[mercury-users] Safe use of uninitialised arrays

[Peter Schachte]

On Wed, Jan 19, 2000 at 10:23:59AM -0800, Ralph Becket wrote:
> Efficiency question.  It's often the case that one
> creates an array and then fills in every slot.
> However, the current array library insists that
> the programmer provides a value with which to
> initialise every element before delivering the new
> array.  Now I think this is good practice in general,
> but there are places where you just don't want the
> overhead.  In much the same way as one is prepared
> to take the consequences by compiling with array
> bounds checking disabled, I'd like to have something
> like
> 
> :- func array__unsafe_init_uninitialised(int) = array(T).

Of course, that'd have to be cc_multi.

> Alternatively, I'd be almost as happy with being able
> to supply an initialising predicate, e.g.
> 
> :- pred array__init_from_fn(
> 	int,				% array size
> 	pred(T2, T2, T1),		% `generating function'
> 	T2,				% `seed value'
> 	T2,				% `final value'
> 	array(T1)
> ).

This sounds reasonable.  You'd probably want the closure to take one
more argument, though: the array index.  Sure, you could keep track of
that yourself in the two arguments you're threading through your
closure, but since the init_from_fn predicate keeps track of this
already, why should you?

I wonder, though, whether the overhead of higher-order calls would
outweigh the savings from not redundantly intializing the array?

-- 
Peter Schachte                     I don't believe in collective guilt, but
mailto:schachte at cs.mu.OZ.AU        I do believe in collective responsibility.
http://www.cs.mu.oz.au/~schachte/      -- Audrey Hepburn 
PGP: finger schachte at 128.250.37.3  
--------------------------------------------------------------------------
mercury-users mailing list
post:  mercury-users at cs.mu.oz.au
administrative address: owner-mercury-users at cs.mu.oz.au
unsubscribe: Address: mercury-users-request at cs.mu.oz.au Message: unsubscribe
subscribe:   Address: mercury-users-request at cs.mu.oz.au Message: subscribe
--------------------------------------------------------------------------