I am looking into using Bioconductor as a slave batch processor for other programs. (BASE, other web tools, etc.) I've found that when using Bioconductor for the statistical analysis, a majority of the CPU time gets spent just loading it and not on the actual computation, which seems like a waste. This added overhead makes providing reasonably interactive output to the user difficult. While it might be possible to set up some sort of persistent R session that gets used, I'd rather reduce the complexity and just run R anew for each analysis. Breaking the dependency of Biobase on reposTools strikes me as a particularly effective way to optimize the load sequence. They both take a long time to load relative to the loading of R itself. (This seems to stem mostly from the use of methods. Profiling shows that 60% of CPU time gets spent in setMethod while loading reposTools.) While it's nice to be able to automagically download and install R libraries, the overwhelming majority of R sessions probably don't use this. (Especially if the R library directory isn't owned by the current user. The warning that pops up when this happens is another annoyance...) Is there some other showstopping reason for Biobase depending on reposTools?

Hi Colin, You have correctly identified one of the sources of slowness in loading. This will very likely go away over the next few months as the methods technology evolves and we put in place a few tools to make that faster. On the otherhand, for your purposes I don't think there is an issue with removing the reposTools dependency (or in fact some of the others). Again, there is an issue with the model (which we are working on as well), that there is only one level of dependency for packages. We hope to address that as well in the near future. Regards, Robert On Mon, Jun 16, 2003 at 12:25:41PM -0700, Colin A. Smith wrote: > I am looking into using Bioconductor as a slave batch processor for > other programs. (BASE, other web tools, etc.) I've found that when > using Bioconductor for the statistical analysis, a majority of the CPU > time gets spent just loading it and not on the actual computation, > which seems like a waste. This added overhead makes providing > reasonably interactive output to the user difficult. While it might be > possible to set up some sort of persistent R session that gets used, > I'd rather reduce the complexity and just run R anew for each analysis. > > Breaking the dependency of Biobase on reposTools strikes me as a > particularly effective way to optimize the load sequence. They both > take a long time to load relative to the loading of R itself. (This > seems to stem mostly from the use of methods. Profiling shows that 60% > of CPU time gets spent in setMethod while loading reposTools.) > > While it's nice to be able to automagically download and install R > libraries, the overwhelming majority of R sessions probably don't use > this. (Especially if the R library directory isn't owned by the current > user. The warning that pops up when this happens is another > annoyance...) Is there some other showstopping reason for Biobase > depending on reposTools? > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor -- +--------------------------------------------------------------------- ------+ | Robert Gentleman phone : (617) 632-5250 | | Associate Professor fax: (617) 632-2444 | | Department of Biostatistics office: M1B20 | | Harvard School of Public Health email: rgentlem@jimmy.harvard.edu | +--------------------------------------------------------------------- ------+