[BioC] Re: trouble installing "affy" package -- FIX

[Jeffrey Chang]

Robert Gentleman and I traced this problem to an incorrectly configured 
R_HOME_DIR variable in the R shell script.  I believe this was caused 
by a faulty R Portfile (port revision 1) in DarwinPorts for OS X and 
will affect anyone who has installed R this way.

I installed R 1.8.1 using DarwinPorts, which:
1) downloads and builds R
2) installs it into a "destroot" temporary directory
3) copies the "destroot" files into /opt/local

R works using two shell files called "R" installed in /opt/local/bin/R 
and /opt/local/lib/R/bin/R .  Both of those shell files contain a 
R_HOME_DIR variable that (among other things) tells R where to find the 
installed libraries.  In /opt/local/bin/R, R_HOME_DIR is correctly set 
to /opt/local/lib/R .  However, in /opt/local/lib/R/bin/R, R_HOME_DIR 
was incorrectly set to destroot/opt/local/lib/R.

The consequence of this is that R would at different points access two 
different sets of libraries.  The installer would install libraries 
into the /opt/local directory, and then be unable to find the library 
in "destroot".

The fix is to make sure the R_HOME_DIR is set correctly (in my case, 
/opt/local/lib/R) in both /opt/local/bin/R and /opt/local/lib/R/bin/R.

I will also forward an email to the port maintainer who can update this 
for future releases.

Thanks much to Robert and everyone else who wrote it with advice and 
suggestions.

Jeff




On Jan 29, 2004, at 5:53 PM, Jeffrey Chang wrote:

> I still don't know what the problem is, but I've hacked my way around 
> it in a *really* ugly way.  The key seems to be in the error message:
>> Error in getClass(Class, where = topenv(parent.frame())) :
>>          "MIAME" is not a defined class
>
> I downloaded the affy_1.3.27.tar.gz source and tried to install it 
> manually (using R CMD INSTALL).  I continued to get the message, and I 
> traced it down to code in AffyBatch.R.  That file begins with the 
> declaration of a new class:
>
> setClass("AffyBatch",
>          representation(cdfName="character",
>                         nrow="numeric",
>                         ncol="numeric"),
>          prototype=list(exprs=matrix(nr=0,nc=0),
>          se.exprs = matrix(nr=0,nc=0),
>          description=new("MIAME"),
>          annotation="",
>          notes="",
>          cdfName="",
>          nrow=0,
>          ncol=0), contains="exprSet")
>
>
> For some reason, R is not importing Biobase correctly (if anyone knows 
> why, please tell me!) so it doesn't know about MIAME and complains.  
> Similarly, it also doesn't know about "exprSet" so can't create the 
> class.
>
> Since I am not planning on using this file, I commented out the entire 
> file and installed the package without this functionality.  The 
> correct fix would be to figure out why R can't find Biobase, but I 
> don't know enough about R to do that.
>
> Jeff
>
>
> On Jan 29, 2004, at 1:15 PM, Jeffrey Chang wrote:
>
>> Hello everybody,
>>
>> I'm trying to install Bioconductor into a freshly compiled R 1.8.1 on 
>> OS X 10.3.2.  I'm sourcing the getBioC.R from 
>> http://www.bioconductor.org and trying to install the "release" 
>> version:
>>
>> > getBioC(relLevel="release")
>>
>> It looks like R is getting sources from
>> http://www.bioconductor.org/repository/release1.3/package/Source
>>
>> However, when it tries to install "affy", it complains about a 
>> missing "Biobase" package and the installation fails.  However, 
>> Biobase seems to be installed properly.  Executing at R prompt
>>     > library("Biobase")
>> seems to complete without error.
>>
>> However, executing
>>     > library("affy")
>>     Error in library("affy") : There is no package called 'affy'
>> fails.
>>
>> I have included a portion of the error messages at the bottom of this 
>> email.  Has anyone seen this before, or know how to fix it?
>>
>> Thanks,
>> Jeff
>>
>>
>>
>> [...]
>> ** R
>> ** data
>> ** demo
>> ** inst
>> ** save image
>> [1] FALSE
>> Warning message:
>> There is no package called 'Biobase' in: library(package, 
>> character.only = TRUE, logical = TRUE, warn.conflicts = 
>> warn.conflicts,
>> [1] TRUE
>> [1] TRUE
>> Error in getClass(Class, where = topenv(parent.frame())) :
>>          "MIAME" is not a defined class
>> Execution halted
>> cat: stdout: Broken pipe
>> ERROR: execution of package source for 'affy' failed
>> ** Removing '/opt/local/lib/R/library/affy'
>> Warning message:
>> Installation of package affy had non-zero exit status in: 
>> installPkg(fileName, pkg, pkgVer, type, lib, repEntry, versForce)
>> From URL:  
>> http://www.bioconductor.org/repository/release1.3/package/Source
>>        tkWidgets version 1.3.0
>>        affy version 1.3.27
>>
>