Lists, Sets, and Associations

[Mark S. Miller]

At 09:36 AM 10/13/98 -0800, Bill Frantz wrote:
>I would prefer a syntax which permitted several different implementations
>of these abstractions.  Lists can be implemented with arrays or linked
>lists for example.  Different uses will find that one implementation is
>better than another.

Feel free to suggest one.

E's use-syntax already has this property -- this is why operators simply
expand to messages.  E's construction syntax invokes one specific
construction function, and this same issue appears in literals.  For
example, in any language with multiple numeric data types, the literal "3"
evaluates to an instance of one of these (with Java & E making different
choices).  However, I find 

	altList([1, 2, 3])

an adequately pleasant construction expression for alternate implementations.

Finally, E's pattern syntax is defined to (but not yet implemented to)
interact with the specimen only by messages, so a tuple-pattern should be
perfectly happy to match against any object that behaves as a tuple.


	Asyncronously, more, in non-deterministic order, later,
	--MarkM