Python测试Mock BigQuery表Schema失败，返回MagicMock对象求排查

我有一个用于获取BigQuery表Schema的简单Python函数：

# Get columns from BigQuery table
def get_schema_for_table(bq_client, source, schema, table):
    bq_table = bq_client.get_table(f'{source}.{schema}.{table}')
    return bq_table.schema

为避免使用真实表，我希望在测试中创建Mock Schema和Mock Client，当前测试代码如下：

def test_get_schema_for_table(mocker):
    mock_client = mocker.patch('google.cloud.bigquery.Client', autospec=True)
    Table.schema = MagicMock(return_value=[
        SchemaField('column_1', 'STRING', 'NULLABLE', None, (), None),
        SchemaField('column_2', 'STRING', 'NULLABLE', None, (), None),
        SchemaField('column_3', 'STRING', 'NULLABLE', None, (), None)
    ])
    source = 'project'
    schema_name = 'dataset'
    table_name = 'existing_table'
    expected_schema = [SchemaField('column_1', 'STRING', 'NULLABLE', None, (), None), SchemaField('column_2', 'STRING', 'NULLABLE', None, (), None), SchemaField('column_3', 'STRING', 'NULLABLE', None, (), None)]
    returned_schema = get_schema_for_table(mock_client,source,schema_name,table_name)    
    print(returned_schema)
    assert returned_schema == expected_schema

但打印returned_schema时得到的是：<MagicMock name='Client.get_table().schema' id='4591155568'>，请问我哪里出错了？

你的问题在于没有正确Mock get_table()方法的返回值，直接给Table.schema赋值并不会关联到mock client调用get_table()后返回的对象上。

修正后的测试代码：

from unittest.mock import MagicMock
from google.cloud.bigquery import SchemaField

def test_get_schema_for_table(mocker):
    # 获取模拟的Client实例
    mock_client = mocker.patch('google.cloud.bigquery.Client', autospec=True).return_value
    # 创建模拟的Table对象并设置预设的schema
    mock_table = MagicMock()
    mock_table.schema = [
        SchemaField('column_1', 'STRING', 'NULLABLE', None, (), None),
        SchemaField('column_2', 'STRING', 'NULLABLE', None, (), None),
        SchemaField('column_3', 'STRING', 'NULLABLE', None, (), None)
    ]
    # 指定get_table方法返回这个模拟Table
    mock_client.get_table.return_value = mock_table

    source = 'project'
    schema_name = 'dataset'
    table_name = 'existing_table'
    expected_schema = [
        SchemaField('column_1', 'STRING', 'NULLABLE', None, (), None),
        SchemaField('column_2', 'STRING', 'NULLABLE', None, (), None),
        SchemaField('column_3', 'STRING', 'NULLABLE', None, (), None)
    ]
    
    returned_schema = get_schema_for_table(mock_client, source, schema_name, table_name)    
    assert returned_schema == expected_schema

核心说明： #

• 原代码中mock_client = mocker.patch(...)得到的是Mock类，而非实例，必须加上.return_value才能拿到可调用的客户端实例。
• 必须让get_table()返回带有预设schema属性的模拟对象，这样函数调用bq_client.get_table()时，才能获取到你定义的schema列表，而非默认的MagicMock占位对象。

内容的提问来源于stack exchange，提问作者nimgwfc