🚀 go-pugleaf

In article <520@ivax.doc.ic.ac.uk> dcw@doc.ic.ac.uk (Duncan C White) writes:
>I'm afraid this is incorrect, Paul: in Modula-2, POINTER TO CHAR is precisely
>what it says: a pointer to a SINGLE character.  Not a pointer to an array of
>characters, or anthing else...

>[suggests using ALLOCATE(pointerToChar, stringSize)]
>However, doing such an operation makes little sense, since you can only
>use

   [pointerToChar^]	

>which is the first character of that block of 10 characters... the remaining
>nine characters can only be accessed by exceedingly messy type conversions,
>which involve type casting the pointer into a pointer to some other type [an
>array or record] and then accessing the fields or elements through the
>fiddled pointer...

The n'th byte is easily extracted by:

  cp := ADDRESS(pointerToChar) + n - 1;
  cp^;

And for iterating over the string:

  VAR cp: POINTER TO CHAR; endCp: ADDRESS;
  ...
  cp := pointerToChar;
  endCp := ADDRESS(cp) + stringSize;
  WHILE ADDRESS(cp) <= endCp DO
    process(cp^);
    cp := ADDRESS(cp) + 1;
  END;


This should actually be faster than using array indices, because
you'll get machine code such as "MOVE.B (An),<destination>",
instead of "MOVE.B 0(An,Dn)" (using m680x0 code as an example),
and a good optimizer should be able to discover the
"MOVE.B (An)+,<destination>" optimization (combining the access 
with the pointer increment).

Alan "String.Compare, when recoded to use this technique, ran more
      than twice as fast on my system" Lovejoy