[Python-3000] sets in P3K?

[Jim Jewett]

On 4/28/06, Greg Wilson <gvwilson at cs.utoronto.ca> wrote:
> - introduce '@' as a prefix operator meaning 'freeze'.

I mostly like the concept, but not the operator.  Using @ limits it to
syntax-created literals (because of ambiguity with decorators).  Since
numbers are already immutable, that basically means lists, dicts, and
maybe sets, but not even user-defined subclasses.  (And you don't have
to do it with tuples, because ... uh ...)  I'm not sure what to use
instead, except maybe for the backwards quote, which means I have to
duck now too.  ("frozen" or "snapshot" might be too long to be
useful.)

I am slightly concerned that it *might* be a bad idea to encourage
immutable containers, in the same way that it can be a bad idea to
prevent subclassing or even to use a __slots__ field.

> - [1, 2, 3] is a mutable sequence (our old friend the list)
>
> - {1, 2, 3} is a mutable set
>
> - @{1, 2, 3} is an immutable set (i.e., a frozen set --- frozen
>   at creation, so all current optimizations possible)
>
> - @[1, 2, 3] is an immutable sequence (i.e., a tuple)

Are there still plans to push the claim that lists hold a homogenous
collection, and tuples don't?  A frozen list (and a mutable tuple)
would make that claim easier to defend, but it still wouldn't be easy.
 I'm inclined to go with real objects and named attributes...

> Now, if only I can figure out whether "@x" means "bind the name x
> permanently to its current value to create a symbolic constant" or "freeze
> the object to which x currently refers"... ;-)

If it is really limited to literals, then I'm not sure it matters; the
freeze should have the only reference.  Unless you were thinking of
supporting horrible things like

    x=@[1,2,3]
    y=x
    y.append(4)   # a different *name*, so we can now treat it mutably

-jJ