Actually, J has extended precision integers and rational numbers. With rational numbers it is possible to compute various functions to arbitrary precision. See for example: http://www.jsoftware.com/jwiki/Essays/Extended_Precision_Functions
Introducing into J arbitrary precision floating numbers (big floats, BF) would be non-trivial, but easier than the impression that Devon gives. Each BF (or all BFs globally) would include an indication of how many digits of precision are represented. So the abstract number 1%3 to 30 decimal digits would be 0.333333333333333333333333333333 and 250 to the same number of digits would be 250.000000000000000000000000000 Suppose BF were implemented in J. If x is the first BF above and y is the second, then you should be able to do: x * y 83.3333333333333333333333333333 x - y _249.666666666666666666666666667 ^. y 5.52146091786224643321951012114 ^ y 3.74645461450267326034995481220e108 Algorithms would have to be found for all the different primitives (as in the essay). %. and H. should be interesting. -------------------------------------------- >From Devon McCormick <[EMAIL PROTECTED]> Sent Monday, February 26, 2007 9:10 am To Programming forum <programming@jsoftware.com> Subject Re: [Jprogramming] Lack of Precision? Geoff - J's extended precision applies to integers only. If you think about it, you'll see that while it's relatively simple to make integers arbitrarily large, floating point is a much more difficult proposition: how would you know when to stop in the potentially infinite expansion of a floating point number? Even something as simple as %3 is a problem. This restriction to integers is why something like Ahmad's <[EMAIL PROTECTED]: 2x*10x^200 works: "<[EMAIL PROTECTED]:" applies both the square root and enforces the integer limitation at the same time. On 2/26/07, Geoff Canyon <[EMAIL PROTECTED]> wrote: > > (this is my first J code) > > I'm executing this to get the square root of 2: > > x:%:x:2e100 > > I also tried: > > <.((%:2)*10^100) > > In both cases I'm looking for extended precision (100 places if J > supports it). > > In both cases it appears the digits take a left turn about 20 digits > in -- they don't match publicly available resources on the web. > > Is there something else I should be doing, or is there a limit to J's > precision? ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
