Getting syntax highlighting to work for custom cell magic

Hey there,

I have a built a package to compile and run C/C++ and FORTRAN sources in a jupyter notebook cell using custom cell magics.

The cell magics are %%CPP %%C and %%FORTRAN.

I can’t seem to get proper syntax highlighting for the cell magics in JupyterLab. This code enables syntax highlighting for ordinary Jupyter notebooks if I put it in ~/.jupyter/custom/custom.js

IPython.CodeCell.options_default.highlight_modes[‘magic_text/x-c++src’] = {‘reg’:[/^%%CPP/]} ;
IPython.CodeCell.options_default.highlight_modes[‘magic_text/x-csrc’] = {‘reg’:[/^%%C/]} ;
IPython.CodeCell.options_default.highlight_modes[‘magic_text/x-fortransrc’] = {‘reg’:[/^%%FORTRAN/]} ;‘kernel_ready.Kernel’, function(){
if (cell.cell_type == ‘code’){ cell.auto_highlight(); } }) ;

I’m trying to do the same thing for Jupyterlab. I used cookiecutter to create a template Javascript extension but I quickly found myself in the stratosphere of complexity and not sure what to do now to get syntax highlighting to work. Perhaps you could have a look at jupyterlab_compiler_magics/index.js at c2c4b77f33341824785c075fa3469c828a8c67fb · drtpotter/jupyterlab_compiler_magics · GitHub and let me know what needs to be there. Any suggestions would be helpful!

Kind regards,

Here’s an in-the-wild example:

And a defunct example:

1 Like

^ describes adding of mode itself.

For dynamic changes of mode depending on cell content, like the use of %%CPP magic, jupyterlab-lsp is one option. Or you could write an extension which re-uses the relevant bits of jupyterlab-lsp. Or we could make it into standalone extension within LSP to let users install it as it seems a common question.

Good suggestions, I’m still not sure how to proceed but it seems that this use case is pretty common and could go into a feature request. All that we need is a way to apply CodeCell’s syntax highlighting for any of its supported languages to any cell magic of our choosing.

So one option is to mirror Bigquery magic by julioyildo · Pull Request #553 · jupyter-lsp/jupyterlab-lsp · GitHub and send a pull request to jupyterlab-LSP and we could include this in the next release (here is a PR with some screenshots: Syntax highlighting by krassowski · Pull Request #319 · jupyter-lsp/jupyterlab-lsp · GitHub). You will need to find out names of CodeMirror modes (possibly here: CodeMirror: C-like mode) and of the Fortran mode. The regular expressions may be much simpler if the magics do not have any arguments.

The real treat is that once someone adds a spec for C/C++/Fortran language server you will also get much, much more coding assistance (advanced autocomplete, linting, etc) for free. See add Pyright langserver spec by yuntan · Pull Request #587 · jupyter-lsp/jupyterlab-lsp · GitHub for an example of a PR adding a spec.


Yes, I forgot to mention the LSP path, which is becoming increasingly viable. I’d personally also be happy to help in getting more (open source) language servers packaged on conda-forge, as that’s where we draw our test dependencies (if possible) for jupyter-lsp.

1 Like


1 Like

By the by, fortran-language-server is now up on conda-forge

1 Like