phischu / haskell-modules Goto Github PK

Currently the output folder is hard-coded for my machine. This of course is not ok.

Extracting modules from a package takes longer than compiling it because we compile it and do extra work. We have to find a way to avoid compilation, maybe by writing fake object files or something.

Currently we put all modules in to the same folder. But it would be nice if we put modules from different packages into different folders named after the package.

We currently take those command line arguments that start with an upper case letter as module names. This breaks down as soon as there is a package that starts with an upper case letter.

If we write a GHC frontend plugin GHC will do the parsing of command line arguments for us. See http://blog.ezyang.com/2017/02/how-to-integrate-ghc-api-programs-with-cabal/

We need to know which C source and C header files a package needs.

We want to extract all relevant modules including those in the global package database that come with GHC. Also including base, ghc-prim and integer-simple or integer-gmp.

array-0.5.0.0 fails with the following error message

user error (The package 'array' requires Cabal library version -any && >=1.10
but no suitable version is installed.)

For some reason this works on my machine.
We have to find out if we can expose a different package-db for the Setup script .

This would also help with #5 and allow us to take advantage of GHC's new -keep-hscpp-files flag.