convert-notebook.md

Bring your notebook and make it reproducible (Convert cells with pip commands automatically)

Let's say I have a notebook I was working on that uses pip in my cells and I want to make it reproducible.

1. Let's open the notebook called make-notebook-reproducible, provided in notebooks folder.

   

   NOTE: A warning from the jupyterlab-requirements extension will appear to tell users what should be used to handle dependencies.

   

2. Run cell with %horus check to check the status of your notebook:

   

   As you can see, initially there are errors reported because no Pipfile or Pipfile.lock exist for this notebook.

3. Run %horus convert, so that the extenstion can convert !pip install cells to commands that allow reproducibility.

   

4. Run the converted cells to add the requirements to your notebook.

   NOTE: If you want to edit some dependencies, you can simply add them again with your specific requirements (%horus requirement --add).

5. Run cell with %horus check to check the status of your notebook.

6. Run %horus lock to lock dependencies using Thoth resolution engine.

   If you are interested in a specific recommendation from Thoth, add --recommendation-type <recommendation-type>, default is latest ("latest", "stable", "performance", "security").

   By default, Thoth will discover the runtime environment you are running on. If you want to receive a recommendation for a specific runtime environment, you can use the following flags:

   • --os-name
   • --os-version
   • --python-version

   You can also use %horus lock --pipenv if you want to lock dependencies with Pipenv resolution engine.

7. Run cell with %horus check to check the status of your notebook or %horus show to show the content of your notebook.

   If you want to show a specific part of your dependencies information stored in the notebook metadata, you can use the following flags:

   • --pipfile
   • --pipfile-lock
   • --thoth-config (only if Thoth resolution engine was used)

Other cases

You can consider another use case for managing dependencies:

• start working on a new notebook

• create dependencies for your existing notebook

• use a reproducible notebook

Next steps

Remove kernels