如何从tenacity中获取底层AssertionError的完整错误信息？

环境与代码 #

使用Tenacity 9.0.0 + Python 3.9，实现重试逻辑的代码如下：

@retry(stop=stop_after_attempt(6), wait=wait_fixed(5))
def assert_response():
    logging.info("Retrying get flow logs")
    es.indices.refresh(index='monitoring')
    response = requests.get(url, headers=headers, json=data, params=params)
    assert response.status_code == 200, f"Response status is {response.status_code} and body: {response.text}"
    response_json = response.json()
    logging.info("Response: " + str(response_json))
    assert response_json['meta']['run_id'] == 1742403068092, "run_id does not match"

try:
    assert_response()
except RetryError as e:
    underlying_error = e.last_attempt.exception()
    error_message = str(underlying_error)
    logging.error(f"Failed to get flow logs after retries: {underlying_error}")
    raise AssertionError(f"Failed to get flow logs after multiple retries: {error_message}") from underlying_error

当前问题 #

日志仅显示片段错误信息，无法获取断言中定义的完整提示（如"run_id does not match"）：

integration-tests-1  | FAILED test_apps/monitoring_service/test_es_monitoring_dao.py::test - AssertionError: Failed to get flow logs after multiple retries: 'run_id'
integration-tests-1  | ====================== 1 failed, 130 deselected in 11.42s ======================
integration-tests-1  | Result of the pytest script: 1

原因分析 #

日志里的'run_id'是KeyError的错误内容，说明在触发run_id匹配断言之前，代码先因为response_json['meta']['run_id']中的meta或run_id键不存在抛出了KeyError，导致断言的错误信息根本没机会触发。

解决方案 #

方案1：先检查键是否存在，再执行断言 #

在原断言前增加键存在性检查，确保只有当键存在时才执行匹配断言，同时给出明确的缺失提示：

@retry(stop=stop_after_attempt(6), wait=wait_fixed(5))
def assert_response():
    logging.info("Retrying get flow logs")
    es.indices.refresh(index='monitoring')
    response = requests.get(url, headers=headers, json=data, params=params)
    assert response.status_code == 200, f"Response status is {response.status_code} and body: {response.text}"
    response_json = response.json()
    logging.info("Response: " + str(response_json))
    # 先检查meta和run_id是否存在
    assert 'meta' in response_json, "meta field not found in response"
    assert 'run_id' in response_json['meta'], "run_id field not found in meta"
    # 再执行run_id匹配断言
    assert response_json['meta']['run_id'] == 1742403068092, "run_id does not match"

方案2：使用.get()安全取值，统一断言提示 #

用字典的.get()方法避免KeyError，同时在断言中包含实际获取到的值，方便排查：

@retry(stop=stop_after_attempt(6), wait=wait_fixed(5))
def assert_response():
    logging.info("Retrying get flow logs")
    es.indices.refresh(index='monitoring')
    response = requests.get(url, headers=headers, json=data, params=params)
    assert response.status_code == 200, f"Response status is {response.status_code} and body: {response.text}"
    response_json = response.json()
    logging.info("Response: " + str(response_json))
    # 安全获取run_id，不存在则返回None
    run_id = response_json.get('meta', {}).get('run_id')
    # 统一断言，覆盖键缺失和值不匹配两种情况
    assert run_id == 1742403068092, f"run_id does not match or is missing. Expected: 1742403068092, Got: {run_id}"

方案3：捕获KeyError并转换为明确的断言错误 #

如果需要保留原有的直接取值方式，可以在函数内部捕获KeyError，将其转换为包含完整信息的AssertionError：

@retry(stop=stop_after_attempt(6), wait=wait_fixed(5))
def assert_response():
    logging.info("Retrying get flow logs")
    es.indices.refresh(index='monitoring')
    response = requests.get(url, headers=headers, json=data, params=params)
    assert response.status_code == 200, f"Response status is {response.status_code} and body: {response.text}"
    response_json = response.json()
    logging.info("Response: " + str(response_json))
    try:
        assert response_json['meta']['run_id'] == 1742403068092, "run_id does not match"
    except KeyError as e:
        raise AssertionError(f"Missing key in response: {str(e)}") from e

验证修改效果 #

修改后，无论是键缺失还是值不匹配，都会抛出包含完整信息的AssertionError，最终日志会显示你定义的完整错误提示，比如：

AssertionError: Failed to get flow logs after multiple retries: run_id does not match

或者

AssertionError: Failed to get flow logs after multiple retries: Missing key in response: 'meta'

内容的提问来源于stack exchange，提问作者tuk