Re: [Gretl-users] r session bug more simply

On Fri, 25 Apr 2014, Ignacio Diaz-Emparanza wrote: > On 25/04/14 15:38, oleg_komashko(a)ukr.net wrote: >> >> to easily see the possible bug: >> >> run >> #begin >> nulldata 100 >> setobs 1 1910 --time-series >> series norm = randgen(N,15000,200) >> # end >> start interactive R session and type in R console >>> gretldata >> What I obtained started as follows: >> index norm >> 1910 1 15 >> 1911 2 1 >> 1912 3 60 >> >> ! > > I see no problem here. Annual data from 1910 to 2009 are correctly shown. > (Gretl from CVS with Ubuntu 12.04). I think it depends on the R version. In R 3.1.0 there's a "new feature" (in fact, a backwardly incompatible change) described as follows: "type.convert() (and hence by default read.table()) returns a character vector or factor when representing a numeric input as a double would lose accuracy." So if we happen to print data for consumption by R with more than (I suppose) 16 significant digits it gets silently borked.