parameter checking

[Brian Gough]

Nigel Wetters writes:
 > I guess this is a question about coding style. For the past few
 > days, I've been stupidly passing zero values to construct gamma
 > distributrions, when the library assumed it would never be passed a
 > zero value. I'm obviously at fault here, but I would have found the
 > problem a lot quicker with a bit of help from the libraries.
 > 
 > If a function expects parameters within certain bounds, should
 > those parameters be checked, or is it the user's responsibility to
 > check bounds before calling a function?

There is currently no error checking on the pdf functions -- see
randist/TODO.  When a function returns a double there is an issue of
how to signal an error.  The only option in this case is to return NaN
for invalid parameters.

-- 
Brian Gough

----------------------------------------------------------------------
Network Theory Ltd           Phone: +44 117 3179309 (UK: 0117 3179309)
15 Royal Park                  Fax: +44 117 9048108 (UK: 0117 9048108)
Bristol BS8 3AL                WWW: http://www.network-theory.co.uk/
United Kingdom               Email: bjg@network-theory.co.uk     
----------------------------------------------------------------------