Question about packaging
hadim opened this issue · 4 comments
@fcollonval There is something I am not sure about: why do I keep pushing new versions to npm (https://www.npmjs.com/package/@hadim/jupyter-archive) if the lab extension is already embedded to both the pip and conda packages.
The only advantage I see is that people can discover it through JLab extension manager but they would still need to install the server extension anyway...
What do you think?
Yes, the good way of packaging extensions is not really solved unfortunately. The two following arguments come to my mind for why a npm package should be published:
- Ensuring that nobody else publish another extension with the same name - albeit this does not require keeping the package up-to-date
- Allow the frontend extension to be used for installation with non-default
JUPYTERLAB_DIR
The latter is actually used in my company to provide a customized jupyterlab without interfering with the original one taking advantage of that approach.
Ok, thanks for the comments.
I'll keep publishing NPM packages then.
I let this issue open for some time in case a Jupyter dev wants to comment.
The status with JupyterLab 3 has been clarified. But this is still advised to publish on npm in case somebody would be interested in using the frontend source code (unlikely though).
Now that the extension is ported on JupyterLab 3. This is clarified, for final user only the python package is required.