Re: Intervention needed?

["Randy Brukardt" <randy@rrsoftware.com>]

"Jere" <jhb.chat@gmail.com> wrote in message 
news:5dbd0a83-2d28-4c39-a973-2f0365d93846@googlegroups.com...
> On Tuesday, March 19, 2019 at 6:26:51 PM UTC-4, Paul Rubin wrote:
>> "Randy Brukardt" writes:
>> > Rust...  You can't build a cursor that way (nor a doubly-linked
>> > list). Perhaps the Rust people have gone beyond simple borrow
>> > semantics -- I don't know.
>>
>> I did a little web searching and it does seem to me that as of a year or
>> so ago, the way to make doubly linked lists in Rust was to call some
>> kind of unchecked library.  I'd like to know how it's done in ATS, and
>> I might ask on a Rust chat if it's still like that in Rust.
>>
>> I think it might be a serious issue for Rust, since one of Rust's
>> driving applications was web browsers.  Those build DOM trees where
>> nodes point to both their children and their parent, so the tree is full
>> of cycles.
>
> If you need either doubly linked lists or maps in Rust, you use the
> ones provided by the standard library (std::collections::list, for
> example, is the doubly linked list).  It's not unchecked at all.

But then why do you need pointers at all? The only cases where they are 
interesting is when they are heavily aliased; otherwise, you're better off 
using containers from the standard library or even your own types (as in Ada 
discriminated records). And it's that sort of aliasing that Rust is aiming 
to eliminate.

I'm sure that there is *some* example (there always is some example :-), but 
hardly enough to make something a programming language feature.

What I'm concerned about primarily is the ability to build ADTs that work 
like the Ada containers. If you have to do a bunch of unchecked stuff to do 
that, then they aren't safe at all. And one can't expect the 
implementation's library to provide *everything* for you.

                                                      Randy.