Making Pytest Scenarios Easy and Scalable

pytest-scenario-files is a pytest plugin that runs unit test scenarios using data loaded from files.

pytest has a feature called parameterization that allows you to run the same test function repeatedly using different inputs to test multiple scenarios. However, managing the test data for parameterization can be a problem. Sometimes the input data is very large or there are many test cases, so that it is impractical to put all of the data into the source code of the test.

This plug-in loads the data for the test scenarios from separate data files that are automatically matched up against the test functions. Each function can have one or more data files associated with it, and each file can contain multiple scenarios. The data files can be in JSON or YAML format.

An additional issue with the basic pytest parameterization API is how the user must provide the parameters. First all of the test case fixture names in a list, followed by a list of lists with the values the fixtures will take on, and then an optional list of test case id's. Since the labels, values, and test case id's are in separate lists it can be difficult to keep track of which fixture corresponds to which value if you have many of them, and also which group of values corresponds to which test id. The file structure uses a dict to keep the test case id's, fixture names, and data values together in a way that is easier on the human brain.

• Loads data for scenarios from files into fixtures
• Multiple scenario data sets may be in one file
• There may be multiple data files for each test
• Fixtures may refer to fixtures in other files
• Can specify indirect parameterization
• Intuitive and sane data file structure

This package is a plug-in for pytest and works with Python 3.9 and up.

• Tested with pytest version 7.4.x, should work with any version 6.2.5 or higher
• Tested with CPython 3.9-3.12 and PyPy 3.9-3.10

While this code currently has a classifier of "Development Status :: 4 - Beta", it is solid and well-tested. I will likely promote it to "Development Status :: 5 - Production/Stable" after a little more real-world usage.

You can install pytest-scenario-files from PyPI by using pip :

$ pip install pytest-scenario-files

To use this plugin you need to make only two changes to your project:

1. Install the plug-in
2. Create the data file(s) with the proper names and formats

You can then access the data from the files via the fixtures defined in those files. The most common usage is to manage test case inputs and expected results. This allows the developer to change and add test cases without making changes to the test code.

Just as with pytest's basic parameterization, the test function must have all of the fixtures in its parameter list. Otherwise, an exception will be raised.

The unit tests for this package are good examples of possible ways to use this package. Look in the files in the tests/ directory and the corresponding files in the tests/pytester_example_files directory.

This package is also designed to be transparent for non-parametric usage. If there are no data files associated with a particular test the fixtures will not be parameterized and everything will work as though the plug-in was not present.

A data file will be loaded if it matches all of the following criteria:

1. The filename starts with data_, followed by the name of the test function with the prefix test_ removed. The remainder of the filename may be any value, and is usually used to identify the tests contained in the file.
2. The filename must end in .json, .yaml, or .yml.
3. The file is contained in a folder at or below the file that contains the test.

For example, for a test function

test_foo(...)

the files

data_foo_part_1.json
data_foo_part_2.yaml
subfolder/data_foo.yaml

would all be loaded.

Caution: Beware of situations where the name of one test is an extended version of another. E.g., if you have two tests named test_foo() and test_foo_bar(), a data file with the name data_foo_bar.yaml will be loaded for both tests. To prevent this, split the two test functions into two separate files in two different directories or change the name of one of the test functions. See test_load_file_extended_name.py and test_load_separate_subdirs.py in the unit test files for this package for concrete examples of what might happen and how to avoid it.

Each data file may contain one or more sets of test data, in either yaml or json format. The top level is a dict whose keys are the test id's. Each test id is a dict whose keys are fixture names and whose values are the test data. The test data may be anything, including container types such as lists or dicts. An example input file data_foo_bar.yaml might contain:

test1:
  input_data_1: 17
  input_data_2: 3
  expected_result: 51

test2:
  input_data_1:
    - abc
  input_data_2: 3
  expected_result:
    - abc
    - abc
    - abc

This would parameterize into two test cases labeled test1 and test2, each with three fixtures, input_data_1, input_data_2, and expected_result.

If the same test case id is present in two different files, the fixtures from the two files will be merged as long as a fixture with the same name is not defined more than once for any particular test case id. For example, for a test function named test_foo() with two data files:

File data_foo_1.yaml;

test_case_one:
  fixture_one: 17

File data_foo_2.yaml;

test_case_one:
  fixture_two: 170

The function will be passed two fixtures fixture_one=17 and fixture_two=170 for a test case with id=test_case_one.

However, if the fixture names are the same there will be a conflict and the code that merges the test cases will raise an exception.

An additional powerful feature is the ability to load the value for a fixture from another data file. You can have fixture data loaded from another file by setting the fixture value to a specially formatted string. It must be prefixed with two underscores and be of the format:

__<Filename>:<test case id>:<fixture name>

For instance, a data file data_other_check_3.yaml might reference the data file data_foo_2.yaml from the previous section:

check_functionality:
  input_data_1: 42
  other_data: __data_foo_2.yaml:test_case_one:fixture_two

This would result in two fixture values being sent into the test function, input_data_1 = 42 and other_data = 170, for a test case with id = check_functionality. (Note that there is nothing preventing an infinite self-referential loop although that is something that should be avoided).

Pytest has a feature called indirect parameterization, where the parameter value is passed to a fixture function, and the return value of the fixture function is then passed downstream. You can specify that a fixture should be marked for indirect parameterization by appending the suffix _indirect to the fixture name in the data file. If the data file contains:

test_case_1:
  variable_A: 51
  variable_B_indirect: 3

test_case_2:
  variable_A: 85
  variable_B_indirect: 5

the corresponding test code would be:

@pytest.fixture
def variable_B(request):
    return request.param * 17


def test_func(variable_A, variable_B):
    assert variable_A == variable_B

The values for fixture variable_A would be passed directly to test_func(), but the values for variable_B_indirect would be passed to the variable_B() function and the return value would be passed in as the variable_B parameter to test_func().

If you encounter any problems, please file an issue including a detailed description and (if possible) an example of the problem.

Since this project is a pytest plug-in, it really does require test-driven development. If you want to contribute a bug fix or new feature, please first create a test case that demonstrates what your new code is supposed to do. Note that you need to set things up using the pytester fixture, rather than testing directly.

This project uses hatch for its environments and build system, as well as pre-commit, ruff, and mdformat for formatting and linting. Before you send in a pull request, please:

• Set up pre-commit and use it to run ruff and mdformat with the settings included in the pyproject.toml and .pre-commit-config.yaml files
• Run tests using the command hatch run test:test, which will run all of the tests against CPython 3.9-3.12 and PyPy 3.9-3.10
• Check test coverage with hatch run cov

Distributed under the terms of the MIT license, pytest-scenario-files is free and open source software.

This plug-in was originally named pytest-parameterize-from-files. It was inspired by the pytest plug-ins pytest-datadir, pytest-datafixtures, and pytest-xpara. I also later found the non-plug-in package parameterize-from-file. To avoid confusion and provide a more descriptive title, I renamed this project to pytest-scenario-files.

• I wanted to load data from files without having to write any additional code. However, pytest-datadir and pytest-datafixtures required code in the test or fixtures specifically to read in the file.
• I liked the way that pytest-xpara loaded data into a fixture, but didn't like that it would only work with one file and that I had to specify the file on the command line.
• After I wrote much of this project I found the package parameterize-from-files which has a similar name. It's a powerful and capable tool, but it's not to my taste as I think it's trying too hard.
  • It requires a decorator per test function, with potentially complex syntax inside the decorator's arguments.
  • It lets the user place code snippets into the data files which will be a maintenance problem down the road. It's cleaner to take advantage of Pytest's indirect parameterization feature instead.
  • Having to import the package in every test file and decorate each function increases the complexity of the test code.

In general I wanted the data file handling to be scalable:

• If you have 50 unit tests you don't have to specify all 50 files to load in code or on the command line.
• You can reference data from other files to keep duplication low.

This pytest plugin was developed using a skeleton generated by cookiecutter along with the cookiecutter-pytest-plugin template, then extensively modified to bring it up to modern standards.