I had a look at the source code and the coercion to double is only done one the sum of each row/column, so the "overhead" (e.g. memory) is only on the summed result. The integer matrix is *not*coerced to double before summing, which could be case if done before calling the native code. The latter would be consume lots of memory.
My $.02 Henrik On Wed, Jul 16, 2008 at 11:57 PM, <[EMAIL PROTECTED]> wrote: > I don't see the cost of doing so paying off. > > storage.mode is really only important if you are passing arguments to > compiled code. > > If you are passing to compiled code, you really need to ensure the > storage mode is what you think it is, anyway. > > Bill Venables. > > > > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of Herve Pages > Sent: Thursday, 17 July 2008 3:48 PM > To: R-devel@r-project.org > Subject: [Rd] rowSums()/colSums() don't preserve the 'integer' storage > mode > > Hi, > > Wouldn't that make sense to have rowSums()/colSums() to preserve the > storage mode? > > m <- matrix(1:15, nrow=5) > > > storage.mode(m) > [1] "integer" > > > storage.mode(sum(m)) > [1] "integer" > > > storage.mode(rowSums(m)) > [1] "double" <------------------- surprising! > > Cheers, > H. > > ______________________________________________ > R-devel@r-project.org mailing list > https://stat.ethz.ch/mailman/listinfo/r-devel > > ______________________________________________ > R-devel@r-project.org mailing list > https://stat.ethz.ch/mailman/listinfo/r-devel > ______________________________________________ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
