Re: [boost] Big integer continued

From: Dave Harris (brangdon_at_[hidden])
Date: 2005-04-29 07:57:31

In-Reply-To: <d4rdqo$vhr$1_at_[hidden]>
kalita_at_[hidden] (Marcin Kaliciñski) wrote (abridged):
> There's an issue with some of the bitwise operations, as they are only 
> well defined for integers of fixed size. What should for example 
> ~boost::integer(0) return? Ideally an infinitely long integer filled 
> with binary 1s.

Some of the functions in n1744 look like they would be most useful when 
using an integer as a kind of vector<bool>. Is that intended as a 
reasonable use?

It could be convenient if there were some way to specify a mask of 
arbitrarily many 1s. In effect, a left-shift with 1-fill. Given that, 
don't see a need for ~boost::integer(0) and I'd rather it didn't compile.

> boost::rational does not have a conversion operator to integer type, 
> which is good as most of rationals are not integers. Should 
> boost::integer have such an operator, with runtime range checking?

The conversion should be explicit in my view, perhaps via 
boost::numeric_cast<>.

> 2. Implementation

Initially I'd be happy if you just provided one portable implementation 
and an efficient one for your favourite platform. Lack of an efficient 
implementation for any given platform should not be a reason to reject the 
library.

-- Dave Harris, Nottingham, UK.