Re: Intervention needed?

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

"Optikos" <optikos@verizon.net> wrote in message 
news:aa13d9dd-69e9-48ac-99d0-3fabef216781@googlegroups.com...
...
> A container could have (as a required fundamental axiom) a singly-linked
>list of spirits of all cursors ever created that reference it.

Matt suggested such an implementation when the containers were designed. But 
only for deep testing, because...

>As each cursor's life ends, that dying cursor directly knows O(1) where its
>spirit is within that linked list, and then removes its dying spirit too.

This means that each cursor has to be controlled (explicitly or with 
something that is effectively the same), and that means that cursor 
operations are going to be much more expensive. Particularly operations that 
return cursors from functions. (Those are probably 10x slower.) And since 
iterations are built on a bunch of such functions, it too would be much 
slower.

I don't think anyone would accept the Ada containers if they are 
intentionally that slow by design. (The place where that was true in Ada 
2012 has spawned a lot of revision in Ada 2020, it would be madness to put 
it back somewhere else.)

                                                 Randy.