You can define 1 or more test functions in a single block. Each test function accepts a data object as an argument.

Within the body of the function, you can write any type of test you want to validate the input data.

After the block’s main code is executed, the output data is passed into each test function for validation. If any tests fail, then the block run will also fail.

​
Example

Here is an example of a transformer block with 2 tests:

from pandas import DataFrame

if 'transformer' not in globals():
    from mage_ai.data_preparation.decorators import transformer
if 'test' not in globals():
    from mage_ai.data_preparation.decorators import test

COLUMNS_TO_USE = ['name']


@transformer
def transform_df(df: DataFrame, *args, **kwargs) -> DataFrame:
    return df.iloc[:1][COLUMNS_TO_USE]


@test
def test_number_of_rows(df) -> None:
    assert len(df.index) >= 2, 'The output has more than 1 row.'


@test
def test_columns(df) -> None:
    assert df.columns[0] != COLUMNS_TO_USE[0], 'The output columns don’t match.'

You can combine all your data validations into 1 test function or you can split them up into multiple test functions. The benefit of splitting them up is that they can run in parallel, speeding up the data validation.

​
Log output

Each test run is recorded and can be viewed in the logs. Here is an example:

Start executing block.
--------------------------------------------------------------
2/2 tests passed.
Finish executing block.

Was this page helpful?

LoggingGreat Expectations