[Nickle] Reference semantics are confusing in Nickle

[Keith Packard]

Around 15 o'clock on Nov 27, "James LaMar" wrote:

> > &int i = reference(5)
> &5

You can also just use

	> &int i = &5;

The box operator '&' works fine on rvalues.

> ->     j[5] = reference (5);
> <stdin>:4: Incompatible types, left 'int', right '&poly', for =
> operation
> 
> Why is this ill-typed?

Remember that references are just syntactic sugar for pointers.  In this 
case, j[5] doesn't refer to the reference at j[5], but rather the value 
pointed at by that reference.  Think of any reference expression as the 
equivalent pointer expression with a '*' prepended, and any reference 
expression with '&' as the bare pointer expression.

	int i;
	int j;
	&int r = &i;
	*int p = &i;

	r = 5		 	<=>	*p = 5
	&r = &j;		<=>	p = &j;

So, to store into the array of references, you do:

	&j[5] = &3;

(I note here the pleasing symmetry of the box operator...)

> > j[5] = 5
> Unhandled exception uninitialized_value ("uninitialized hash element")
> <stdin>:5:     j[5] = 5;
> 
> Why doesn't this work?

Same reason; j[5] dereferences the value at j[5], which hasn't been 
initialized yet.

	> &j[5] = &0
	> j[5] = 5

works just fine.

The examples with arrays appear identical to the above discussion with 
hashes.

-keith


-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 228 bytes
Desc: not available
Url : /pipermail/nickle/attachments/20041128/c1607857/attachment.pgp