Re: pointer elimination in C

Robin Popplestone <pop@dcs.gla.ac.uk> writes:
|> I would be interested to know what that language is - any serious language
|> -does- in effect provide pointers, but restricts the operations you can do
|> on them in the interests of hygene. LISP for example, provides almost
|> nothing but pointers - typically only short integers and possibly short
|> floats will not be pointers.


APL and J support recursive data structures this way:


      xyz =. <X
(denoted box or enclose)
where X is any array, produces a scalar which is a representation of
the array X. These scalars (which are implemented as pointers)
can be catenated, selected, etc, as any other array elements.


      >xyz
(denoted open or disclose)
produces an array which is the value of X as of the time that the
enclose was performed.


Thus X <--> ><X for all X.


Enclose gives you the ability to build tree-like data structures,
manipulate large array collections as simple entities, etc.


In the above example, changing X (assuming it was a named variable, rather
than a temp) would NOT affect the value pointed to by xyz. Techniques
such as static analysis or reference counts ensure that such side effects
can't occur.


The end result is that enclose/box produces an alias-free method of
pointer support. The only problem with this approach is that since it's
side-effect free, you cannot change the original X via a reference from
xyz, even if you want to.


If you could, we'd be back into the world of aliasing problems.


Bob


ps: Source and binaries for J are available by anonymous ftp from
watserv1.waterloo.edu (aka watserv1.uwaterloo.ca) in /languages/j.


Robert Bernecky rbe@soliton.com
Snake Island Research Inc (416) 203-0854
18 Fifth Street, Ward's Island
Toronto, Ontario M5J 2B9, Canada
--

Return to the comp.compilers page.
Search the comp.compilers archives again.