【コピペOK】pytestで“壊れないPython”を作る12ステップ

project/
  ├─ src/
  │   ├─ features.py      # 前処理/特徴量
  │   ├─ report.py        # Excel/図の出力
  │   └─ db.py            # SQLite I/O
  ├─ tests/
  │   ├─ test_features.py
  │   ├─ test_db.py
  │   └─ conftest.py      # 共有fixture
  ├─ data/ (raw/warehouse)  # 非追跡推奨
  ├─ reports/               # 出力（非追跡推奨）
  ├─ requirements.txt
  ├─ pyproject.toml        # pytest設定
  └─ README.md

pip install pytest pytest-cov pandas
pytest -q   # テスト探索→実行

# src/features.py
import pandas as pd


def add_amount(df: pd.DataFrame) -> pd.DataFrame:
# 失敗時の原因がすぐ分かるよう、前提条件を明示
assert {"qty", "price"}.issubset(df.columns)
# 副作用なし（受け取ったDFを壊さない）
return df.assign(amount=df["qty"] * df["price"])

# tests/test_features.py
import pandas as pd
from pandas.testing import assert_frame_equal
from src.features import add_amount


def test_add_amount_multiplies_qty_and_price():
# Arrange（入力準備）
df = pd.DataFrame({"qty": [2, 3], "price": [500, 1200]})
# Act（関数を呼ぶ）
got = add_amount(df)
# Assert（期待と比較）
expect = pd.DataFrame({
"qty": [2, 3], "price": [500, 1200], "amount": [1000, 3600]
})
assert_frame_equal(got, expect)

# tests/conftest.py
import pandas as pd
import sqlite3
import pytest
import random, numpy as np


@pytest.fixture
def sample_df():
return pd.DataFrame({
"order_date": ["2024-04-01"], "store": ["S01"], "qty": [2], "price": [500]
})


@pytest.fixture
def sqlite_conn(tmp_path: pytest.TempPathFactory):
db = tmp_path / "test.sqlite"
con = sqlite3.connect(db)
yield con
con.close()


# すべてのテストで乱数を固定（再現性の土台）
@pytest.fixture(autouse=True)
def fixed_seed():
random.seed(42); np.random.seed(42)
yield

import pytest
import pandas as pd
from src.features import add_amount

@pytest.mark.parametrize("qty, price, amount", \[
(2, 500, 1000),
(0, 500, 0),
(5, 0, 0),
])
def test\_add\_amount\_param(qty, price, amount):
df = pd.DataFrame({"qty":\[qty], "price":\[price]})
got = add\_amount(df)
assert int(got.loc\[0, "amount"]) == amount

from pandas.testing import assert_series_equal
from src.features import add_amount

def test\_columns\_and\_dtypes(sample\_df):
got = add\_amount(sample\_df)
assert set(\["order\_date","store","qty","price","amount"]).issubset(got.columns)
assert got\["qty"].dtype.kind in "iu"   # int/uint
assert got\["price"].dtype.kind in "if"  # int/float
assert not got.isna().any().any()      # 必要に応じて欠損も確認

# src/db.py
import sqlite3, pandas as pd


def write_sales(con: sqlite3.Connection, df: pd.DataFrame) -> None:
df.to_sql("sales", con, if_exists="append", index=False)


def read_monthly(con: sqlite3.Connection) -> pd.DataFrame:
q = """
WITH m AS (
SELECT strftime('%Y-%m', order_date) AS ym, store,
SUM(qty*price) AS sales
FROM sales GROUP BY ym, store
) SELECT * FROM m
"""
return pd.read_sql_query(q, con)

# tests/test_db.py
import pandas as pd
from src.db import write_sales, read_monthly


def test_sqlite_roundtrip(sqlite_conn):
df = pd.DataFrame({
"order_date": ["2024-04-01", "2024-04-02"],
"store": ["S01", "S01"],
"qty": [2, 3], "price": [500, 1200],
})
write_sales(sqlite_conn, df)
got = read_monthly(sqlite_conn)
assert set(["ym", "store", "sales"]).issubset(got.columns)
assert float(got.loc[0, "sales"]) > 0

# src/report.py
from pathlib import Path


def save_text(path: Path, content: str) -> Path:
path.parent.mkdir(parents=True, exist_ok=True)
path.write_text(content, encoding="utf-8")
return path

# tests/test_report.py
from pathlib import Path
from src.report import save_text


def test_save_text_writes_to_tmp(tmp_path):
p = tmp_path / "reports/hello.txt"
got = save_text(p, "hello")
assert got.exists() and got.read_text() == "hello"

# src/plot.py
import matplotlib.pyplot as plt


def plot_trend(df, out):
plt.figure()
plt.plot(df["x"], df["y"], marker="o")
plt.tight_layout()
plt.savefig(out, dpi=120, bbox_inches="tight")

# tests/test_plot.py
import pandas as pd
from src.plot import plot_trend


def test_plot_trend_saves_png(tmp_path):
df = pd.DataFrame({"x": ["2024-04", "2024-05"], "y": [10, 12]})
out = tmp_path / "trend.png"
plot_trend(df, out)
assert out.exists() and out.stat().st_size > 0

import numpy as np
from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import StratifiedKFold, cross_val_score
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression

def test\_cv\_auc\_is\_reasonable():
Xy = load\_breast\_cancer()
X, y = Xy.data, Xy.target
pipe = Pipeline(\[
("sc", StandardScaler()),
("lr", LogisticRegression(max\_iter=1000, random\_state=42, n\_jobs=-1))
])
cv = StratifiedKFold(n\_splits=5, shuffle=True, random\_state=42)
scores = cross\_val\_score(pipe, X, y, scoring="roc\_auc", cv=cv, n\_jobs=-1)
assert np.mean(scores) > 0.98  # 下限
assert np.std(scores) < 0.02   # ばらつき上限

import sys, pytest

@pytest.mark.skipif(sys.platform == "win32", reason="OS依存のためスキップ")
def test\_posix\_only():
assert True

@pytest.mark.xfail(reason="将来対応予定の仕様")
def test\_future\_behavior():
assert 1/0 == 0

# pyproject.toml（一例）
[tool.pytest.ini_options]
minversion = "7.0"
addopts = "-q --maxfail=1 --disable-warnings"
testpaths = ["tests"]
filterwarnings = ["ignore::DeprecationWarning"]

# カバレッジ（任意）
pip install pytest-cov
pytest --cov=src --cov-report=term-missing

assert 1 == 1
import math; assert math.isclose(0.1+0.2, 0.3, rel_tol=1e-9)
from pandas.testing import assert_frame_equal, assert_series_equal

import pytest
from src.features import add_amount
import pandas as pd

def test\_add\_amount\_raises\_when\_columns\_missing():
with pytest.raises(AssertionError):
add\_amount(pd.DataFrame({"qty":\[1]}))

pytest -k amount        # 名前にamountを含むテストだけ
pytest tests/test_db.py::test_sqlite_roundtrip -q
pytest -x               # 最初の失敗で停止
pytest -n auto          # 並列実行（pytest-xdist）