juvix/notes/tooling.md

Tools used so far:

• cabal-edit

• hlint and checkout https://github.com/kowainik/relude/blob/main/.hlint.yaml for a complex configuration and better hints.

• stan

• summoner

• ghcup

• implicit-hie generates hie.yaml.

• For a good prelude, I tried with Protolude, but it seems a bit abandoned, and it doesn't have support the new Haskell versions. Relude just offered the same, and it is better documented. Let us give it a shot. NoImplicitPrelude plus base-noprelude. https://kowainik.github.io/projects/relude

• For Pretty printer, we will use the package https://hackage.haskell.org/package/prettyprinter, which supports nice annotations and colors using Ansi-terminal subpackage: cabal-edit add prettyprinter-ansi-terminal.

• Two options for the kind of container we can use for Context. Using the package container and for performance, unordered-containers. The latter seems to have the same API naming than the former. So, in principle, it doesn't matter which we use.

• See gluedeval. During elaboration different kind of evaluation strategies may be needed.

  • top vs. local scope.

• On equality type-checking, see abstract

• To document the code, see https://kowainik.github.io/posts/haddock-tips

Initial task order for Minijuvix indicated between parenthesis:

1. Parser (3.)
2. Typechecker (1.)
3. Compiler (2.)
4. Interpreter (4.)

• On deriving stuff using Standalone Der. See https://kowainik.github.io/posts/deriving.

• To avoid boilerplate in the cabal file, one could use common stanzas. At the moment, I'm using cabal-edit to keep the bounds and this tool does not support stanzas. So be it.

• Using MultiParamTypeClasses to allow me deriving multi instances in one line.

• TODO: make a ref.bib listing all the repositories and the source-code from where I took code, inspiration, whatever thing.

• The haskell library https://hackage.haskell.org/package/capability seems to be a right choice. Still, I need to check the details. I will use Juvix Prelude.

• Let us use qualified imports to prevent namespace pollution, as much as possible. Checkout:

  • https://www.haskell.org/onlinereport/haskell2010/haskellch5.html
  • https://ro-che.info/articles/2019-01-26-haskell-module-system-p2
  • https://mmhaskell.com/blog/2017/5/8/4-steps-to-a-better-imports-list.

• TODO: https://kowainik.github.io/posts/2018-09-09-dhall-to-hlint So far, I have proposed a hlint file, it's clean, but for refactoring, seems to me, the link above shows a better approach.

• Let us use the Safe pragma. https://stackoverflow.com/questions/61655158/haskell-safe-and-trustworthy-extensions