🚀 go-pugleaf

In article <BxIoDv.72J@mentor.cc.purdue.edu> hrubin@pop.stat.purdue.edu
(Herman Rubin) writes:
[various uses of pointers to functions, to arrays of pointers, etc.,
deleted]

	Indeed these things all have their uses, but they too are all
in Algol, and have been for 24 years now.  Two examples that may [or
may not!] gladden Dr Rubin's heart:

  a) a self-seeding PRNG:

	PROC REAL myrand := REAL: (myseed; myrand := myprng);

     ["myrand" is a pointer to a procedure that delivers a real.  It is
      initialised to point to a procedure that first does the seeding, then
      makes "myrand" point instead to the real PRNG.  The result of following
      the pointer and then obeying the procedure is delivered as the first
      PRN.  Uses of "myrand" after the first go straight to "myprng";  note
      that only one procedure is actually called, as "myrand" itself is a
      pointer, *not* a procedure.]

   b) a self-buffering geometric random-number generator:

	INT bufsize = 1024;
	[bufsize] REAL buffer;  INT bufp;
	[bufsize] PROC REAL geombuff;
	PROC REAL geom := geombuff[bufsize] :=
	   REAL: (fill buffer; geom := geombuff[1]; buffer[bufp := 1]);
	FOR i TO bufsize - 1
	   DO geombuff[i] := REAL: (geom := geombuff[bufp +:= 1]; buffer[bufp])
	   OD;
	PROC fill buffer = ( ... some code to generate the random numbers
		and fill buffer with them ...);

     ["geom" points to one of an array of (anonymous) procedures.  All but
      one of these simply returns an element of "buffer", having updated
      "geom";  the last instead fills the buffer and reinitialises the
      various pointers.  Here there is an implied context switch (but only
      one, except when the buffer is filled) each time "geom" is used, so
      we don't get the full benefits of the buffering unless a heavily-
      optimising compiler recognises that the procedure denotation inside
      the DO-loop can be in-lined.]

-- 
Andy Walker, Maths Dept., Nott'm Univ., UK.
anw@maths.nott.ac.uk