Archive note: The functionality this repository provided is now integrated into the workflow process run over individual tutorial repositories. See the build.yml workflow in a given tutorial repository and the action-- repositories in this organization.

nbcollection

Tools for building collections of Jupyter notebooks into web pages for public consumption.

This project serves as a thin wrapper around nbconvert to enable converting and executing directories or directory structures full of Jupyter notebooks to static HTML pages.

License

nbcollection is free software made available under the MIT License. For details see the LICENSE file.

---

Example usage

Default behavior:

Converting a directory structure of specific notebook files

Imagine we have a directory containing Jupyter notebook files and some other sub-directories that also contain notebook files, such as:

my_notebooks
├── notebook1.ipynb
└── sub_path1
    ├── notebook2.ipynb
    └── notebook3.ipynb

From the top level, we could use nbcollection to execute and convert all of these notebook files to HTML by running:

nbcollection convert my_notebooks

With no options specified, this will create a directory within the specified path, my_notebooks/_build, to store the executed notebooks and the converted HTML pages:

my_notebooks
└── _build
    ├── notebook1.ipynb
    ├── notebook1.html
    └── sub_path1
        ├── notebook2.ipynb
        ├── notebook3.ipynb
        ├── notebook2.html
        └── notebook3.html

If you are only interested in executing the notebooks, you can instead use the execute command:

nbcollection execute my_notebooks

which still creates a new _build path but now only contains the executed notebook files:

my_notebooks
└── _build
    ├── notebook1.ipynb
    └── sub_path1
        ├── notebook2.ipynb
        └── notebook3.ipynb

Converting a list of specific notebook files

Instead of running on a full directory, it is also possible to convert or execute single notebook files (but you should probably use jupyter nbconvert directly), or lists of notebook files. For example, to convert a set of specific notebook files within the above example directory layout:

nbcollection convert my_notebooks/notebook1.ipynb my_notebooks/sub_path1/notebook2.ipynb

Because these files could in principle be in two completely separate paths, the build products here will instead be written to the current working directory by default (but see the command option --build-path below to customize). So, the above command would result in:

_build
├── notebook1.ipynb
└── sub_path1
    └── notebook2.ipynb

Command options:

Several options are available to modify the default behavior of the nbcollection commands.

Customizing the build path

As outlined above, the default locations for storing the executed notebooks or converted HTML pages is either in a parallel directory structure contained within a _build directory created at the top-level of the specified path, or in a _build path in the current working directory (if a list of notebook files are specified). However, the build path can be overridden and specified explicitly by specifying the --build-path command line flag. For example, with the notebook directory structure illustrated in the above examples, we could instead specify the build path with:

nbcollection convert my_notebooks --build-path=/new/path/my_build

With this option specified, the executed notebook files and converted HTML notebooks would be placed under /new/path/my_build instead.

Flattening the built file structure

If your notebook files are spread throughout a nested directory structure, you may want to place all of the converted notebook files in a single path rather than reproduce the relative path structure of your content. To enable this, use the --flatten boolean flag. For example, if your content has the following path structure:

my_notebooks
├── notebook1.ipynb
└── sub_path1
    ├── notebook2.ipynb
    └── notebook3.ipynb

You can convert all of the notebooks to a single build path with:

nbcollection convert my_notebooks --flatten

This will result in:

my_notebooks
└── _build
    ├── notebook1.ipynb
    ├── notebook2.ipynb
    ├── notebook3.ipynb
    ├── notebook1.html
    ├── notebook2.html
    └── notebook3.html

This command also works in conjunction with --build-path if you want to, e.g., convert a list of individual notebook files and have the build products end up in the same root path.

Specifying a custom template file

nbconvert allows specifying custom jinja2 template files for exporting notebook files to HTML. We support this through the --template command-line flag, which allows specifying a path to a jinja2 template file. For example:

nbcollection convert my_notebooks --template-file=templates/custom.tpl

Extracting figures and other preprocessors

You can enable additional preprocessors for the HTML exporter by passing one or more preprocessor names to the --preprocessors option. A useful preprocessor is ExtractOutputPreprocessor, which extracts figures from the HTML into separate files for better page loading performance:

nbcollection convert my_notebooks --preprocessors=nbconvert.preprocessors.ExtractOutputPreprocessor

Only execute the notebooks

Though the primary utility of nbcollection is to enable converting a collection of notebook files to static HTML pages, you can also use the nbcollection execute command to instead only execute a collection of notebooks. This command is used the same way as nbcollection convert, but also enable executing the notebook files in place as a way to test the notebooks. To execute a collection of notebooks in-place (i.e., this will not create a _build path with the executed notebooks):

nbcollection execute my_notebooks --inplace