我最终为我们的test-utils库编写了一个方便的方法

/**
 *  Utility method to test for a specific error class and message in Jest
 * @param {fn, expectedErrorClass, expectedErrorMessage }
 * @example   failTest({
      fn: () => {
        return new MyObject({
          param: 'stuff'
        })
      },
      expectedErrorClass: MyError,
      expectedErrorMessage: 'stuff not yet implemented'
    })
 */
  failTest: ({ fn, expectedErrorClass, expectedErrorMessage }) => {
    try {
      fn()
      expect(true).toBeFalsy()
    } catch (err) {
      let isExpectedErr = err instanceof expectedErrorClass
      expect(isExpectedErr).toBeTruthy()
      expect(err.message).toBe(expectedErrorMessage)
    }
  }

如果你正在使用Promises:

await expect(Promise.reject(new HttpException('Error message', 402)))
  .rejects.toThrowError(HttpException);

这有点奇怪，但它是有效的，以我之见是很好的可读性:

it('should throw Error with message \'UNKNOWN ERROR\' when no parameters were passed', () => {
  try {
      throwError();
      // Fail test if above expression doesn't throw anything.
      expect(true).toBe(false);
  } catch (e) {
      expect(e.message).toBe("UNKNOWN ERROR");
  }
});

Catch块捕获您的异常，然后您可以测试您引发的Error。奇怪的期待(真正的).toBe(假);如果没有抛出预期的错误，则需要测试失败。否则，这一行永远无法到达(Error应该在它们之前引发)。

@Kenny Body提出了一个更好的解决方案，如果你使用expect.assertions()，可以提高代码质量:

it('should throw Error with message \'UNKNOWN ERROR\' when no parameters were passed', () => {
  expect.assertions(1);
  try {
      throwError();
  } catch (e) {
      expect(e.message).toBe("UNKNOWN ERROR");
  }
});

请参阅原始答案和更多解释:如何在Jest中测试抛出异常的类型

2022年编辑:

要使用这种方法而不触发no-conditional-expect规则(如果你使用eslint-plugin-jest)，该规则的文档建议使用错误包装器:

class NoErrorThrownError extends Error {}

const getError = async <TError>(call: () => unknown): Promise<TError> => {
  try {
    await call();

    throw new NoErrorThrownError();
  } catch (error: unknown) {
    return error as TError;
  }
};

describe('when the http request fails', () => {
  it('includes the status code in the error', async () => {
    const error = await getError(async () => makeRequest(url));

    // check that the returned error wasn't that no error was thrown
    expect(error).not.toBeInstanceOf(NoErrorThrownError);
    expect(error).toHaveProperty('statusCode', 404);
  });
});

参见:no-conditional-expect文档

开玩笑地说，你必须将一个函数传递给expect(函数)。toThrow(<空白或错误类型>)。

例子:

test("Test description", () => {
  const t = () => {
    throw new TypeError();
  };
  expect(t).toThrow(TypeError);
});

或者如果你还想检查错误信息:

test("Test description", () => {
  const t = () => {
    throw new TypeError("UNKNOWN ERROR");
  };
  expect(t).toThrow(TypeError);
  expect(t).toThrow("UNKNOWN ERROR");
});

如果需要测试现有函数是否抛出一组参数，则必须将其包装在expect()中的匿名函数中。

例子:

test("Test description", () => {
  expect(() => {http.get(yourUrl, yourCallbackFn)}).toThrow(TypeError);
});

检查toThrow方法。

您必须将代码包装在一个额外的回调函数中!

您应该同时检查错误消息及其类型。

例如:

expect(
  () => { // additional function wrap
    yourCodeToTest();
  }
).toThrow(
  new RangeError('duplicate prevArray value: A')
);

由于额外的回调换行，代码不会立即运行，因此jest将能够捕获它。

您应该始终检查错误消息，以确保您正在检查正确的抛出情况，而不会得到代码可能抛出的另一个错误。

检查错误类型也很好，因此客户端代码可以依赖它。

一个好方法是创建自定义错误类并模拟它们。然后你可以断言任何你想要的东西。

MessedUpError.ts

type SomeCrazyErrorObject = {
  [key: string]: unknown,
}

class MessedUpError extends Error {
  private customErrorData: SomeCrazyErrorObject = {};

  constructor(err?: string, data?: SomeCrazyErrorObject) {
    super(err || 'You messed up');

    Object.entries(data ?? {}).forEach(([Key, value]) => {
      this.customErrorData[Key] = value;
    });
    Error.captureStackTrace(this, this.constructor);
  }

  logMe() {
    console.log(this.customErrorData);
  }
}

export default MessedUpError;

messedUpError.test.ts

import MessedUpError from './MessedUpError';

jest.mock('./MessedUpError', () => jest.fn().mockImplementation((...args: any[]) => ({
  constructor: args,
  log: () => {},
})));

type MessedUpErrorContructorParams = Expand<typeof MessedUpError['prototype']>
const MessedUpErrorMock = MessedUpError as unknown as jest.Mock<MessedUpError, [MessedUpErrorContructorParams]>;

const serverErrorContructorCall = (i = 0) => ({
  message: MessedUpErrorMock.mock.calls[i][0],
  ...MessedUpErrorMock.mock.calls[i][1] || {},
});

beforeEach(() => {
  MessedUpErrorMock.mockClear();
});

test('Should throw', async () => {
  try {
    await someFunctionThatShouldThrowMessedUpError();
  } catch {} finally {
    expect(MessedUpErrorMock).toHaveBeenCalledTimes(1);
    const constructorParams = serverErrorContructorCall();
    expect(constructorParams).toHaveProperty('message', 'You messed up');
    expect(constructorParams).toHaveProperty('customErrorProperty', 'someValue');
  }
});

断言总是在finally子句中。通过这种方式，它将始终被断言。即使测试没有抛出任何错误。