[Python-Dev] Possible resolution of generator expressionvariablecapture dilemma (original) (raw)

Jeremy Hylton jeremy at zope.com
Thu Mar 25 15:08:28 EST 2004

On Thu, 25 Mar 2004 02:34:10 -0500 "Raymond Hettinger" <python at rcn.com> wrote:

I do not view genexps as "convenient" short-forms for real generators. Instead, they represent memory/resource friendly versions of list comps which have been fabulously successful and have helped move us away from map/lambda forms. I believe they will make the language more fluid, cohesive, and expressive.

It would be great to revise some existing code to use generator expressions. I find it hard to judge their effect, but am inclined to agree with you. I like list comprehensions a lot.

In deciding on semantics, we would all rather trade away some power in exchange for greater clarity about what the expression will do without having to run it. Put another way, human interpretability is more important than supporting weird usages. With that criteria, early-binding is looking to be least nuanced approach.

If readability counts, I think the standard binding rules make the most sense. The examples I've seen that depend on differences in binding all seem like edge cases to me. I don't expect generator expressions in for loops will be a common usage. Put another way, in the cases where it matters, I don't expect either option to be ovious or intuitive. If there's no obvious solution, I'd rather see the consistent one.

Jeremy

More information about the Python-Dev mailing list