Re: GSL Error Handling

["Gregory R. Warnes" <warnes at biostat dot washington dot edu>]

On Mon, 17 Jul 2000, Gerard Jungman wrote:

  GJ>> The problem with library design in this area is not that
  GJ>> there is too much "side-band" interaction with the application,
  GJ>> but too little. Applications need to know what is going on
  GJ>> in a numerical implementation; they have a right to know.
  GJ>> The problem is arranging some form of communication channel
  GJ>> without breaking encapsulation. This is not an easy problem,
  GJ>> and archaic languages make it nearly impossible. So we
  GJ>> are stuck with "error codes" for now.

Yes, it is difficult to do this properly using "obsolete" languages (C,
Fortran, ..).  That's why I wonder about the choice to do GSL in C rather
than in C++ or Java, ;^) .  The try-catch mechanism in C++ (and Java) is
designed to allow a great deal of interaction in the case of errors
without hampering the efficiency and clarity of the normal code path.

  GJ>> 
  GJ>> 
  GJ>> -- 
  GJ>> G. Jungman
  GJ>>