[BioC] Rgraphviz: Edge attributes not displaying

Kasper Daniel Hansen khansen at stat.Berkeley.EDU
Fri Feb 16 06:03:45 CET 2007


I believe I have been a bit quick. I should be able to include the  
fix in the autoconfigure script of Rgraphviz, so don't go and apply  
the stuff below, unless you have a problem with command-line Graphviz  
or other applications using Graphviz. Hopefully this will be nailed  
down tonight.

Kasper

On Feb 15, 2007, at 7:56 PM, Kasper Daniel Hansen wrote:

> A fix has been found to the problem reported by Jarrett: the fact
> that edge labels are not showing up on Mac (and possibly 64bit Linux
> as well as reported by Li).
>
> Unfortunately the fix is distinctively not nice :) I have only tested
> this on Graphviz 2.12 on a Mac G4. You need to modify the Graphviz
> header files which usually (if you have a default installation)
> resides in
>    /usr/local/include/graphviz
> The file in question is called
>    logic.h
>
> At the start of the file, just below the lines
>
> #ifndef GV_LOGIC_H
> #define GV_LOGIC_H
>
> (which are at lines 17 and 18 in my installation), add the two lines
>
> #define HAVE_BOOL 1
> #define HAVE_STDBOOL_H 1
>
> (you will probably need to have superuser privileges to do so, on a
> Mac do
> # sudo pico logic.h
> )
>
> And reinstall Rgraphviz from source (you do not need to recompile
> Graphviz, but you need to recompile Rgraphviz).
>
> I am very curious if this works with
> 1) Simon's Graphviz-2.8 binary from r.research.att.com
> 2) 64bit Linux
> 3) Intel Mac (not sure if there was a problem to begin with on that
> platform)
> so please send me an email if you try this.
>
> For background, read the post by Emden R. Gansner on the Graphviz-
> interest email-list. Apparently they were being a bit too smart
> regarding the use of bool, which on Apple's machines have sizeof
> (bool) = 4.
>
> Kasper
>
> On Feb 5, 2007, at 9:28 AM, Kasper Daniel Hansen wrote:
>
>>
>> On Feb 5, 2007, at 9:22 AM, Li.Long at isb-sib.ch wrote:
>>
>>>> Nope, I get something very different
>>>>
>>>>>
>>>> character(0)
>>>>> AgEdge(g2)[[10]]@txtLabel at labelText
>>>> character(0)
>>>>
>>>> Given what I saw looking at the edgeAttrs of the Ragraph object
>>>> itself, (namely, Null), I'm guessing that the problem is there.
>>>> This
>>>> may also be why edge style (solid v. dashed) wasn't getting picked
>>>
>>> The problem with "lty" is different from this, we didn't even try
>>> to set
>>> that entry.  Here we do try, but don't get it right.
>>>
>>>> up.  Why it is different on linux instead of a mac, I am uncertain,
>>>> unless there are different binaries in the Bioconductor archive.
>>>
>>> Kasper,
>>>
>>> I see this problem on a 64-bit Linux machine as well.  So this
>>> might be a
>>> 64-bit vs. 32-bit issue.
>>>
>>> The size of a pointer is different on 64-bit machine vs. 32-bit
>>> machine,
>>> I'd pay some attention to that, since we are looking at the pointers
>>> here...
>>
>> There is also the endian issue on the PPC (I am on a 32 bit machine)
>>
>>> I'll take a closer look at it tomorrow.
>>
>> It is rather bewildering. I have quite a lot of experience of looking
>> at how _node_ labels are handled in graphviz' C structures, and in
>> this case the edge labels are certainly behaving differently. Of
>> course, I guess edge labels are not used as much, so it could be a
>> bug somewhere.
>>
>> The fact you see it on 64 bit Linux is at least a bit comforting :) I
>> guess it is using Graphviz 2.12?
>>
>> Like I just wrote: I'll follow up on the Graphviz list later today.
>>
>> Kasper
>>
>>
>>> Li
>>>
>>
>> _______________________________________________
>> Bioconductor mailing list
>> Bioconductor at stat.math.ethz.ch
>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>> Search the archives: http://news.gmane.org/
>> gmane.science.biology.informatics.conductor
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/ 
> gmane.science.biology.informatics.conductor



More information about the Bioconductor mailing list