Re: Let's gather some evidence!

[Marco Nehmeier <nehmeier@xxxxxxxxxxxxxxxxxxxxxxxxxxx>]

Arnold Neumaier schrieb:
> John Pryce wrote:
>
> > - How do you envisage "returning more information besides an interval" 
> > being implemented? If it is by the mechanism of an extra output 
> > argument, so c = a + b becomes c = plus(a,b,flag) as I believe is the 
> > scheme in the draft C++ interval standard, how should one remedy the 
> > huge inconvenience to the user, who has to rewrite existing function 
> > code?
>
>
> With two datatypes, one for decorated and one for undecorated
> intervals, this is easy.
>
> Expressions like x=a+b can still be interpreted as before - if a,b are
> undecorated intervals, the result will be undecorated, if they are
> decorated intervals the result will be decorated.
>
> Alternatively, one could use compiler directives that tell how a given
> block of code is to be treated, the default being using no decorations.
>
> Automatic detection of what is needed is also a possibility.
>
>
> Arnold Neumaier


In C++ this is easy,

I have written a test-implementation with the two datatypes
Interval and ExpressionResult (a flagged interval).

With an overloaded assignment-operator it is easy and comfortable.
You don't need any expression like c = plus(a,b,flag). You can simply write

Interval a = ...
Interval b = ...

ExpressionResult c = a + b;

if (c.getFlag())
	Interval i = c;

But it is also possible to write

Interval c = a + b;

without flags.

--
     o           Marco Nehmeier, Lehrstuhl fuer Informatik II
    / \          Universitaet Wuerzburg, Am Hubland, D-97074 Wuerzburg
InfoII o         Tel.: +49 931 / 31 88684
  / \  Uni       E-Mail: nehmeier@xxxxxxxxxxxxxxxxxxxxxxxxxxx
 o   o Wuerzburg