Meta: PEP discussion (was Re: PEP 255: Simple Generators)

[Tim Peters]

[Greg Ewing]
> In Icon, there isn't any implementation *to* expose.

Of course there is -- Griswold wrote a whole book about it.

> There are no generator-iterators in Icon; the closest equivalent
> is simply the activation record of a procedure.

Some implementations of Icon work that way, but they've used several
approaches over the years.  In no case did they expose any of the internals,
though.  We didn't *have* to expose Python's generator-iterator object
either, but doing so adds (IMO) significant power over Icon's generators --
and it's more Pythonic to do so anyway.