Looping over lists

[Alex Martelli]

Dustan <DustanGroups at gmail.com> wrote:

> On May 5, 3:15 am, kaens <apatheticagnos... at gmail.com> wrote:
> > I think the for i in range() is more readable (Maybe because I'm
> > coming from a c-style syntax language background) -  but what would
> > the benefits of using enumerate be (other that being more . . .
> > pythonesque?)
> 
> It doesn't create a whole new list just for iterating.

As the example list was of length 5, that's not all that important in
this case.  In cases where it _is_ crucial, you can use xrange.

The performance of the various ways of looping is substantially the
same:

$ python -mtimeit -s'n=5; a=n*[23]' 'for i in range(n): x=a[i]'
1000000 loops, best of 3: 1.4 usec per loop
$ python -mtimeit -s'n=5; a=n*[23]' 'for i in xrange(n): x=a[i]'
1000000 loops, best of 3: 1.18 usec per loop
$ python -mtimeit -s'n=5; a=n*[23]' 'for i,v in enumerate(a): x=v'
1000000 loops, best of 3: 1.49 usec per loop
$ 

Here, xrange is minutely faster and enumerate slower, but each speed
difference "in the noise".  Focusing on clarity is thus well warranted.

> > > > for i in range(n):
> > > >     for j in range(i+1, n):
> > > >         print a[i], a[j]
> >
> > >         Ah, but wouldn't the cleaner Python be something like
> >
> > > >>> a = [1, 2, 3, 4, 5, 3, 6]   #just to confuse matters
> > > >>> for pos, val in enumerate(a):
> > > ...     for v2 in a[pos+1:]:
> > > ...             print val, v2

This "breaks symmetry", by using enumerate in the outer loop and a slice
in the inner loop; the symmetrical construction of using range in both
loops is a big conceptual/clarity win -- the reader of the code needs to
"grasp" one fewer concept (locally).  Using xrange in both loops would
be just as good from this POV.


Alex