On Wed, 21 Nov 2007, Robin Hankin wrote: > > On 21 Nov 2007, at 08:30, Prof Brian Ripley wrote: > >> On Tue, 20 Nov 2007, Tim Hesterberg wrote: >> >>> I wrote the original rowSums (in S-PLUS). >>> There, rowSums() does not coerce integer to double. >> >> Actaully, neither does R. It computes a double answer but does no coercion >> per se. >> >>> However, one advantage of coercion is to avoid integer overflow. >> >> Indeed, as I told Robin Hankin privately, that was the design reason. >> > > > Brian Ripley also reminded me that the sum() of integers is an integer, > behaviour that I find desirable. > > The reason for my starting this thread is that > sometimes I actually *want* sums of > integers to overflow: my interest is in exact computations > where I must be absolutely certain that there can be no rounding error. > > If the sum cannot be represented > in integers, I want this fact to be flagged with extreme vigour as it signals > what > might be catastrophic loss of precision.
Doubles hold integers to a much higher precision (up to 2^53-1), so you can just check if any of the results exceed .Machine$integer.max. It's very unlikely that you are summing enough integers for there to be a possibility of floating-point imprecision in an intermediate sum. > > At least, that's my current thinking. > > > > best wishes > > > rksh > > >>> >>> Tim Hesterberg >>> >>>> ... So, why does rowSums() coerce to double (behaviour >>>> that is undesirable for me)? >>> >>> ______________________________________________ >>> R-help@r-project.org mailing list >>> https://stat.ethz.ch/mailman/listinfo/r-help >>> PLEASE do read the posting guide >>> http://www.R-project.org/posting-guide.html >>> and provide commented, minimal, self-contained, reproducible code. >>> >> >> -- >> Brian D. Ripley, [EMAIL PROTECTED] >> Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ >> University of Oxford, Tel: +44 1865 272861 (self) >> 1 South Parks Road, +44 1865 272866 (PA) >> Oxford OX1 3TG, UK Fax: +44 1865 272595 > > -- > Robin Hankin > Uncertainty Analyst > National Oceanography Centre, Southampton > European Way, Southampton SO14 3ZH, UK > tel 023-8059-7743 -- Brian D. Ripley, [EMAIL PROTECTED] Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UK Fax: +44 1865 272595 ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
