Comparing xeus-Haskell and ihaskell kernels | Hacker News Reader

Comparing xeus-Haskell and ihaskell kernels

datahaskell.org

13 pointsmchav8 comments

When I was toying with Haskell I was unable to get GHCi to interpret multiline definitions (Haskell is declarative, which means you don't make commands, but you declare commands which then get executed).

Jupyter+conda/anaconda is almost always not installable and broken for most of its kernels (it was for Haskell), so I ended up piping GHCi to a websocket server that I found on GitHub and a minimal HTML/JS frontend to talk to it: a reorderable list of textareas.

Then I added some security in the form of passing and verifying long random strings (via files, URL arguments and the user, that one is annoying). GHCi is capable of shelling out, so letting every browser tab on my computer talking to it is not ideal.

AFAIK Jupyter does not have any advanced features like dependency tracking between cells, or tracking variables, data, etc. So if you want Haskell in the browser, I recommend my approach, it is small, fast, the UI is configurable, and the whole process is hackable with hooks and transformations at any place.

https://discourse.haskell.org/t/a-lighter-weight-alternative... talks about workarounds for multiline definitions

It certainly can be inefficient depending on what you want to do, but loading files with functions into the scope of your GHCi session works quite well for quick debugging. When cabal repl all necessary deps for the project get loaded and it just works.

RE Jupyter not having advanced features.

Yeah it's a bummer. It seems that notebooks that support these sort of "reactive" workflows are custom built around that model. Marimo, Pluto.jl, and observable are mostly language specific. Creating one would be non trivial.

Do you have your approach documented (tutorial style) anywhere?

Wow ! Great use of MicroHS. What is the performance difference between interpreted MicroHS and warm compiled Haskell for a practical program eg. Shellcheck or similar ?

The rule of thumb is somewhere between 5 and 10x difference. Which is large if you're going to do anything heavy but for most practical purposes it's fine. Roughly the difference between C and Python.

The author of Xeus-Haskell is here. Ask me anything.

you can/should post that in the jupyter zulip chat.

Reply on news.ycombinator.com