[Python-Dev] PEP 218 update questions

[Raymond Hettinger]

> 1)
> The long-term proposal in PEP 218 lists a single built-in conversion
> function, set(iterable).  This made sense when there was a single Set
> class, but now this needs to be updated.  
> 
> How would instances of a built-in immutable set type be created?
> Would there be a second immutable_set() built-in, or would the set()
> function take an additional argument: set(iterable, immutable=True)?
> 

-1 on the property version.  As it stands now, immutables can be 
constructed directly from a class with its own methods or, more
likely, is an automagically created proxy used internally by
mutuable sets.  I think of immutable sets as regular sets that are
frozen for purposes of inclusion in another set -- there is almost
no reason for instantiatiating them directly (other than to save
coercion time for sets whose contents are known and fixed
at the time of instatiation).

> 2)
> The PEP proposes {1,2,3} as the set notation and {-} for the empty
> set.  Would there be different syntax for an immutable and a mutable
> set?  
> 
> My thought is, "no"; {1,2,3} or {-} is always mutable, and an
> immutable set would be written as immutable_set({1,2,3}).  That's up
> to python-dev, though, but whatever gets chosen needs to be specified
> in the PEP.

+1.  Immutable sets have their own weirdness.  No use reflecting
that in the core language syntax.  Besides, they may yet evolve 
further (perhaps with an animmutable.refresh() method).


Raymond Hettinger