[Haskell-cafe] Parsing cabal files to calculate average number of dependencies

[Gwern Branwen]

On Fri, Jul 1, 2011 at 5:37 PM, Gwern Branwen <gwern0 at gmail.com> wrote:
> Looking at it, the index tarball contains the .cabal files for all
> versions known to Hackage, which isn't necessarily the interesting set
> of cabal files - I'm usually more interested in just the cabal files
> of the latest version of every package. No doubt there's a scripting
> solution (loop over the untarred directory of packages, and take the
> lexically last cabal file?), but it was easier to just exploit cabal
> fetch's behavior of fetching only the latest version and work with
> those tarballs.

The version using just the index tarball is kind of ugly; the
filtering and extracting doesn't seem terribly easy, so the best
script I could come up with was:

    cd ~/.cabal/packages/hackage.haskell.org && for DIR in */; do (for
CABAL in `tar --wildcards "$DIR" -tf 00-index.tar|head -1`; do (tar
-Oxf 00-index.tar $CABAL | runhaskell ~/deps.hs); done); done

(Parentheses aren't necessary but make it more readable.)

-- 
gwern
http://www.gwern.net