---
author: "Marc Wouts"
date: "June 16, 2018"
jupyter:
  kernelspec:
    display_name: Python
    language: python
    name: python3
---

(examples/custom_formats)=
# Custom Notebook Formats

You can designate additional file types to be converted to Notebooks, and then executed / parsed in the same manner as regular Notebooks, by adding the following configuration to your `conf.py`:

```python
nb_custom_formats = {
  ".mysuffix": "mylibrary.converter_function"
}
```

- The string should be a Python function that will be loaded by `import mylibrary.converter_function`
- The function should take a file's contents (as a `str`) and return an [nbformat.NotebookNode](nbformat:api)

If the function takes additional keyword arguments, then you can specify these as dictionary in a second argument.
For example this is what the default conversion would look like:

```python
nbsphinx_custom_formats = {
    '.ipynb': ['nbformat.reads', {'as_version': 4}],
}
```

:::{important}

By default, Markdown cells in the Notebook will be parsed using the same MyST parser configuration as for other Markdown files ([see available configuration options](start/config-options)).

But, if this is incompatible with your file format, then you can specify for the Markdown to be parsed as **strictly CommonMark**, using a third argument:

```python
nbsphinx_custom_formats = {
    '.ipynb': ['nbformat.reads', {'as_version': 4}, True],
}
```

:::

Finally, for text-based formats, MyST-NB also searches for an optional `source_map` key in the output Notebook's metadata.
This key should be a list mapping each cell to the starting line number in the original source file, for example for a notebook with three cells:

```json
{
  "metadata": {
    "source_map": [10, 21, 53]
  }
}
```

This mapping allows for "true" error reporting, as described in [](start/error-reporting).

## Using Jupytext

A common conversion tool is [jupytext](https://github.com/mwouts/jupytext), which has been used to convert this very `.Rmd` file to a notebook!

The configuration looks like:

```python
nb_custom_formats = {
  ".Rmd": ["jupytext.reads", {"fmt": "Rmd"}]
}
```

:::{important}
For full compatibility with `myst-nb`, `jupytext>=1.8.0` should be used.
:::

For example:

```
\```{python echo=TRUE}
import pandas as pd
series = pd.Series({'A':1, 'B':3, 'C':2})
pd.DataFrame({"Columne A": series})
\```
```

```{python echo=TRUE}
import pandas as pd
series = pd.Series({'A':1, 'B':3, 'C':2})
pd.DataFrame({"Columne A": series})
```

```
\```{python bar_plot, echo=FALSE, fig.height=5, fig.width=8}
series.plot(kind='bar', title='Sample plot')
\```
```

```{python bar_plot, echo=FALSE, fig.height=5, fig.width=8}
series.plot(kind='bar', title='Sample plot')
```

## Acknowledgements

Thanks to [nbsphinx](https://nbsphinx.readthedocs.io), for providing the initial implementation which this was built on.