PEP 276 Simple Iterator for ints

[David Eppstein]

In article <3BF40F8A.378CC6F8 at ccvcorp.com>,
 Jeff Shannon <jeff at ccvcorp.com> wrote:

> Which is exactly why I think that the Haskell-ish syntax is a bad idea.  This
> syntax cries out to be a closed interval, and it would be very confusing if
> used as a half-open interval.  But that would make it the *only* case of
> Python using a closed interval.  I find that inconsistency to be *far* more
> jarring than the use of range().

I don't buy this argument.  Currently AFAIK the *only* cases of Python 
using intervals at all are range() and xrange(), which I've argued are 
unsatisfactory -- what's the point of changing to something better if you 
require it to be consistent with the bad solution currently in place?
Don't tell me it's for consistency with C, either -- in C, closed and open 
intervals are equally easy.

Let me see if I can list some likely uses of ranges:

- Loops over the elements of a sequence. Better handled by "for item in 
sequence" or, as someone recently suggested, "for index in 
indices(sequence)".

- Numbering lines of output.  Most commonly done from 1..n instead of from 
0..n-1.

- Functional programming e.g. "def factorial(n): 
reduce(multiply,range(1,n+1))" vs "reduce(multiply,[1,2,...n])".  A closed 
interval would be more natural.

- Loops where you want to execute a given number of times but you don't 
actually use the index within the loop.  "for index in range(n)" and "for 
index in [1, 2, ... n]" are equally natural.

- Dynamic programming type algorithms as in the lecture notes I cited 
earlier.  It's not uncommon to see loops over ranges like [0, 1, 2, ... n], 
for instance I think this is the natural way to index the longest common 
subsequence dynamic program.  It's also not uncommon for some of the loop 
indices to run backwards.  These are very unnatural to express in half-open 
format -- e.g. the interval above is range(0,n+1,1) or shorter range(n+1) 
while the backwards one is range(n,-1,-1) -- quick, which -1 is the 
increment and which is the end of the interval?

Looking through the Python code I've worked with recently, I found only one 
example of range, in code by Thierry Bousch for reading EXIF data:
    for i in range(entries):
        do something with (ifd + 2 + 12*i)
It seems to me equally straightforward to do this as a 1-based closed 
interval:
    for i in [1, 2, ... entries]:
        do something with (ifd + 12*i - 10)
-- 
David Eppstein       UC Irvine Dept. of Information & Computer Science
eppstein at ics.uci.edu http://www.ics.uci.edu/~eppstein/