[bitc-dev] Primary string type

[Jonathan S. Shapiro]

On Fri, 2006-05-19 at 00:01 -0400, Swaroop Sridhar wrote:
> Jonathan S. Shapiro wrote:
> > On Thu, 2006-05-18 at 23:38 -0400, Swaroop Sridhar wrote:
> > 
> >>I just noticed that the string type is not listed in
> >>Section 3.2 Primary Types. This is probably because we initially defined 
> >>string as (vector char) but subsequently changed over to a primitive 
> >>string type.
> > 
> > 
> > String isn't a primary type. It is a sequence type.
> > 
> > That having been said, I agree that it needs to be added. However, what
> > to add is not straightforward. We need first to discuss string literals
> > and their relationship to mutability.
> 
> If I write:
> 
> (define x:(mutable string) "abcd")
> 
> and then if we say:
> 
> (set! x "efgh")
> 
> x now points to a different string. x was mutated, not the string. In 
> the same sense,
> 
> (define y "efgh")
> 
> (set! x y)
> 
> x and y can point to (since strings cannot be unboxed) the same string.
> 
> Also, the total memory consumption in this system is constant because 
> once the string literals are created, copies only need to point to 
> existing strings. Am I missing something?

Possibly yes, because there is a question about whether a string literal
is a shorthand for an application, in which case copies may not point to
existing strings. See my note coming out momentarily.

> However, by string-set!, do you mean that parts of the string can be 
> overwritten?

If strings are sequences of mutable characters, then yes. The (mutable
string) test isn't relevant to the question. Your later vector test *is*
relevant.

> If strings are a sequence type, is there a plan to allow indexing into 
> strings?

Definitely yes.


shap