#eclipsefp

EclipseFP clearly doesn't deal with very large Haskell files, at least those generated by Happy.  The PHP parser in Phaestus has a source file of about 1400 lines and generates a Haskell source file that's about three quarters of a megabyte in size.  Trying to work with this file in EclipseFP makes my system nearly unusable due to frequent launches of buildwrapper which suck down 2 gigabytes or more of memory, which I suspect is due to buildwrapper trying to deal with the very large Happy output file.  Unfortunately, there doesn't appear to be a good way to tell EclipseFP or buildwrapper to disregard the Haskell file generated by Happy.

I'm seriously considering attempting to rewrite this parser in Parsec, but the problem I have there is that PHP really needs a modal lexer, and I need to figure out how to do this.  I suspect the answer is going to be a custom monad type, but I don't really understand Parsec well enough to be sure.  It might be possible to use the Alex lexer I already have with Parsec.

I spent much of today trying to get my Haskell toolchain into order, in preparation for actually starting to do development again. In the past I’ve used the EclipseFP plugin to Eclipse, and since I still have it installed I figured I’d just update it and the underlying Haskell tools it needs.

Updating EclipseFP was easy. Unfortunately, the GHC toolchain it uses was less cooperative. There is a problem of some sort with scion-browser not getting along with persistent, which prevents me from actually building any version of scion-browser. I’ve tried a half dozen different combinations, and none has worked.

Make sure there are no additional packages installed into a global scope! (not even a new version of cabal-install)

# cd ~/.cabal # mkdir sandbox # cd sandbox/ # cabal sandbox init # cabal install buildwrapper scion-browser hoogle hlint htf alex happy stylish-haskell uuagc --force-reinstalls # cp ./.cabal-sandbox/bin/* ~/bin # cd # hoogle data all