Home

Awesome

ipytest - Pytest in Jupyter notebooks

PyPI | Usage | Global state | How does it work? | Changes | Reference | Development | Related packages | License

ipytest allows you to run Pytest in Jupyter notebooks. ipytest aims to give access to the full pytest experience and to make it easy to transfer tests out of notebooks into separate test files.

Usage

Install ipytest by running

pip install ipytest

The suggested way to import ipytest is

import ipytest
ipytest.autoconfig()

Afterwards in a new cell, tests can be executed as in

%%ipytest -qq

def test_example():
    assert [1, 2, 3] == [1, 2, 3]

This command will first delete any previously defined tests, execute the cell and then run pytest. For further details on how to use ipytest see the example notebook or the reference below.

Global state

There are multiple sources of global state when using pytest inside the notebook:

  1. pytest will find any test function ever defined. This behavior can lead to unexpected results when test functions are renamed, as their previous definition is still available inside the kernel. Running %%ipytest per default deletes any previously defined tests. As an alternative the ipytest.clean() function allows to delete previously defined tests.
  2. Python's module system caches imports and therefore acts as a global state. To test the most recent version of any module, the module needs to be reloaded. ipytest offers the ipytest.force_reload() function. The autoreload extension of IPython may also help here. To test local packages, it is advisable to install them as development packages, e.g., pip install -e ..
  3. For async code, IPython will create an event loop in the current thread. This setup may interfere with async tests. To support these use cases, ipytest supports running tests in a separate thread. Simply setup ipytest via ipytest.autoconfig(run_in_thread=True).

How does it work?

In its default configuration (via autoconfig()), ipytest performs the following steps:

  1. Register pytest's assertion rewriter with the IPython kernel. The rewriter will rewrite any assert statements entered into the notebook to give better error messages. This change will affect also non test based code, but should generally improve the development experience.
  2. Ensure the notebook can be mapped to a file. ipytest will create a temporary file in the current directory and remove if afterwards.
  3. Register the notebook scope temporarily as a module. This step is necessary to allow pytest's doctest plugin to import the notebook.
  4. Call pytest with the name of the temporary module

NOTE: Some notebook implementations modify the core IPython package and magics may not work correctly (see here or here). In this case, using ipytest.run() and ipytest.clean() directly should still work as expected.

Reference

autoconfig | %%ipytest | config | exit_code | run | clean | force_reload | Error | ipytest.cov

<!-- minidoc "function": "ipytest.autoconfig", "header_depth": 3 -->

ipytest.autoconfig(rewrite_asserts=<default>, magics=<default>, clean=<default>, addopts=<default>, run_in_thread=<default>, defopts=<default>, display_columns=<default>, raise_on_error=<default>, coverage=<default>)

Configure ipytest with reasonable defaults.

Specifically, it sets:

See ipytest.config for details.

<!-- minidoc -->

%%ipytest ...

<!-- minidoc "function": "ipytest._impl.ipytest_magic", "header": false, "header_depth": 3 -->

IPython magic to first execute the cell, then execute ipytest.run().

Note: the magics are only available after running ipytest.autoconfig() or ipytest.config(magics=True).

It cleans any previously found tests, i.e., only tests defined in the current cell are executed. To disable this behavior, use ipytest.config(clean=False).

Any arguments passed on the magic line are interpreted as command line arguments to to pytest. For example calling the magic as

%%ipytest -qq

is equivalent to passing -qq to pytest. The arguments are formatted using Python's standard string formatting. Currently, only the {MODULE} variable is understood. It is replaced with the filename associated with the notebook. In addition node ids for tests can be generated by using the test name as a key, e.g., {test_example} will expand to {MODULE}::test_example.

The keyword arguments passed to ipytest.run() can be customized by including a comment of the form # ipytest: arg1=value1, arg=value2 in the cell source. For example:

%%ipytest {MODULE}::test1
# ipytest: defopts=False

is equivalent to ipytest.run("{MODULE}::test1", defopts=False). In this case, it deactivates default arguments and then instructs pytest to only execute test1.

NOTE: In the default configuration %%ipytest will not raise exceptions, when tests fail. To raise exceptions on test errors, e.g., inside a CI/CD context, use ipytest.autoconfig(raise_on_error=True).

<!-- minidoc --> <!-- minidoc "function": "ipytest.config", "header_depth": 3 -->

ipytest.config(rewrite_asserts=<keep>, magics=<keep>, clean=<keep>, addopts=<keep>, run_in_thread=<keep>, defopts=<keep>, display_columns=<keep>, raise_on_error=<keep>, coverage=<default>)

Configure ipytest

To update the configuration, call this function as in:

ipytest.config(rewrite_asserts=True)

The following settings are supported:

<!-- minidoc -->

ipytest.exit_code

The return code of the last pytest invocation.

<!-- minidoc "function": "ipytest.run", "header_depth": 3 -->

ipytest.run(*args, module=None, plugins=(), run_in_thread=<default>, raise_on_error=<default>, addopts=<default>, defopts=<default>, display_columns=<default>, coverage=<default>)

Execute all tests in the passed module (defaults to __main__) with pytest.

This function is a thin wrapper around pytest.main and will execute any tests defined in the current notebook session.

NOTE: In the default configuration ipytest.run() will not raise exceptions, when tests fail. To raise exceptions on test errors, e.g., inside a CI/CD context, use ipytest.autoconfig(raise_on_error=True).

Parameters:

The following parameters override the config options set with ipytest.config() or ipytest.autoconfig().

Returns: the exit code of pytest.main.

<!-- minidoc --> <!-- minidoc "function": "ipytest.clean", "header_depth": 3 -->

ipytest.clean(pattern=<default>, *, module=None)

Delete tests with names matching the given pattern.

In IPython the results of all evaluations are kept in global variables unless explicitly deleted. This behavior implies that when tests are renamed the previous definitions will still be found if not deleted. This method aims to simply this process.

An effective pattern is to start with the cell containing tests with a call to ipytest.clean(), then defined all test cases, and finally call ipytest.run(). This way renaming tests works as expected.

Parameters:

<!-- minidoc --> <!-- minidoc "function": "ipytest.force_reload", "header_depth": 3 -->

ipytest.force_reload(*include, modules: Optional[Dict[str, module]] = None)

Ensure following imports of the listed modules reload the code from disk

The given modules and their submodules are removed from sys.modules. Next time the modules are imported, they are loaded from disk.

If given, the parameter modules should be a dictionary of modules to use instead of sys.modules.

Usage:

ipytest.force_reload("my_package")
from my_package.submodule import my_function
<!-- minidoc --> <!-- minidoc "class": "ipytest.Error", "header_depth": 3 -->

ipytest.Error(exit_code)

Error raised by ipytest on test failure

<!-- minidoc --> <!-- minidoc "module": "ipytest.cov", "header_depth": 3 -->

ipytest.cov

A coverage.py plugin to support coverage in Jupyter notebooks

The plugin must be enabled in a .coveragerc next to the current notebook or the pyproject.toml file. See the coverage.py docs for details. In case of a .coveragerc file, the minimal configuration reads:

[run]
plugins =
    ipytest.cov

With this config file, coverage information can be collected using pytest-cov with

%%ipytest --cov

def test():
    ...

ipytest.autoconfig(coverage=True) automatically adds the --cov flag and the path of a generated config file to the Pytest invocation. In this case no further configuration is required.

There are some known issues of ipytest.cov

ipytest.cov.translate_cell_filenames(enabled=True)

Translate the filenames of notebook cells in coverage information.

If enabled, ipytest.cov will translate the temporary file names generated by ipykernel (e.g, ipykernel_24768/3920661193.py) to their cell names (e.g., In[6]).

Warning: this is an experimental feature and not subject to any stability guarantees.

<!-- minidoc -->

Development

Setup a Python 3.10 virtual environment and install the requirements via

pip install -r requirements-dev.txt
pip install -e .

To execute the unit tests of ipytest run

python x.py test
python x.py integration

Before committing, execute python x.py precommit to update the documentation, format the code, and run tests.

To create a new release execute:

python x.py release

Related packages

ipytest is designed to enable running tests within an interactive notebook session. There are also other packages that aim to use test full notebooks: these packages run the notebook and compare the output of cells to the output of previous runs. These packages include:

While PyTest itself is generally supported, support for PyTest plugins depends very much on the plugin. The following plugins are known to not work:

See ipytest.cov on how to use ipytest with pytest-cov.

Please create an issue, if I missed a packaged or mischaracterized any package.

License

The MIT License (MIT)
Copyright (c) 2015 - 2024 Christopher Prohm

Permission is hereby granted, free of charge, to any person obtaining a
copy of this software and associated documentation files (the "Software"),
to deal in the Software without restriction, including without limitation
the rights to use, copy, modify, merge, publish, distribute, sublicense,
and/or sell copies of the Software, and to permit persons to whom the
Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
DEALINGS IN THE SOFTWARE.