Re: [Haskell-cafe] Re: ANNOUNCE: fgl-5.4.2.3

On 12.07.2010 09:25, Ivan Lazar Miljenovic wrote:

A couple of points I meant to make here but forgot (I was busy hacking on this and my other three graph-related packages for over a week now, and especially this past weekend it cut into my sleeping...):

* Apart from bug-fixes, I don't intend on touching the 5.4 series any more. That said, I believe that this version is suitable for replacing 5.4.2.2 in the platform (what's the process on that?).

* After I get my generic graph class sorted out at AusHac this coming weekend, I intend to make a 5.5.0.0 release which extends the classes in this new library; this will probably _not_ be suitable for the platform and is intended to serve as a stepping stone to the replacement library Thomas Bereknyei and I are working on.

With that last point: Thomas and I are willing to call this new version/replacement something like "inductive-graphs" if that is the preference of the community. Does anyone know of a website that would let us have a survey we can use to determine which option people would prefer?

how about http://doodle.com? have fun, keep hacking. martin

Note that even if we give it a new name (rather than just a new major version number), we still intend on using the Data.Graph.Inductive module namespace (as it makes even more sense with the new name), so there will still be clashes between this new version and fgl.

Ivan Lazar Miljenovic writes:

...

I'm pleased to present the first new release of fgl [1] since Thomas Bereknyei took over maintaining it from Martin Erwig.

[1] http://hackage.haskell.org/package/fgl

Before people start panicking, rioting, etc., please check the version number: this is just a bug-fix release, and not the complete re-write version which we've been talking about (since we got a little sidetracked, etc.). As such, the API hasn't changed, and this should fit right in to packages already using fgl (sorry to all those people who followed my advice and put "fgl == 5.4.2.2" in the build-depends fields of their packages' .cabal files, but I didn't expect to make another 5.4.y release).

The exact change that has been made is to fix a bug pointed out to me by Tristan Allwood, in that Data.Graph.Inductive.PatriciaTree didn't support multiple edges (and furthermore this wasn't specified in the documentation). This has now been rectified. As an indication of what these changes mean, see this sample call graph produced by my SourceGraph program; when using PatriciaTree from fgl-5.4.2.2 the lines were all the same thickness; now there is among other things a loop of width 32 on getExp and a line of width 7 from getExp to maybeEnt.