clonenotebooks 1.0.3

NBViewer extension and Jupyter notebook extension for cloning notebooks viewed in NBViewer to user's home directory.

This was built starting from Daniel Allan's notebook clone extension as a basis. It currently requires branch step7 of my fork of NBViewer in order to run.

Installation Instructions

The easiest and quickest way to get a version of this up and running to test out would be to use the Docker image and setup instructions included in the Docker subfolder of this repository.

Otherwise install branch step7 of my fork of NBViewer using the same installation instructions as for the Jupyter master branch, in particular the setup instructions for using NBViewer as a JupyterHub service. Then download this repository, and in the folder run (the dot is important, it means "present working directory"):

pip install .

if you don't want to make changes to the source code, otherwise

pip install -e .

This installs the corresponding Python package, but we still need to tell Jupyter to use the appropriate sub-module as a notebook server extension. To do this, run:

jupyter serverextension enable clonenotebooks.cloners --sys-prefix

Make sure you have nbviewer installed as a Python package, via Pip or Conda or whatever else, on order for this to work.

Then copy the templates folder to your preferred location, and add to the command for NBViewer in your jupyterhub_config.py file

--template-path=/your/preferred/location

Then make sure to include an nbviewer_config.py file (this is a major why for now using my fork of NBViewer is necessary) with the lines:

c.NBViewer.handler_settings    = {'clone_notebooks' : True}

c.NBViewer.local_handler       = "clonenotebooks.renderers.LocalRenderingHandler"
c.NBViewer.url_handler         = "clonenotebooks.renderers.URLRenderingHandler"
c.NBViewer.github_blob_handler = "clonenotebooks.renderers.GitHubBlobRenderingHandler"
c.NBViewer.github_tree_handler = "clonenotebooks.renderers.GitHubTreeRenderingHandler"
c.NBViewer.gist_handler        = "clonenotebooks.renderers.GistRenderingHandler"
c.NBViewer.user_gists_handler  = "clonenotebooks.renderers.UserGistsRenderingHandler"

By default notebooks are cloned into the root directory as determined by the value of c.Spawner.notebook_dir in jupyterhub_config.py. For cloning notebooks into another directory, the line c.NBViewer.handler_settings in nbviewer_config.py needs to be modified as follows:

c.NBViewer.handler_settings    = {'clone_notebooks' : True, 'clone_to_directory' : <desired directory here>}

One can include {username} as a stand-in for the JupyterHub user's name, analogous to the settings for c.Spawner.notebook_dir and c.Spawner.default_url in jupyterhub_config.py. For example the following is valid:

c.NBViewer.handler_settings    = {'clone_notebooks' : True, 'clone_to_directory' : '/users/{username[0]}/{username}'}

will cause notebooks to be cloned into /jupyter/users/f/foo for user foo and /jupyter/users/b/bar for user bar if the value of c.Spawner.notebook_dir is '/jupyter', and will cause notebooks to be cloned into /users/f/foo for user foo and /users/b/bar for user bar if the value of c.Spawner.notebook_dir is '/'. In particular, the destination where notebooks is cloned will always be relative to the contents manager's root directory (which will usually equal the value of c.Spawner.notebook_dir).

An example copy of nbviewer_config.py is also included in this repository, in the Docker subfolder. Ideally this should have everything configured, but admittedly these setup instructions are more vague than they could be and might not have suggested an important step.

I recommend comparing with the Dockerfiles in the Docker subfolder of this repository for an example setup if any difficulties arise. Please give any and all feedback about any ways in which the documentation could be improved, since it will be much appreciated. Here is a link to the issues page for requests for improved documentation and/or general feedback.

Kernelspec Cloning

For notebooks from almost any source (local, Gist, URL), clonenotebooks checks for a "local" kernelspec (kernel.json) file located in the same directory as the notebook being cloned, with the assumption that this kernelspec can be used at the clone destination to load the environment needed to run the environment. If it finds one, the kernelspec is installed in addition to the notebook being cloned. The name given to the kernelspec (i.e. the name of the corresponding directory in <environment_path>/share/jupyter/kernels) is by default the name of the enclosing directory. ("Kernel name" as used here should not be confused with the display_name attribute of the kernel.json, which is what is visible to the end-user and does not need to be unique.) (In the case of notebooks from URLs or Gist, "enclosing directory" refers to the "base name" of the URL "path" excluding the filename, e.g. test in https://example.com/test/notebook.ipynb.) If a kernelspec with the same name is already found, the previous one is overwritten. In particular, if you update the kernelspec (kernel.json) file in the directory and then clone another notebook from that directory, the updated kernelspec will replace the previous one.

The behavior is somewhat different for notebooks sourced from GitHub. First, clonenotebooks checks for a "global" kernelspec located at the repository's root directory. Second, it checks for a "local" kernelspec in the notebook's directory as in the cases above. If a "local" kernelspec is found, regardless of whether a "global" kernelspec is also found, the "local" kernelspec will be installed, and under the name <repo_name>-<branch_name>-<enclosing_directory> (so as to avoid name conflicts with any "global" kernelspecs). If a "global" kernelspec is found, but no "local" kernelspec is found, then the "global" kernelspec is installed under the name <repo_name>-<branch_name>. (For notebooks located in the repository root, any kernel.json located also in the repository root will be treated as a "global" kernelspec, and thus installed under <repo_name>-<branch_name>, even though in this case the kernel.json is technically also a "local" kernelspec.)

"Global" kernelspecs could probably be implemented for other notebook sources if there is demand for the feature and if sensible defaults can be chosen for the locations to check for the "global" kernelspec.

Features

This extension can clone notebooks served from:

• URL
• GitHub tree view
• GitHub individual file view
• individual GitHub Gists
• Gist's from a user's page
• local files from a directory
• individual local files

Attributions

Example notebooks included in multiple-container Dockerfile demo are from the Jupyter gallery of interesting notebooks. Credit for them belongs to their respective authors (Filipa Rodrigues, Jason Chin, Shashi Gowda).