On Nov 30, 2007 10:21 AM, Arie Groeneveld <[EMAIL PROTECTED]> wrote: > Try this (extending input range): > > (2+i.50) -: 10 d0f0 10*^:(<50)10 > 0 > (2+i.50) -: 10 d0f0 10*^:(<50)10x > 1 > > (1+i.50) -: 10 d0f0 _1+10^1+i.50x > 1 > (1+i.50) -: 10 d0f0 _1+10^1+i.50 > 0 > > It seems that all the proposed verbs produce wrong values > for some input values, unless you provide some of them with > ext. prec. input values. Even build in ext. prec. (d0fx) > doesn't seem to work.
I'm not completely sure, but I believe that's not a bug. Those big powers of ten cannot be represented exactly as a double, and the closest double to them, when rounded back to an integer, sometimes actually start with 999... (try 0j_16&": on those numbers). This means that the verb d0f0 is actually precise. Ambrus ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
