G'day all. On Thu, Oct 10, 2002 at 11:50:39AM +0200, Jerzy Karczmarczuk wrote:
There are of course more serious approaches: intervals, etc. The infinite- precision arithmetic is a mature domain, developed by many people. Actually the Gosper arithmetic of continued fractions is also based on co-recursive expansion, although I have never seen anybody implementing it using a lazy language, and a lazy protocol.
Anybody wants to do it with me? (Serious offers only...)
Already did it. It's not pretty, but I'll send you my implementation off-list. One thing I'd like to see is a lazy implementation of linear fractional transformations. The reason I'd like to see this is that it's easier to implement fixpoint-style computations (e.g. transcendental functions) using LFTs than using continued fractions. Cheers, Andrew Bromage