Skip to main content

Basic notebook checks. Do they run? Do they contain lint?

Project description

See Build Status on Travis CI See Build Status on AppVeyor

nbsmoke

Basic notebook smoke tests: Do they run ok? Do they contain lint?

WARNING: early stage proof of concept; work in progress. Use at your own risk.

In particular, this extension is supposed to handle ipython magics as far as possible, but has not yet been widely tested.


This Pytest plugin was generated with Cookiecutter along with @hackebrot’s Cookiecutter-pytest-plugin template.

Installation

You can install nbsmoke via pip from PyPI:

$ pip install nbsmoke

Or you can install nbsmoke via conda from anaconda.org:

$ conda install -c pyviz/label/dev -c conda-forge nbsmoke

Usage

Check all notebooks in examples/ run without errors:

$ pytest --nbsmoke-run examples/

Check all notebooks in examples/ run without errors, and store html to look at afterwards:

$ pytest --nbsmoke-run --store-html=/scratch examples/

Lint check notebooks in examples/:

$ pytest --nbsmoke-lint notebooks/

Instead of all files in a directory, you can specify a list e.g.:

$ pytest --nbsmoke-run notebooks/Untitled*.ipynb

Additional options are available by standard pytest ‘ini’ configuration in setup.cfg, pytest.ini, or tox.ini:

[pytest]
# when running, seconds allowed per cell (see nbconvert timeout)
cell_timeout = 600

# notebooks to skip running; one case insensitive re to match per line
skip_run = ^.*skipme\.ipynb$
           ^.*skipmetoo.*$

# case insensitive re to match for file to be considered notebook;
# defaults to ``^.*\.ipynb``
it_is_nb_file = ^.*\.something$

nbsmoke supports # noqa comments to mark that something should be ignored during lint checking.

What’s the point?

Although more sophisticated testing of notebooks is possible (e.g. see nbval), just checking that notebooks run from start to finish without error in a fresh kernel (or on a neutral CI service) can be useful during development. Practical experience of working on several projects with notebooks confirms this, but that’s all the evidence I have.

Checking notebooks for lint might seem trivial/pointless, but it frequently uncovers unused names (typically unused imports). It’s also quite common to find python 2 vs 3 problems, and sometimes undefined names - in a way that’s faster than running the notebook (over multiple versions of python).

Unused imports/names themselves might seem trivial, but they can hinder understanding of a notebook by readers, or add dependencies that are not required.

Using # noqa: explanation in a notebook might seem like overkill, but the intention is to encourage unavoidable/desirable ‘mysterious imports’ to be clarified. E.g. if you’re importing something for its side effects, it’s very helpful to inform the reader of that.

Pyflakes is used as the underlying linter because “Pyflakes makes a simple promise: it will never complain about style, and it will try very, very hard to never emit false positives.”

Contributing

First, install using pip install -e .. Then run the tests using tox or pytest -v tests/.

New release to PyPI and anaconda.org: git tag -a vX.Y.Z -m "Something about release" && git push --tags.

License

Distributed under the terms of the BSD-3 license, “nbsmoke” is free and open source software.

Issues

If you encounter any problems, please file an issue (ideally including a copy of any problematic notebook).

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

nbsmoke-0.2.1.tar.gz (31.1 kB view hashes)

Uploaded Source

Built Distribution

nbsmoke-0.2.1-py2.py3-none-any.whl (15.6 kB view hashes)

Uploaded Python 2 Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page