Re: OT: A bit of Sudoku

["Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de>]

On Fri, 06 Jun 2014 11:21:51 +0200, Georg Bauhaus wrote:

> On 06/06/14 09:51, Dmitry A. Kazakov wrote:
>> Or reduce overhead
> 
> Will there be less overhead if programmers can announce
> some controlled jump, back to where "it" came from?

A model of error handling based on returning back to the point of raising
was in PL/1 (on-error stuff). It proved to be wrong, because the whole idea
of exception is that you cannot go back because you cannot and will not
handle it where you detected it.

Regarding overhead, it is stack winding which takes most time. This
overhead is necessary per the logic of exception processing = leaving all
contexts compromised by the unexpected/undefined/undesired state. You must
finish these contexts regardless exception propagation or if-then-goto.

And testing/checking state is sometimes far more expensive that handling
it. Which is why exceptions are often less overhead than doing looking
ahead for an unanticipated condition, like EOF. Or an extreme example:

   if Check_Syntax (Ada_Program) then
      Compile (Ada_Program);
   else
      Put_Line ("I won't compile that mess!");
   end if;

If you moving forward checking each step you made, you are much slower than
when walking normally.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de