Skip to main content

Pytest plugin for functional testing of data analysis pipelines

Project description

pytest-pipeline is a Python3-compatible pytest plugin for functional testing of data analysis pipelines. They are usually long-running scripts or executables with multiple input and/or output files + directories.

It is meant for end-to-end testing where you test for conditions before the pipeline run and after the pipeline runs (output files, checksums, etc.).


pip install pytest-pipeline


For our example, we will use a super simple pipeline that writes a file and prints to stdout:

#!/usr/bin/env python

from __future__ import print_function

if __name__ == "__main__":

    with open("result.txt", "w") as result:
    print("Result computed")

At this point it’s just a simple script, but it should be enough to illustrate the plugin. Also, if you want to follow along, save the above file as run_pipeline.

With the pipeline above, here’s how your test would look like with pytest_pipeline:

import os
import shutil
from pytest_pipeline import PipelineRun, PipelineTest, mark, utils

# one pipeline run is represented by one class that subclasses PipelineTest
class TestMyPipeline(PipelineTest):

    # define the pipeline execution via PipelineRun objects
    run = PipelineRun(
        # the actual command to start your pipeline

    # before_run-marked functions will be run before the pipeline is executed
    def test_prep_executable(self):
        # copy the executable to the run directory
        shutil.copy2("/path/to/run_pipeline", "run_pipeline")
        # testing if the file is executable
        assert os.access("run_pipeline", os.X_OK)

    # after_run-marked tests will only be run after pipeline execution is finished
    def test_result_md5(self):
        assert utils.file_md5sum("result.txt") == "50a2fabfdd276f573ff97ace8b11c5f4"

    # ordering for all tests annotated by after_run can be set manually
    # here we want to test the exit code first after the run is finished
    def test_exit_code(self):
        assert == 0

    # we can also check the stdout that we capture as well
    def test_stdout(self):
        assert open("run.stdout", "r").read().strip() == "Result computed"

If the test above is saved as, you can then run the test by executing py.test -v You should see that four tests were executed and all four passed.

What just happened?

You just executed your first pipeline test. The plugin itself gives you:

  • Test directory creation (one class gets one directory). By default, testdirectories are all created in the /tmp/pipeline_test directory. You can tweak this location by supplying the --base-pipeline-dir command line flag.

  • Automatic execution of the pipeline. No need to import subprocess, just define the command via the PipelineRun object. We optionally captured the standard output to a file called run.stdout as well. For long running pipelines, you can also supply a timeout argument which limits how long the pipeline process can run.

  • Test ordering. Pipelines by definition are simply series of commands executed subsequently. The plugin allows you to also order your tests accordingly via the before_run and after_run decorators. In the code above, we first test for the exit code before testing the output files. Using the command line flag --xfail-pipeline, if the first test after the pipeline run fails then the rest will be marked as failed immediately.

And since this is a py.test plugin, test discovery and execution is done via py.test.

Getting + giving help

Please use the issue tracker to report bugs or feature requests. You can always fork and submit a pull request as well.




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

pytest-pipeline-0.1.0.tar.gz (16.9 kB view hashes)

Uploaded Source

Built Distribution

pytest_pipeline-0.1.0-py2.py3-none-any.whl (11.4 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