[BioC] automatic package installs

[Kasper Daniel Hansen]

So in case you are not too familiar with affy chips: in general you  
can be certain that there is no automatic package installation in R/ 
BioC. The packages Jim is referring to is (kind of) annotation  
packages for working with specific microarray chips. These packages  
does not really contain code, but rather some data which is necessary  
to manipulate the chip (basically the chip layout). When you read an  
Affymetrix data file (CEL file) it contains information about the  
type of chip. And then affy/gcrma tries to download the relevant chip  
packages (of course this only happens if the packages are not already  
installed). The package download only happens when the user tries to  
do something that requires the data package. So in your case it is  
perfectly fine to read in the cel file, but as soon as you try to  
print it, it checks for the package.

I would guess the sys admins are primarily afraid of possible  
versioning conflicts on a multi-user system. Generally these layout  
packages (cdf, probe) are tied to the chip and there is no reason to  
update the packages - unless there have been some internal changes in  
R. As Jim is pointing out, it is possible to configure R so that by  
default any package installation happens in the given user's home  
directory, instead of in the system area. They might want to look  
into the R function .libPaths and the environment variable R_LIBS.

Having said all of this, there is in principle nothing that prevents  
a package writer from including code that installs arbitrary packages  
when executed. And that will override the configuration settings  
above. So if they are super paranoid, they should not allow people to  
run R form an account that has write permission to the R_HOME folders.

Kasper


On Nov 3, 2006, at 1:27 PM, James W. MacDonald wrote:

> Hi Patrick,
>
> Votruba, Patrick G. wrote:
>> Hello,
>>
>> My colleague recently received the following error message and it has
>> caused some concern among our system administrators.  The
>> getProbePackage function tries to download and install a package  
>> that's
>> not locally available and fails because the user doesn't have write
>> permission to the library directory.  There are, however, a core  
>> group
>> of R users who do have write permission and we'd like to prevent
>> unintentional package installations.  Is there an easy way to disable
>> this function?
>
> Not really, but if one is trying to run a function that requires a
> particular package to work, is it really an unintentional package
> installation?
>
> Anyway, the canonical method to handle such situations is to create a
> local library in one's home directory (or somewhere that the  
> particular
> user has write access), and then set .libPaths() to include that
> directory. Then the package will be installed there, the user can  
> do the
> intended analysis, and the 'usual' R library directory is unaffected.
>
>
>>
>> Also, this has sparked some debate as to whether bioconductor  
>> software
>> also automatically updates existing packages.  If so, can this  
>> also be
>> disabled?
>
> I don't know of any BioC software that automatically updates existing
> packages. If there are some, you would need to disable them
> individually, as AFAIK, this isn't a common practice.
>
>>
>> Do automatic package installs/updates occur for all packages or  
>> just a
>> subset (e.g. annotation packages)?
>
> I only know of two packages that do automatic installs of packages;  
> affy
> and gcrma. These packages will install cdf packages (affy) or probe
> packages (gcrma), both of which are necessary for the respective
> packages to function. The only way to disable the automatic  
> installs for
> these packages is to not have write access to the library, which as  
> you
> found causes them to error out.
>
> Best,
>
> Jim
>
>
>
>
> -- 
> James W. MacDonald, M.S.
> Biostatistician
> Affymetrix and cDNA Microarray Core
> University of Michigan Cancer Center
> 1500 E. Medical Center Drive
> 7410 CCGC
> Ann Arbor MI 48109
> 734-647-5623
>
>
> **********************************************************
> Electronic Mail is not secure, may not be read every day, and  
> should not be used for urgent or sensitive issues.
>
> _______________________________________________
> 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