I have mixed feelings about this change. I see why it's useful and arguably it's consistent with all the other features cabal2nix has gotten over the years, i.e. support for hpack, support for fetching git repositories, etc. So from that point of view, I think it's perfectly fine to add that feature.

On the other hand, I feel like all these changes are going into a completely wrong direction. cabal2nix is a tool that converts a Cabal file into a Nix expression. That's its domain. So why should it know about git, and http and hpack and --subpath and whatnot else? None of this has anything to do with converting Cabal files into Nix expressions.

What is happening now feels like feature creep. Meanwhile, the list of dependencies one needs to build cabal2nix is getting crazy long and complicated. We already need regular releases to accomodate API breaking changes in hpack. Now we're adding dhall into the mix, too?

I don't know ... Personally, if I had may way then I'd throw out all of these features in a potential 3.0 release and restrict cabal2nix's functionality to reading a Cabal file from the local disk and writing a Nix expression to the local disk -- nothing else.

If other maintainers of this code feel that this is a useful addition that we should merge, then I'll happily defer to their judgement, but I am probably not going to merge these changes.