Re: [q-lang-users] Special values or failing functions

[Albert G. <Dr....@t-...>]

John Cowan wrote:
> Presumably "const bottom;" somewhere in the standard library is enough?
> I like the name "bottom".

No, that's a misunderstanding. I was talking about an operation which
bails out of the current reduction, not a general-purpose sentinel value
to be returned by a function. That is, an operation like 'fail'
[http://q-lang.sourceforge.net/qdoc/qdoc_10.html#IDX287], but making the
entire reduction fail (effectively turning the redex being reduced into
a normal form), not just the current equation. Example:

foo X Y = _FAIL_ if not (isnum X and then isnum Y);
        = /* ... */;

So, e.g., 'foo 99 bla' would be a normal form, but subsequent equations
may be tried if X and Y are both numbers.

Alexander, I think that this is what you wanted, right?

> IMHO it's better to use "nan" than "inf" in the numeric case.  It has
> exactly the right semantics.

The specific application was shortest path computations, where it makes
perfect sense to use 'inf' for the length of a non-existing edge,
because X+inf = inf and max X inf = inf if X is either a number or inf.
That's also what you'll see in many text book algorithms for the
shortest path problem, whether it is Bellman-Ford, Dijkstra, or the
maxtrix-based algorithms. Of course this only works for algorithms where
you don't have to distinguish between non-existing and infinite-length
edges.

Albert

-- 
Dr. Albert Gr"af
Dept. of Music-Informatics, University of Mainz, Germany
Email:  Dr....@t-..., ag...@mu...
WWW:    http://www.musikinformatik.uni-mainz.de/ag