Re: [boost] Interest in an arbitrary precision library?

From: Joel Eidsath (jeidsath_at_[hidden])
Date: 2005-09-03 16:00:23

>
>Maths my field too (I thought you might figure that out of the expression 
>'Z[sqrt(2)])(log(2))'  ) but now I do not understand your original point (9)
>  
>
Sorry about that.  It looks like I got some words got cut out from my 
original post at 9).  Basically it's a choice between:

//first implementation of "set_precision"
rational a,b;
a.set_precision(50);  //a's precision now set at 50
b.set_precision(150)  //b's precision now set at 150, a's precision 
still at 50
//Do some math
cout << (a + b) << endl << (b + a)  //issues with the return values of 
the expressions here

//second implementation
rational a,b;
set_precision(25)  //a's and b's precision now at 25

I much prefer doing things the second way, but I wanted to get some 
comments.  Thread safety will be an issue the second way.

>Anyway I can say I agree with you on almost everything. Do you want to move 
>forward with this idea?
>  
>
I'd like to. 

I discovered N1692 and N1794 while looking through the C++0x proceedings 
yesterday.  Here is the second:  
http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2005/n1781.pdf

It is a proposal for an arbitrary integer library.  It doesn't really 
say anything about implementation, but I think that I'll follow its 
guidelines -- whether it is adopted or not (has it been already? -- I 
don't know the comittee process), it's a good interface.  Once that is 
finished, I'll get started on the rationals side of things.

Joel Eidsath