JupyterLab Bundle

Cookiecutter template to generate conda package of Jupyterlab with prepackaged extensions independent of a standard JupyterLab installation. The package created has two scripts:

project_slug-lab equivalent for the bundle to jupyter lab
project_slug-labextension equivalent for the bundle to jupyter labextension

Requirements

To generate a recipe, you will need to install cookiecutter.

Using conda:

conda install -c conda-forge cookiecutter

To generate the JupyterLab bundle, you will need conda package manager and conda-build package builder.

Using conda:

conda install -c conda-forge conda-build

Usage

Generate a new recipe from the cookiecutter template:
```
cookiecutter gh:measejm1/jupyterlab_delux
```
Adapt the template to bundle the extensions you want
1. Add the JupyterLab extension installation commands in recipe/bld.bat and recipe/build.sh They will have the following structure jupyter labextension install extension_name --no-build
2. Add the Jupyter server extensions needed by the JupyterLab extensions in the run section of recipe/meta.yaml
Generate the conda package
```
conda build -c conda-forge recipe
```

As displayed by the conda build command, this will create a conda package in your anaconda/conda-bld/noarch directory.

Comments:

You can specify JupyterLab extension versions to install using the npm notation extension_name@version.
You can specify JupyterLab and Jupyter server extensions versions using conda notation package =X.Y.Z or package >=X.Y.Z
On Windows, the build process may crash due to the folder name length limit. To overcome that problem, you can specify a custom folder to build conda package into:

conda build -c conda-forge --croot C:\Temp recipe

Implementation Notes

To have a clear separation between the standard jupyterlab and the version bundled with preinstalled extensions, we take advantage of the LabApp.app_dir configurable property (see the documentation).

That parameter is more easily specified through the environment variable JUPYTERLAB_DIR as you can see in the building scripts.

To launch easily this customized JupyterLab, two entry points are created:

project_slug-lab calling jupyterlab_project_slug.labapp:main equivalent for the bundle to jupyter lab
project_slug-labextension calling jupyterlab_project_slug.labextensionapp:main equivalent for the bundle to jupyter labextension

Limitations

This approach does not address the open question of how individual extensions should be distributed as conda packages and installed in an offline environment.

See jupyterlab/jupyterlab#2065 for discussion on that front.

License

This project is licensed under the terms of the BSD 3-Clause License.