Help with Jupyter/IPython notebook kernel for Elm

[Austin Bingham]

Hi everyone,

I started hacking a bit today on a Jupyter notebook kernel for elm. You can see it here:

    https://github.com/abingham/jupyter-elm-kernel

It doesn't quite work yet, though, and I think I need help from someone who knows javascript/requirejs/web stuff a bit better than me. The proximal problem I'm seeing is that when jupyter tries to run the elm-make-generated javascript, it thinks that 'Elm' is undefined during (I think) some part of the AMD machinery. Jupyter show this as the output for the cell:

    Javascript error adding output!

    ReferenceError: Elm is not defined

    See your browser Javascript console for more details.

As far as I can see, Elm *should* be defined, and certainly the generated code looks like any other Elm output I've looked at. So I'm a bit stumped.

My approach to the kernel is currently very simple. The kernel is implemented in Python, and it receives a blob of Elm source code. I dump this to a temp file, use a subprocess to run "elm-make" to make the output, read the output, and ship it back to jupyter. As far as I can see, all of that is working properly. I run into problems when jupyter tries to execute the stuff I return. This design may or may not be optimal in the long run, but I want to get the plumbing working first.

So if someone feels up to the challenge, I'd love any help I could get. This seems like it should be pretty straightforward, but perhaps I'm being naive and/or missing something obvious.

Of course, I'm also happy to discuss other aspects of the kernel (e.g. design, compilation technique, etc.), but my priority is to just get something into an output cell.

[Austin Bingham]

FWIW, I see that when the generated javascript is passed back to the Jupyter client, the client calls eval() on the javascript. This seems to work OK, insofar as it doesn't throw an exception. However - and this is where I'm a bit out of my depth - none of the bindings made in the javascript seem to be available in the calling context. Maybe this is the nature of eval(), but I'm not sure. The "ReferenceError" I reported initially seems to actually come from when I try to use the "Elm" object to embed the app in a DOM object.

[Berry Groenendijk]

Hi Austin,

I like Jupyter/iPython a lot as a concept. I did not use it a lot though. And Elm seems to be like a natural fit.

My dream would be to have live aka. living documents where static text and live code are intermingled. Jupyter is a step in the right direction. But, you still need a server.

Just recently I found a thing called Klipse, see: http://blog.klipse.tech/javascript/2016/06/20/blog-javascript.html. It looks a lot like Jupyter (minus the document editing, but that should be trivial), but all the code is executed in the browser! It currently has support for clojure, ruby, javascript, python, scheme, es2017, jsx, brainfuck, c++ and Lua. Quite impressive. And if javascript works, them Elm should also work. Perhaps it is already possible... import elm.js... hmmm... I don't know.

Sorry, if this post isn't directly related to Jupyter and your challenge to integrate Elm in Jupyter.

Op maandag 27 maart 2017 20:10:33 UTC+2 schreef Austin Bingham:

[Noah Hall]

If you join the https://elmlang.slack.com and join #elm-dev, I can
help you out (eeue56)

> --
> You received this message because you are subscribed to the Google Groups
> "Elm Discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to elm-discuss...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Austin Bingham]

Klipse looks interesting, but I'm hoping to use Jupyter to make a presentation so I'm going to stick with that for now.

Regarding Jupyter needing a server: there are places that host notebooks, so the server requirement is only partly true. Your point is a good one, though. Sometime the complexity of jupyter can seem a bit unwieldy.

[Michael Lamb]

Are you still looking for a hand with this, Austin? 

I'm not an expert, or anything but I'm really interested in making a good Jupyter kernel for Elm to use with nteract. If you haven't seen it, check it out: https://nteract.io/
It mimics Jupyter as a desktop application, so you do not need a server.

[Austin Bingham]

The basic kernel is working at this point, and you can get it here: https://github.com/abingham/jupyter-elm-kernel

But there's still quite a bit to be done, and some of the design decisions I made might be short-sighted. Take it for a spin and let me know what you think. Any feedback, ideas, PRs, etc. would be great.

--
You received this message because you are subscribed to a topic in the Google Groups "Elm Discuss" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/elm-discuss/4VUQuYukGXM/unsubscribe.
To unsubscribe from this group and all its topics, send an email to elm-discuss...@googlegroups.com.

[douglas smith]

very nice! just found this thread today... seems to install and run just fine. i have a test running on aws 

is it possible to have single expressions eval as in elm-repl ?

i know they are two different systems  

let me know if i can help with anything- doug

[Austin Bingham]

There isn't currently any elm-repl support, though there's an open issue for it: https://github.com/abingham/jupyter-elm-kernel/issues/9

I'm not entirely sure how it should work, both from the user's perspective or technically, so any ideas would be great.

--