ipynb-py-convert-databricks

Function for un- and re- doing the conversion from .ipynb ipython notebook to .py python file which Databricks performs every time it exports a notebook. Building on code from kiwi0fruit/ipynb-py-convert

Why?

Databricks exports .ipynb files (e.g. when version controlling) as .py, which is pretty handy as it's easier to deal with a straight up python script file than a JSON formatted IPython notebook. However, it would be nice to perform the reverse operation, e.g. when we download a databricks notebook and want to run it in a local anaconda notebook.

How?

Conversion is possible both ways; from the .py version made by databricks to an IPython notebook:

convert_databricks_nb('databricks_nb.py', 'databricks_nb.ipynb')

And from an IPython notebook to a .py file (i.e. a function presumably similar to the one Databricks itself runs before exporting notebooks):

convert_databricks_nb('databricks_nb.ipynb','databricks_nb.py')

Example

Let's say we write the following notebook in databricks:

On, exporting, databricks will convert it to a .py file which looks like this:

Although this is a perfectly functioning .py script, it's lost the IPython notebook formatting of the original databricks notebook.

By running:

convert_databricks_nb('example_databricks_notebook.py', 'example_databricks_notebook_conv.ipynb')

we convert this file to an IPython notebook which looks like this:

Notice that all magic commands (except for markdown) remain commented out.