【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

import pandas as pd

def add\_amount(df: pd.DataFrame) -> pd.DataFrame:
assert {"qty","price"}.issubset(df.columns)
return df.assign(amount=df\["qty"] \* df\["price"])  # 純関数

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
df = pd.DataFrame({"qty":\[2,3], "price":\[500, 1200]})
\# Act
got = add\_amount(df)
\# Assert
expect = pd.DataFrame({"qty":\[2,3], "price":\[500,1200], "amount":\[1000,3600]})
assert\_frame\_equal(got, expect)

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()

# 乱数固定（全テストに適用）

@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

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)

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

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

from pathlib import Path

def test\_save\_text\_writes\_to\_tmp(tmp\_path):
from src.report import save\_text
p = tmp\_path / "reports/hello.txt"
got = save\_text(p, "hello")
assert got.exists() and got.read\_text() == "hello"

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")

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

[tool.pytest.ini_options]
minversion = "7.0"
addopts = "-q --maxfail=1 --disable-warnings"
testpaths = ["tests"]
filterwarnings = ["ignore::DeprecationWarning"]

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）