Re: Passing procedures as parameters to procedures.

[cis.ohio-state.edu!magnus.acs.ohio-state.edu!zaphod.mps.ohio-state.edu!ho]

What amuses in this discussion, is the sub-thread about the
storage management penalties (allegedly?) imposed by procedures-
as-parameters. The amusement stems from the fact that:

********Ada(83) already requires that valid implementations pay
********an even greater penalty than that required by mere
********provedures-as-parameters.

To cut a long story short, read ``Full Functional Programming
in a Declarative Ada Dialect'', in the proceedings of the 1992
TRI-Ada conference (pp. 350-358), by P. Bailes (that's me), D.
Johnston, E. Salzman and L. Wang. The really keen can contact me
for a copy of University of Queensland Department of Computer
Science Technical Report 225 ``First Class Functions for Ada''.

The idea is that function-*valued*-functions, like in functional
programming languages (and not just procedures/functions as
parameters to procedures/functions) are avaialble in Ada through
direct (ie non-interpretive) translation into a complex of tasks.
In other words, Ada tasks require Ada implementations already to
be equipped with all the (heap etc.) storage management facilities
needed to implement a language construct much more ambitious than
this thread has so far contemplated.

The interesting question should therefore be why 9X doesn't have
function-valued-functions - sure, there are lots of sociological
etc. issues that surround a language design, but let's keep this
on a purely technical, almost Platonic, plane. (I've had a few words
with the 9X gurus at TRI-Ada about this, but to my mind inconclusively.)
--
Paul A Bailes
paul@cs.uq.oz.au