It would be nice to have some way of combining multiple cells into a single program. Right now each cell is an independent program. This may be clunky with Elm, esp. compared to Python.

so far, i tried running jupyter in the project directory but i have no output but when moved to the src directory, i was able to run it but i had issue with external modules not being available.

when i move elm.json to the src directory too, jupyther hangs during compilation.

can you please assist?

I installed the elm-kernel module via pip and realized that it requires 'elm-make', which is elm 0.18.

Then I installed the kernel via pip install -e . and it worked as expected, using elm 0.19.1

Steps to reproduce:

git clone https://github.com/abingham/jupyter-elm-kernel
copy default.nix and requirements.txt there

In a nix-enabled shell run:

nix-shell -I nixpkgs=https://github.com/NixOS/nixpkgs-channels/archive/nixos-18.03.tar.gz

I had to use nixos-18.03 to overcome the requirement to upgrade to elm 0.19 (elm-make renamed to elm make which would require kernel.json edits).

Also had to use pyzmq==17.1.3 instaed of pyzmq==18.0.0 (current latest version) to overcome zeromq/pyzmq#946 (had exact same error).

Reproducing the following error when opening the notebook after jupyter notebook:

[I 09:16:59.447 NotebookApp] Adapting to protocol v5.1 for kernel 8b6ce4f9-db58-4082-a242-8d67b7580676
/mnt/c/Users/myuser/pip-inside-nix-shell/venv/lib/python3.6/site-packages/notebook/base/zmqhandlers.py:284: RuntimeWarning: coroutine 'WebSocketHandler.get' was never awaited
super(AuthenticatedZMQStreamHandler, self).get(*args, **kwargs)

They should be able to specify the elm-make command-line, while still providing reasonable defaults. We need to figure out the proper way to do this in the Jupyter framework. Perhaps in the profile information or something.

The jupyter kernel API supports code completion and inspection. It would be nice if we could do this. Will elm-oracle help out here?

It is not possible to import other modules, e.g Http, since it's not part of the core language its needs to be installed with elm-package.

Take for example: http://elm-lang.org/examples/http

The error I got when running this:
b"I cannot find module 'Http'.\n\nModule 'Main' is trying to import it.\n\nPotential problems could be:\n * Misspelled the module name\n * Need to add a source directory or new dependency to elm-package.json\n"

When subprocess.run is running elm-make, it does install the core dependencies but it also tells that I have unsatisfied dependencies and it does not install it for me because it does not know what should be installed.

Currently the only way I found to make this work is to install the dependency with elm-package install elm-lang/http and make elm kernel copy elm-package.json to the temporary directory where the code is compiled.

So, the idea is if user wants to import modules, he will have to specify them in elm-package.json in the directory where notebook is running, then elm-kernel will copy elm-package.json and compile the code.

I can make a pull request if that makes sense.

We're at a point where the compile Elm is getting shipped back to the client, and we're loading it into the client, but we don't seem to be able to actually use it. I'm still getting to grips with all of the moving parts. There's a discussion about it on elm-discuss.

Right now we're using 3.6+ APIs in (at least) our use of subprocess.run()'s encoding parameter. We should see if we can support earlier versions, maybe 3.3+.

Esp. when there are compiler errors. We should capture those.

Does it currently only work if installed from source?

I installed the Node kernel, have Python3 by default and did pip install elm_kernel, see that it's installed with conda list but don't get an option for an elm notebook when I click the new dropdown after launching jupyter notebook.

Does this project allow to run full elm codes or only partial codes?

actually when i execute the code, it hangs with no output but the hello world example works.

It would be nice to have a few tests to verify that everything works as expected. I suppose we'll need to use something like nbconvert and its Python API to execute and inspect notebooks.

You can see some rough work on the smoke-test branch. This work is incomplete and doesn't do it's job, but it might serve as an inspiration.

I am new to elm, but shouldn't the new command be elm make instead of elm-make?

https://github.com/abingham/jupyter-elm-kernel/blob/master/elm_kernel/kernel.py#L78-L80

I am using nteract to write my notebooks. I don't have syntax highlighting in there using the elm-kernel ... I ask myself if this is a task of the kernel module ? I am not experienced with those things...

distutils is pretty limiting...

I get the following error when i try the SVG example:

I cannot find module 'Svg'.

Module 'Main' is trying to import it.

Potential problems could be:

• Misspelled the module name
• Need to add a source directory or new dependency to elm-package.json

I am using:
Notebook server version: 5.2.2
Python 3.6.4 on Win64

On first "run" it looks like the core packages getting installed as the error message looks like the following:

I cannot find module 'Svg'.

Module 'Main' is trying to import it.

Potential problems could be:

• Misspelled the module name
• Need to add a source directory or new dependency to elm-package.json
  Starting downloads...

● elm-lang/html 2.0.0
● elm-lang/virtual-dom 2.0.4
● elm-lang/core 5.1.1

Packages configured successfully!

Hydrogen is a jupyter front-end plugin for Atom.
It fails to execute elm code because the language field in the kernel json is text instead of elm.
We should change that in install.py to say "language":"elm".

In the meantime, I found a workaround:
We can create a language mapping in Hydrogen's package settings from text to elm so that it can find the correct kernel for the language. (Language Mappings: {"text":"elm"})

The current compilation strategy is simple but slow. It downloads all dependencies every time, and we might be able to speed things up by re-using the elm-stuff directory.

https://github.com/jupyter/echo_kernel/blob/master/echo_kernel/install.py

I have an elm notebook in my elm project directory. To get the notebook compiling, I had to 'hardcode' the absolute path to the source-directories property. I just had an array in there:
["src"], now I have like ["absolutepath/project/src"] ...

The compiler told me:

-- MISSING SOURCE DIRECTORY ------------------------------------------- elm.json

I need a valid elm.json file, but the "source-directories" field lists the
following directory:

    src

I cannot find it though. Is it missing? Is there a typo?


                            
Dependencies ready!

Right now we support code as it would appear in an Elm source file, but it might be nice to have an elm-repl version as well. Not sure if this needs to be a separate kernel or how it might interact with the existing mode. Maybe look for other kernels with similar needs (e.g. IHaskell?)