niltest

Executable specifications, fixed development mocks, and lightweight checks—next to the Python function they describe.

日本語 · Documentation · GitHub · Issues

niltest lets one expect.case() definition serve three purposes:

• readable behavior close to the implementation;
• a fixed mock when databases or external APIs are unavailable;
• an executable check against the real function.

It has no runtime dependencies and supports Python 3.10+ on Windows, macOS, and Linux. Pydantic-backed type validation is available through the optional niltest[pydantic] extra.

Install

# Active virtual environment (Windows, macOS, or Linux)
pip install niltest

# Windows
python -m pip install niltest

# Ubuntu / macOS
python3 -m pip install niltest

A complete example

import niltest
from niltest import Mode, expect, scenario

niltest.configure(mode=Mode.MOCK)  # Configure before @scenario is evaluated.

@scenario("Shipping fee")
def shipping_fee(subtotal: int, premium: bool = False) -> int:
    if expect:
        expect.case(
            "Premium members ship free",
            given={"subtotal": 1_000, "premium": True},
            returns=0,
        )
        expect.case(
            "Standard shipping",
            given={"subtotal": 1_000, "premium": False},
            returns=500,
        )

    return 0 if premium or subtotal >= 5_000 else 500


assert shipping_fee(1_000, premium=True) == 0

niltest.configure(mode=Mode.TEST)
result = niltest.run_tests(shipping_fee)
assert result.success

CLI and CI

niltest run your_package.specs --language en
niltest run your_package.specs --json

The CLI exits with 0 when all cases pass, 1 for specification failures, and 2 for usage or import errors. JSON output and the structured RunResult API make niltest easy to compose with CI and other developer tools.

Selecting a mode from Python

Use the Mode enum for completion and type checking. String values remain available when brevity or compatibility matters.

from niltest import Mode, configure

configure(mode=Mode.TEST)   # Recommended
configure(mode="test")      # Also supported

Expectations

returns supports:

• plain values and collections;
• dataclass instances;
• Pydantic models;
• type-only checks such as returns=UserData;
• custom validators such as returns=lambda result: result["count"] > 0;
• synchronous and asynchronous functions.

Typed specifications and inspection

Version 1.1 adds conforms_to(), backed by Pydantic's TypeAdapter. It validates models and arbitrary annotations such as list[User], unions, and Annotated constraints. During specification execution, case inputs are checked against the function signature and annotations, then normalized (for example, a dictionary can become a Pydantic model) before the implementation runs. Normal application calls and production mode inputs are never transformed by niltest.

from niltest import case, conforms_to, docs, scenario

@scenario("User lookup")
@docs(case("existing", given={"user_id": 1}, returns=conforms_to(User)))
def fetch_user(user_id: int) -> dict[str, object]:
    return {"id": user_id, "name": "Alice"}

Use niltest inspect your_package.services as a compact architecture map. Replace your_package.services with an importable module path such as your_package/services.py. Inspect imports the module and lists registered scenarios; it does not execute their cases.

niltest inspect your_package.services                 # terminal-friendly text
niltest inspect your_package.services --format json   # CI and other tools
niltest inspect your_package.services --format markdown # reviews and design docs

--json remains a compatibility alias for --format json.

Pytest integration and exception specifications

Version 1.2 exposes each niltest case as an individual pytest item. The plugin is inert unless --niltest is present, so existing pytest behavior is unchanged.

pytest --niltest --niltest-module=your_package.specs
pytest --niltest --junitxml=report.xml --cov=your_package

[tool.pytest.ini_options]
niltest_modules = ["your_package.specs"]

Use exactly one of returns or raises. match is a regular expression applied to the exception message. Exception cases validate the real implementation and never act as fixed mocks.

@scenario("Withdraw funds")
@docs(case(
    "insufficient funds",
    given={"balance": 100, "amount": 150},
    raises=ValueError,
    match="insufficient",
))
def withdraw(balance: int, amount: int) -> int:
    if amount > balance:
        raise ValueError("insufficient funds")
    return balance - amount

Inspection reports are available as text, JSON, or portable Markdown:

niltest inspect your_package.specs --format markdown

Production path

NILTEST_MODE=production is the safe default. Set NILTEST_MODE=test or NILTEST_MODE=mock before importing decorated modules when you want niltest's development features. In production mode, @scenario returns the original function without creating a wrapper. The explicit if expect: truth check remains; niltest documents this measurable cost instead of claiming absolute zero overhead.

For zero niltest cost at function-call time, use the declaration-style API. The existing inline API remains fully supported.

from niltest import case, docs, scenario

@scenario("Shipping fee")
@docs(case("Premium", given={"premium": True}, returns=0))
def shipping_fee(premium: bool) -> int:
    return 0 if premium else 500

With the default NILTEST_MODE=production, this returns the original function: no wrapper and no if expect: branch in its body.

Localization

Japanese and English are built in. niltest detects the OS locale, falls back to English, and exposes register_locale() plus a validated template for additional languages.

Verifying release provenance

Official wheel and source distributions are published through PyPI Trusted Publishing and carry Sigstore digital attestations. GitHub Releases contain the same files with GitHub build-provenance attestations tied to the source commit and CI workflow. An isolated, trusted SLSA generator also produces non-forgeable SLSA Build Level 3 provenance for every distribution.

After downloading a release asset, verify its origin with the GitHub CLI:

gh attestation verify niltest-*.whl --repo disnana/niltest
gh attestation verify niltest-*.tar.gz --repo disnana/niltest

For strict SLSA verification, download the matching .intoto.jsonl file from the GitHub Release and use slsa-verifier:

slsa-verifier verify-artifact niltest-*.whl \
  --provenance-path multiple.intoto.jsonl \
  --source-uri github.com/disnana/niltest \
  --source-branch main

Verification confirms provenance; it does not replace reviewing the package or its dependencies for your use case.

日本語

niltestは、Python関数の先頭に代表的な入力と期待値を書くことで、同じ定義を「読める仕様」「開発用モック」「実装チェック」に再利用する軽量ライブラリです。

python -m pip install niltest
niltest run your_package.specs --language ja

日本語の詳しい使い方、モック・非同期関数・バリデータ・本番モード・応用例は公式ドキュメントで確認できます。

MIT © 2026 Disnana