[mercury-users] uniqueness

[Robert Ernst Johann JESCHOFNIK]

On Mon, 4 Oct 1999, Fergus Henderson wrote:

> Currently we don't support nested unique modes.

Damn.
Just out of curiosity, how far away are they, roughly?

> When we support nested unique modes, then the object that you deconstruct
> would need to have the following inst
> 
> 	:- inst uniq_f_matrix = unique(f_matrix(ground, ground, uniq_array(ground))).
> 
> in order for the array__set call to work.

Well, I spent a bit of time playing (fighting) with my code before I
received/read your reply, and I had managed to come up with the following
code:

:- inst uniq_f_matrix(I, J, K) = unique(f_matrix(I, J, K)).
:- inst uniq_f_matrix == uniq_f_matrix(ground, ground, uniq_array).

:- mode f_matrix_di == di(uniq_f_matrix).
:- mode f_matrix_uo == out(uniq_f_matrix).
:- mode f_matrix_ui == in(uniq_f_matrix).

:- type f_matrix --->
	f_matrix(
		 int,		% M
		 int,		% N
		 array(float)	% the elements of the matrix
		).

and the compiler does not give me any errors or warnings. Infact, I have a
predicate defined as follows:

:- pred f_matrix_set(int, int, float, f_matrix, f_matrix).
:- mode f_matrix_set(in, in, in, f_matrix_di, f_matrix_uo) is det.

f_matrix_set(I, J, Val, M_in, M_out) :-
	M_in = f_matrix(M, N, Array0),
	Location = I * N + J,
	array__set(Array0, Location, Val, Array),
	M_out = f_matrix(M, N, Array).

and it seems to be working. Before I got all the insts right, I was
getting uniqueness warnings here.

I wrote most of that inst code without having any idea really what I was
doing... It was more a case of copying style/syntax from declarations I
see in the library modules; but, having done it, I think I actually get
the drift of what is going on as I read back over it. (Although, I am
still not sure why the f_matrix(I, J, K) line is needed, and similarly
unsure why you can only use type variables in it.)

Am I correct in saying that the above code only works because of the
work-around in the array library, where uniq_array isn't really a unique
inst, but bound instead?


I think it is all slowly starting to fall into place...

thanks for your help, Fergus. especially since it is so late at night.
Rob
--------------------------------------------------------------------------
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
--------------------------------------------------------------------------