PEP 309 - Built-in closure type (with tentative syntax proposal)

[Peter Harris]

Alex Martelli <aleax at aleax.it> wrote in message news:<XwV1a.217687$AA2.8558127 at news2.tin.it>...
> 
> It wouldn't be a bad idea, except for the name (which should
> be curry, not 'closure').  The type should live in a standard
> library module -- it's not going to be used often enough to
> make it a built-in.
> 
OK, I'll change the PEP to suggest 'curry' as the preferred name
since enough people have set me straight on that one.

Curried functions are a bit like list comprehensions in that a
beginner won't immediately see the point, but once you find out
what they are good for, you wonder how you did without them.

If you get them for free as a built-in, they will be used more
than if you have to code them yourself.  How much is often enough?
Who knows?  But I don't see the point of another standard library
module with only one type in it, just so I have to do
  import curry from curry
at the top of most of my non-trivial programs.

 
> > Is the @ syntax too ugly to be worth it?
> 
> It's ugly, unpythonic, and absolutely unwarranted besides:

OK it's ugly. It's unpythonic too in that Python doesn't do it
at the moment. But no other language does either, so if Python
did, the @ would be de facto *uniquely* Pythonic :)

However we have hash-table literals, tuple literals and list
literals all neatly differentiated by punctuation rather than
named constructors.  It's probably not too big a stretch to
have functional-section literals too, if they were going to be
useful enough.

Apparently everyone thinks they would not, and I am not pushing
for it. It was probably not such a good idea.

> 
> Alex

Peter Harris