PEP: 281 Title: Loop Counter Iteration with range and xrange Version: $Revision$ Last-Modified: $Date$ Author: magnus@hetland.org (Magnus Lie Hetland) Status: Rejected Type: Standards Track Created: 11-Feb-2002 Python-Version: 2.3 Post-History: Abstract This PEP describes yet another way of exposing the loop counter in for-loops. It basically proposes that the functionality of the function indices() from PEP 212 [1] be included in the existing functions range() and xrange(). Pronouncement In commenting on PEP 279's enumerate() function, this PEP's author offered, "I'm quite happy to have it make PEP 281 obsolete." Subsequently, PEP 279 was accepted into Python 2.3. On 17 June 2005, the BDFL concurred with it being obsolete and hereby rejected the PEP. For the record, he found some of the examples to somewhat jarring in appearance: >>> range(range(5), range(10), range(2)) [5, 7, 9] Motivation It is often desirable to loop over the indices of a sequence. PEP 212 describes several ways of doing this, including adding a built-in function called indices, conceptually defined as def indices(sequence): return range(len(sequence)) On the assumption that adding functionality to an existing built-in function may be less intrusive than adding a new built-in function, this PEP proposes adding this functionality to the existing functions range() and xrange(). Specification It is proposed that all three arguments to the built-in functions range() and xrange() are allowed to be objects with a length (i.e. objects implementing the __len__ method). If an argument cannot be interpreted as an integer (i.e. it has no __int__ method), its length will be used instead. Examples: >>> range(range(10)) [0, 1, 2, 3, 4, 5, 6, 7, 8, 9] >>> range(range(5), range(10)) [5, 6, 7, 8, 9] >>> range(range(5), range(10), range(2)) [5, 7, 9] >>> list(xrange(range(10))) [0, 1, 2, 3, 4, 5, 6, 7, 8, 9] >>> list(xrange(xrange(10))) [0, 1, 2, 3, 4, 5, 6, 7, 8, 9] # Number the lines of a file: lines = file.readlines() for num in range(lines): print num, lines[num] Alternatives A natural alternative to the above specification is allowing xrange() to access its arguments in a lazy manner. Thus, instead of using their length explicitly, xrange can return one index for each element of the stop argument until the end is reached. A similar lazy treatment makes little sense for the start and step arguments since their length must be calculated before iteration can begin. (Actually, the length of the step argument isn't needed until the second element is returned.) A pseudo-implementation (using only the stop argument, and assuming that it is iterable) is: def xrange(stop): i = 0 for x in stop: yield i i += 1 Testing whether to use int() or lazy iteration could be done by checking for an __iter__ attribute. (This example assumes the presence of generators, but could easily have been implemented as a plain iterator object.) It may be questionable whether this feature is truly useful, since one would not be able to access the elements of the iterable object inside the for loop through indexing. Example: # Printing the numbers of the lines of a file: for num in range(file): print num # The line itself is not accessible A more controversial alternative (to deal with this) would be to let range() behave like the function irange() of PEP 212 when supplied with a sequence. Example: >>> range(5) [0, 1, 2, 3, 4] >>> range('abcde') [(0, 'a'), (1, 'b'), (2, 'c'), (3, 'd'), (4, 'e')] Backwards Compatibility The proposal could cause backwards incompatibilities if arguments are used which implement both __int__ and __len__ (or __iter__ in the case of lazy iteration with xrange). The author does not believe that this is a significant problem. References and Footnotes [1] PEP 212, Loop Counter Iteration http://www.python.org/peps/pep-0212.html Copyright This document has been placed in the public domain. Local Variables: mode: indented-text indent-tabs-mode: nil fill-column: 70 End: