[Python-Dev] Re: PEP 289 - Generator Expressions MoveForward (original) (raw)

Tim Peters tim.one at comcast.net
Fri Apr 30 17:28:28 EDT 2004

[Russell E. Owen]

I completely agree. Generator expressions and list comprehensions are similar enough that it seems a wart to have to carefully differentiate between two in normal use.

In normal use you won't have to distinguish, where "normal use" is pretty much defined by "cases where it doesn't matter" .

Personally, I'd only like to see generator expressions added to Python if they can allow list comprehensions to be deprecated -- to avoid language clutter.

Listcomps can be deprecated, although not all existing uses of listcomps will translate directly into genexps.

I think this has become a "practicality beats purity" feature. When Python had only 3 scopes, it was fine to ignore "explicit is better than implicit": the scope model was so bare that explictness would have been overkill. When multiple nested scopes got introduced, and function bodies start appearing in places other than lambda and def (which is the case for genexps), scope gets much more mysterious (and so also potentially surprising) in the absence of explicit scope declarations. But simple genexps will work fine anyway almost all the time, so it doesn't appear to matter much that devious uses will have nightmarish semantics.

One question, though: is "if" part of generator expressions or not?

Yes. If the PEP isn't clear about this, it's a flaw in the PEP. I don't think the semantics of genexps are clear in the PEP either (unless that's changed recently).

More information about the Python-Dev mailing list