PEP 276 Simple Iterator for ints (fwd)

[Peter Hansen]

David Eppstein wrote:
> 
>  Peter Hansen <peter at engcorp.com> wrote:
> 
> > I agree that sounds like a worthy goal.  I disagree that your
> > suggestion is any more likely to be understood by anyone,
> > other than perhaps a mathematician.  [...]
> 
> From context, it seems you mean Ewing's suggestion rather than mine, right?

Yes, quite right.  Sorry.  :)

> > Maybe we should look back to BASIC (for i = 1 to 5: next)
> > or one of the suggestions for something which looks more
> > explicitly like a list of items (for i in [0..5]) or
> > (for i in [0, 1, .. 5] ).  Especially this latter would
> > definitely be understandable to anyone, IMHO.
> 
> Your final option is very close to my suggestion, much earlier in this
> thread, which seems to be disliked by many due to its requiring closed
> intervals while other Python features use half-open ones.

I was skimming the thread by that point, but I can say only
that I don't see that there needs to be consistency between
these two things.  range(5) is a function with a definition
which does what it does.  It could just as well return 
true or false, if that's what the docs said it did.

[0, 1, ... 5 ] looks like a list and should act like a list.
In fact, it should *produce* a list, or in the new world 
an iterator.  That list should "clearly" contain both 0
and 5.  Anything else would be surprising.

I suppose it's just the consistency argument that detractors
dislike, but maybe a second look at it would convince them
it's not important.

The suggestions of doing [0, 1, ... 5 ) are, however sheer
insanity :-).  Obscure (to all but mathematicians), error
prone, unbalanced braces, rarely needed, etc...

If we just wanted to have another way to type range() it
would be a waste.  If we are trying to improve the 
readability of the language (especially for newbies)
the [0, 1, ... 5] option (or something close) would be
a good idea.

-- 
----------------------
Peter Hansen, P.Eng.
peter at engcorp.com