Cabal or Stack

What we do at IRIS Connect

Posted on 2020-04-11

Cabal or Stack? At IRIS connect we have built an app that:

  • reads in a YAML file that describes the stack in terms of a Stackage LTS snapshot plus extra dependencies that override and extend the snapshot;

  • we maintain for each of our own library packages a database associating the packages with the GitHub repositories and the SHA of the commit for the current release, which gets used instead of the version number in stack.yaml files;

  • we also upload our packages into a local Hackage server;

  • we build a stack.yaml and then use stack ls dependencies to extract the vector of packages that makes up our stack;

  • we generate the stack.yaml and cabal.project files to be checked into the source tree;

  • stack loads our packages from our GitHub organization while cabal gets them from our Hackage server;

  • developers are free to build with stack or cabal, whichever works best with their workflow.

I know of others that base their stacks on Stackage releases even if cabal might be used to build in the end.

It is a remarkably satisfying arrangement of some outstanding tools.

Cabal and stack; Hackage and Stackage.

Got an issue with any of this? Please share or drop me a line (see below).