[BioC] Rgraphviz load problem

[Byron Ellis]

On Oct 28, 2005, at 7:55 PM, Seth Falcon wrote:

> Hi Byron,
>
>
> On 28 Oct 2005, ellis at stat.harvard.edu wrote:
>
>> How to Build Rgraphviz From Source on OS X
>> or
>> The Path Of Pain
>> or
>> Can I Have Those Two Hours Back Now?
>>
>> A Cautionary Tale
>>
>
> Heh, I walked a similar path of path to get graphviz to build on
> Windows.  I wonder if you might consider posting your steps to the
> graphviz-devel list ;-)

Possibly. I don't how they'd take emails titled "A Cautionary Tale"  
though. I had hoped that the move to Cairo would fix some of the  
rendering dependency issues, but unfortunately they haven't seemed to  
grasp the fact that Cairo is indeed a cross-platform library so they  
require gtk+, which of course I don't have since I use the quartz/ 
ATSUI backend. *sigh*

>
> The most frustrating thing is that all of those bits that don't
> compile are parts that Rgraphviz does not need.  AFAICS, graphviz
> couples graph layout with graph _rendering_ and that pulls all sorts
> of dependencies into the mix when all one wants to do is retrieve the
> layout :-/
>

Indeed. The ability to

> On OS X I have graphviz 2.2.0 installed (which built without pain) and
> it works fine for Rgraphviz.  But I don't seem to have the tarball
> anywhere so I don't know where to suggest retrieving it.

That must have been prior to the release of Xcode 2.1---it no longer  
builds so easily. In particular, the custom yacc and lex scripts bail  
and I was too lazy (strangely enough, but I really dislike mucking  
with yacc) to attempt to fix them. You would also need to update the  
framework building to put them somewhere useful (like /Library)  
rather than @executable_path/../.. since you'd have to drop them into  
R.app otherwise.

Of course, now that Rgraphviz is at least a reasonable facsimile of  
working its unlikely I'll do any deeper investigation between now  
and, say, June ;-)

>
> + seth
>
>

---
Byron Ellis (ellis at stat.harvard.edu)
"Oook" -- The Librarian