This article become outdated as there is a sandbox support in cabal starting from 0.18.
See <a href="http://coldwa.st/e/blog/2013-08-20-Cabal-sandbox.html#for-the-users-of-cabal-dev">blog post</a>
for details

Q: how to use cabal?

A: just cabal install cabal-dev

Cabal is a great system of building and packaging Haskell libraries but is has some downsides such as:

  1. it has very poor handling of library versions and tracking dependecies.

  2. it can’t have multiply versions of same library (really it can but then it becomes a pain to use them)

So sometimes it becomes a pain to fix haskell tree.

Cabal-dev is a cabal-install wrapper that build project and dependencies in sandboxed enviromnent, this allowes to use multiple version of the same library and so chanses to break system are very low. Additionally it generates image of installed application so you can check how it will be installed on your system without real installation.

Of cause as there are some problems:

  1. cabal-dev is slow, because you should install all dependencies in sandbox environment (unless they are installed system wide).

  2. cabal-dev can’t handle some Setup.hs options at least it was true in Dec-2011.

First problem easily fixed in gentoo, one can install basic dependencies with emerge, so portage will track tree consitency with haskell-updater <http://www.haskell.org/haskellwiki/Gentoo#haskell-updater> (replacement tool for ghc-updater) and has all other features out of the box. So cabal-dev will be using most of libs from system and some other will be sandboxed.

For second one I had a workaround by creating makefile that run setup.hs and then cabal-dev intall.

Also cabal-dev has additional feature of running built program in ghc (i.e. run ghci and load every module in project) it can be done with

cabal-dev build && cabal-dev ghci




comments powered by Disqus