On Fri, Jan 3, 2014 at 2:55 PM, Adam Bergmark
Hi Daniel,
If I'm getting this correctly, than Halberd uses some kind of static database, right?
Right.
To be more clear, it uses the database built by hs-gen-iface [0], which is a compile-like tool that generates an interface database for your packages. You have to make a separate `cabal` call to use it and build the database for each package you compile. Hopefully in the future, cabal will have a way to automatically call compiler-like tools (think haddock as wel). But for now, it's a bit impractical.
I think the nice approach of 'vim-hsimport' is, that by using 'hdevtools' in conjunction with a 'cabal sandbox', only the modules of the packages are considered, which your project depends on, and this information is dynamically "established" by 'hdevtools' reading the package database of the 'cabal sandbox'.
I recall Erik and others discussing ways of restricting halberd to a sandbox as well, but you'd have to ask him about the progress.
Right now halberd uses the global and user database. It should be fairly trivial to add support for cabal sandboxes, although there will probably be some duplication with cabal in the sandbox flags at least. Let me know if you're interested. Regards, Erik [0] http://hackage.haskell.org/package/hs-gen-iface