Skip to content

Latest commit

 

History

History
151 lines (107 loc) · 4.98 KB

README.md

File metadata and controls

151 lines (107 loc) · 4.98 KB
Raw

nbtest

nbtest is Elastic's Search Labs testing and validation tool for Python notebooks.

Installation

pip install elastic-nbtest

Usage

nbtest my-notebook.ipynb another-notebook.ipynb ...

How it works

nbtest runs all the code cells in the notebook in order from top to bottom, and reports two error situations:

  • If any code cells raise an unexpected exception
  • If any code cells that have output saved in the notebook generate a different output (not counting especially designated sections, as described in the "Configuration" section below)

Something to keep in mind when designing notebooks that are testable is that for any operations that are asynchronous it is necessary to add code that blocks until these operations complete, so that the entire notebook can execute in batch mode without errors.

Configuration

nbtest looks for a configuration file named .nbtest.yml in the same directory as the target notebook. If the configuration file is found, it is imported and applied for all the notebooks in that directory.

There is currently one supported configuration variable, called masks. This variable can be set to a list of regular expresssions that capture details in the output of code cells that vary from one run to the next and should be masked when comparing the previously stored output against output from the current run.

Here is an example .nbtest.yml file:

masks:
- "'name': '[^']+'"
- "'cluster_name': '[^']+'"
- "'cluster_uuid': '[^']+'"
- "'build_flavor': '[^']+'"
- '[0-9]+\.[0-9]+\.[0-9]+'
- "'build_hash': '[^']+'"
- "'build_date': '[^']+'"

Handling of the getpass function

The nbtest script installs an alternative version of the getpass() function that looks for requested values in environment variables instead of starting an interactive prompt.

Consider the following example, which is used in many Elastic notebooks:

CLOUD_ID = getpass("Elastic Cloud ID:")
ELASTIC_API_KEY = getpass("Elastic Api Key:")

The getpass() function used by nbtest takes the prompt given as an argument and converts it to an environment variable name with the following rules:

  • Spaces are converted to underscores
  • Non-alphanumeric characters are removed
  • Letters are uppercased

In the above example, the variables that will be used to source these prompts are ELASTIC_CLOUD_ID and ELASTIC_API_KEY.

As a convenience, nbtest imports all variables defined in a .env file if found. The --env-file option can be used to provide an alternative environment file.

Set up and tear down procedures

Sometimes it is necessary to perform "set up" and/or "tear down" operations before and after a notebook runs. nbtest will look for notebooks with special names designated as set up or tear down and execute those notebooks to perform any necessary actions.

Set up notebooks

nbtest will look for the following notebooks names and execute any that are found before running the target notebook:

  • _nbtest_setup.ipynb
  • _nbtest_setup.[notebook-name].ipynb

The first one can be used for general set up logic that applies to all the notebooks in the directory. The NBTEST["notebook"] expression can be used inside this notebook to obtain the name of the notebook under test.

The second one should be used for set up actions that are specific to one notebook.

A global setup notebook can also be provided in the --setup-notebook command line argument. This notebook is executed once, before any tests.

Tear down notebooks

nbtest will look for the following notebooks names and execute any that are found after running the target notebook, regardless of the testing having succeeded or failed:

  • _nbtest_teardown.[notebook-name].ipynb
  • _nbtest_teardown.ipynb

These notebooks are inteded for cleanup that needs to happen after a text, for example to delete indexes that were created. As in the set up case, NBTEST["notebook"] is set to the notebook that was tested.

A global teardown notebook can also be provided in the --teardown-notebook command line argument. This notebook is executed once, after all the tests.

Mocking and/or Monkey-Patching

For specific cases in which it is necessary for tests to alter the behavior of functions or methods, nbtest provides the --mocks option. The argument to this option is a Python module name, exactly as it would be entered in an import statement. The module will be imported in the context of the notebook and any defined setup and teardown procedures.

The module passed to the --mocks option can apply any patching techniques as needed by the tests. The following example patches the Elasticsearch class to always connect to a locally hosted instance without authentication, regardless of the connection arguments that are passed:

import elasticsearch

orig_es_init = elasticsearch.Elasticsearch.__init__


def patched_es_init(self, *args, **kwargs):
    return orig_es_init(self, 'http://localhost:9200')


elasticsearch.Elasticsearch.__init__ = patched_es_init