Why I think range is a wart.

[Gonçalo Rodrigues]

On 13 Mar 2002 12:53:50 -0500, com-nospam at ccraig.org (Christopher A.
Craig) wrote:

>Gonçalo Rodrigues <op73418 at mail.telepac.pt> writes:
>
>> For starters, range (and xrange) is a perfectly sensible, reasonable and
>> useful built-in. But the biggest use I make of it is in iterating through
>> a list, e.g. as in
>> 
>> for index in range(len(mylist)):
>>     <whatever>
>> 
>> and in the <whatever> body I need to make use of the indexes. I am not
>> sufficiently qualified to say if this is Pythonic or not, but I find
>> myself doing this quite often.
>
>I don't understand how range is a reasonable and useful built-in if
>this use is not.  

There are other useful, reasonable and sensible uses for the range
built-in. My comment is that this particular use looks warty to me
because any sequence, especially if it obeys the "list-protocol" (that
is, it has methods like __getitem__, __setitem__, etc.) should be able
to generate the indexes (or keys) of its members. Just like a
dictionary.

>
>If you have a list [2, 3, 5, 7, 11] and you want to iterate through it
>you don't need range:
>
>for item in list: 
>  <work on items>
>
>If, on the other hand, you want to iterate though the indicies of said
>list rather than the items themselves you need to come up with some
>way to generate a zero based list of integers up to one less than the
>size of the list (i.e. a list of indicies).  When you create this list
>you aren't doubling information that the list already has.  A list
>does not contain a list of indicies into itself.  The list knows its
>elements, in order, and how many elements it has.  From this you can
>generate a list of indicies, but there are no 'keys' stored in a list
>object like there are in a dictionary.

This is where I beg to disagree. If the elements are in order, as they
are in a list, it is perfectly reasonable to ask, "you are a list,
right? then hand me the indexes of your members." If the list has the
keys in memory or not, that is an implementation issue.

Prolly the best way to implement this is to add a single method
(iteritems?) that gives an iterator through the indices.

>
>To go from the information that is stored in the list (an ordered list
>of elements, and the size of said list) to the information you need to
>iterate through indicies (an ordered sequence of indicies valid for the
>given list) you need to take the length of the list (N), and create a
>new list of N integers from 0 to N-1.  'range(len(mylist))' says
>precisely that.  I don't see how this is the least bit kludgey.
>
>.keys() and .items() methods on lists seem odd to me.  That is
>inforamtion that the list does not contain.  Sure it can easily be
>generated given information the list does contain, but why can't the
>programmer generate it himself from the same information?

Of course the programmer can do it for himself. But since the idiom
expounded in my post seems to be prevalent it would be at least
reasonable to make it a built-in.

Best regards,
Gonçalo Rodrigues