On Monday, June 9, 2014 7:22:36 AM UTC-4, Stéphane Laurent wrote: > > That's why we need BigRational > <https://github.com/andrioni/BigRationals.jl> :) >> >> >>>> Doesn't Rational{BigInt} work fine? Of course, the problem with all bigint rational approaches is that the denominator generically grows exponentially as you do more operations, and so even trivial problems like summation have exponential complexity.
Floating-point arithmetic is a pretty well though-out compromise for doing arithmetic in a subset of the rationals. It would be nice to have decimal floating-point support, though, for cases where you need to represent human (decimal) inputs exactly.
