為資料管線撰寫單元測試

使用 Python 的 ETL 與 ELT

Jake Roach

Data Engineer

用單元測試驗證資料管線

單元測試：

常見於軟體工程流程
確保程式碼如預期運作
協助驗證資料

典型的資料管線驗證架構，包含單元測試與端到端測試。

用 pytest 進行單元測試

from pipeline import extract, transform, load

# Build a unit test, asserting the type of clean_stock_data
def test_transformed_data():
    raw_stock_data = extract("raw_stock_data.csv")
    clean_stock_data = transform(raw_data)
    assert isinstance(clean_stock_data, pd.DataFrame)

> python -m pytest

test_transformed_data .                                                     [100%]
================================ 1 passed in 1.17s ===============================

assert 與 isinstance

pipeline_type = "ETL"

# Check if pipeline_type is an instance of a str
isinstance(pipeline_type, str)

True

# Assert that the pipeline does indeed take value "ETL"
assert pipeline_type == "ETL"

# Combine assert and isinstance
assert isinstance(pipeline_type, str)

AssertionError

pipeline_type = "ETL"

# Create an AssertionError
assert isinstance(pipeline_type, float)

Traceback (most recent call last):
  File "<stdin>", line 4, in <module>
AssertionError

用 fixture 模擬資料管線元件

import pytest

@pytest.fixture()
def clean_data():
    raw_stock_data = extract("raw_stock_data.csv")
    clean_stock_data = transform(raw_data)
    return clean_stock_data

def test_transformed_data(clean_data):
    assert isinstance(clean_data, pd.DataFrame)

對 DataFrame 做單元測試

def test_transformed_data(clean_data):
    # Include other assert statements here
    ...

    # Check number of columns
    assert len(clean_data.columns) == 4

    # Check the lower bound of a column
    assert clean_data["open"].min() >= 0

    # Check the range of a column by chaining statements with "and"
    assert clean_data["open"].min() >= 0 and clean_data["open"].max() <= 1000

一起來練習吧！

使用 Python 的 ETL 與 ELT