Edit on GitHub

IJulia

WebIO Not Detected

For troubleshooting information, visit the WebIO Not Detected section of the documentation.

JupyterLab Considerations

JupyterLab presents a few issues that are different that than those associated with the classic notebook. JupyterLab extensions are bundled together with the core application code (so that all of the JavaScript associated with JupyterLab is served as a single file).

This can sometimes present issues if there are multiple versions of JupyterLab installed on a system (e.g. for multiple virtual environments or if installed using Conda and Pip) because these bundles are separate for each JupyterLab installation.

Note

For example, if you installed JupyterLab via pip install --user jupyterlab and pip install jupyterlab in a virtualenv at ~/venv, then there will be two bundles (at ~/.local/share/jupyter/lab/ and ~/env/share/jupyter/lab/ respectively, on Linux).

This presents issues when the JupyterLab that is launched is not the JupyterLab that WebIO was installed for (for example, the labextension was installed for the Conda version of JupyterLab but you're launching the version that is installed in a Python virtual environment).

Fixing Things Up

If you suspect that WebIO isn't installed for a specific version of JupyterLab, you can check with the jupyter labextension list command. First, make sure that you're using the version of JupyterLab that you mean to be using (this usually entails activating the correct Conda environment or Python virtual environment).

$ which jupyter
/home/user/.local/bin/jupyter

Then, make sure that the WebIO extension is installed. The output should look something like this (though the specific versions and paths will likely be different).

$ jupyter labextension list
JupyterLab v0.35.5
Known labextensions:
   app dir: /home/user/.julia/conda/3/share/jupyter/lab
        @webio/jupyter-lab-provider v0.8.3  enabled  OK*

   local extensions:
        @webio/jupyter-lab-provider: /home/user/.julia/dev/WebIO/packages/jupyter-lab-provider

   linked packages:
        @webio/webio: /home/user/.julia/dev/WebIO/packages/webio

If the WebIO extension isn't present, then simply fire up Julia and install it. This can either be done by running Pkg.build("WebIO") or by running

using WebIO, IJulia
WebIO.install_jupyter_labextension()

API Reference

WebIO.find_jupyter_cmd — Function
find_jupyter_cmd([; conda=false])

Find the most likely candidate for the jupyter executable. This will locate jupyter by searching the PATH environment variable and, if not found, tries to return Conda.jl's jupyter. If both of these approaches fail, an error is thrown.

source
WebIO.install_jupyter_labextension — Function
install_jupyter_labextension([jupyter]; condajl=false)

Install the Jupyter Lab extension for WebIO using the specified jupyter executable. The executable defaults to the first one found in the PATH or Jupyter installed via IJulia/Conda.jl. To force using Conda.jl's jupyter, specify the force_conda_jupyter=true keyword; this might be necessary if you launch Jupyter via IJulia in the Julia REPL.

The IJulia provider documentation provides some more information (and caveats) about the relationship between Jupyter Lab and WebIO.

source
WebIO.install_jupyter_nbextension — Function
install_jupyter_nbextension([jupyter])

Install the Jupyter Notebook extension (nbextension) for WebIO. This copies the nbextension code to the appropriate place and writes the appropriate configuration files.

source
WebIO.install_jupyter_serverextension — Function
install_notebook_config()

Install necessary configuration for the jlstaticserve notebook (server) extension. This function only configures the notebook extension, not the browser nbextension or Jupyter Lab extension.

  • Adds the path to ./deps to Python's sys.path so that we can load the jlstaticserve.py extension. This is done in jupyter_notebook_config.py because there's no way to add to sys.path from the JSON config file.
  • Adds jlstaticserve to the list of extensions loaded in the notebook server. This is done in jupyter_notebook_config.json because that file has higher precedence when both the .py and .json files exist (IPyWidgets, for example, writes to the JSON file, so if we only wrote to the .py file, that directive would take precedence and the jlstaticserve extension would not be loaded).
source