jupyterlite doesn't have print option

[NAbdolkhani]

In this new interface, the print option is gone and there is no way to save the notebook as pdf. Is there an alternative? or can't you just bring the print option back?

[bollwyvl]

just bring the print option back?

The print feature upstream relies on server-side nbconvert to generate the "dead pixels" HTML.

To get to something better we have a few options:

• run nbconvert in python in the browser
  • this would probably be the most faithful to the existing experience, and might support plugins, etc.
  • at present nbconvert cannot be installed directly, due to a missing tornado and pyzmq
    • we could potentially shim these... or provide a patched nbconvert
  • we've been trying to avoid requiring any heavy-duty wasm inside the server component
• re-implement some subset of nbconvert in JS
  • this would probably be the most performant
  • theoretically, we could pack all of the jinja templates from nbconvert and use nunjucks, but...
• work with more directly with in-app HTML
  • this would potentially be interesting, as it might handle a broader swathe of outputs (canvas, etc.)

[pierrepo]

Dear @bollwyvl I had a similar issue while trying to export a notebook as html. If I understand correctly your previous answer, this is also because nbconvert (or any substitute) is not (yet) available in jupyterlite. Am I correct?

[bollwyvl]

nbconvert (or any substitute) is not (yet) available in jupyterlite.

Yep.

[bollwyvl]

nbconvert (with mad hacks) is now (somewhat) available in (a soon-to-be-released) jupyterlite.

#756 (comment)

[sean-morris]

@bollwyvl I am hung up on the "somewhat" available part. From the link above, it appears that v0.1.0b12 includes the ability to print a notebook to a pdf. I working with this version.

The example from this gist is pretty intense. Is the implication that you need to run through the commands on the page referenced to create a context for converting and downloading a notebook to a PDF?

I have moved through the gist but ran into issues right at the end of the process. The HTMLExporter in nbconvert has been moved for one, that was not a big deal -- nbconvert.exporters.HTMLExporter -- but the second involves a problem with bleach not having the attribute ALLOWED_STYLES defined. The error suggests upgrading bleach to 5 or greater. But this did not work.

I am able to read in the notebook which is great -- I just need the last step!

Am I missing something here?

[bollwyvl]

I haven't revisited this in a while, and while certainly no upstreams have made this easier, I'm not particularly surprised the gist no longer works. PDF export is not really going to happen, natively, through any of the nbconvert mechanisms (pandoc/latex, headless browser, qt). Some of these _have_ been compiled to wasm, which indeed would be a lower-friction way to deliver them to users, as they are all very tricky to get/keep working on many platforms. To this end, I'm still of the mind that trying to do this _in_ a kernel (or even python-in-wasm-but-not-kernel) is not worth the effort, and a proper implementation should start with a lab extension that reimplements the stack on top of JS, with e.g. nunjucks. This would allow it to, for example, emulate "--execute" with any kernel, using only real contents and kernel APIs, and therefore work in a full jupyterlab/notebook(7).

[sean-morris]

Ok, I am with you. I will look into an extension. Thanks that helps a lot.

[westurner]

Is there any reason nbconvert cannot be compiled into a jupyterlite build (w/ emscripten)?

When I:

• open Intro.ipynb
  • Firefox w/ linked instances of Jupyter [1] and [2]
  • [1] https://jupyterlite.readthedocs.io/en/latest/_static/lab/?path=notebooks%2FIntro.ipynb
  • [2] https://jupyter.org/try-jupyter/lab/?path=notebooks%2FIntro.ipynb
• try File > Print and/or Ctrl P
• BUG: Print prints a 404 page
• It looks like the 404 is from the url path that the jupyterlite instance is hosted on, and not the notebook.
• ; it does not print the notebook
• nbconvert in WASM without pandoc could generate printable notebook HTML of the Inputs and Outputs
• Is there a reason we can't just print the notebook HTML as it is without nbconvert; by changing the @media print { ... } CSS to only display the current notebook tab at print time?

[westurner]

nbconvert (with mad hacks) is now (somewhat) available in (a soon-to-be-released) jupyterlite.

#756 (comment)
[...]

"Running nbconvert inside JupyterLite"
https://gist.github.com/bollwyvl/6b3cb4c46b1764c6d9ae1e5831f86d7a#file-nbconvert-in-jupyterlite-ipynb

Future Work

• explore packaging this technique for JupyterLite
  • by creating a custom JupyterLite addon, it would be possible to fire up a kernel and service the expected JupyterLab UI components
  • do a better shim of nbclient
    • this could use JupyterLite machinery to actually run kernels
  • further simplify the nbconvert API
    • nbclient could be an optional, or at least not catastrophically-failing dependency

BUG,404: JupyterLite: File > Print does not print nbconvert'ed notebooks