class TestStructuredSanitization:
    """结构化抹除 —— 按键名整值替换，不依赖文本正则。

    回归背景：此前用文本正则脱敏，有三种逃逸形态 —— 值以空白开头时
    整体匹配失败（**完全不脱敏**）；值含空格/逗号时只抹首段；值含引号时
    既残留明文又破坏 JSON 结构。密码校验只查字节数、前后端都不 trim，
    这些形态都是正常可用密码。
    """

    def test_前导空白密码被完全抹除(self) -> None:
        """回归：此前 `" abc12345"` 原样明文入库。"""
        text = _serialize_request_data({"new_password": " abc12345"}) or ""
        assert "abc12345" not in text
        assert json.loads(text) == {"new_password": "***"}

    def test_含空格逗号引号均整体抹除且JSON完整(self) -> None:
        for password in ("abc 12345", "abc,def", 'abc"def', "abc&def"):
            text = _serialize_request_data({"password": password}) or ""
            assert password not in text
            assert json.loads(text) == {"password": "***"}

    def test_嵌套结构与键名大小写(self) -> None:
        text = (
            _serialize_request_data({"user": {"New_Password": "s3"}, "list": [{"TOKEN": "t"}]})
            or ""
        )
        assert "s3" not in text
        assert json.loads(text) == {"user": {"New_Password": "***"}, "list": [{"TOKEN": "***"}]}

    def test_非敏感键不受影响(self) -> None:
        text = _serialize_request_data({"keyword": "a_b", "remark": "正常备注"}) or ""
        assert json.loads(text) == {"keyword": "a_b", "remark": "正常备注"}

    def test_孤立代理字符不再抛错(self) -> None:
        """回归：直接 encode('utf-8') 遇 \\ud800 抛 UnicodeEncodeError，
        该请求的审计记录会被兜底逻辑吞掉、静默丢失。"""
        text = _serialize_request_data({"username": "a\ud800b"})
        assert text is not None