How is irange() any different from range() or xran

I was going through Python Generators Wiki when I came across this RangeGenerator page which talks about irange() -

This will let us iterator over large spans of numbers without resorting to xrange, which is a lazy list as opposed to a generator.

I can't seem to understand the test suite and the implementation described on that page. I know that range() creates a list in the memory (from Python 2.7 point of view) and xrange() is a generator. How is irange() any different?

回答1:

irange() returns a generator type, which can only be iterated over. Nothing else. Once you iterated over it, the generator is exhausted and can not be iterated over again.

The Python 2 xrange() type and Python 3 range() type are sequence types, they support various operations that other sequences support as well, such as reporting on their length, test for containment, and indexing:

>>> xr = xrange(10, 20, 3)
>>> len(xr)
4
>>> 10 in xr
True
>>> xr[0]
10
>>> xr[1]
13

You can iterate over these objects more than once:

>>> for i in xr:
...     print i,
... 
10 13 16 19
>>> for i in xr:
...     print i,
... 
10 13 16 19

You can even use the reversed() function to iterate over them in reverse, efficiently:

>>> for i in reversed(xr):
...     print i,
... 
19 16 13 10

The Python 3 range() type is an improved version of xrange(), in that it supports more sequence operations, is more efficient still, and can handle values beyond sys.maxint (what would be a long integer in Python 2).

It supports slicing, for example, which results in a new range() object for the sliced values:

>>> r = range(10, 20, 3)
>>> r[:2]
range(10, 16, 3)

You can use negative indices just like you can with other Python sequences, to get elements counting from the end:

>>> r[-2]
16
>>> r[-2:]
range(16, 22, 3)

and the type supports testing for equality; two range() instances are equal if they'd yield the same values:

>>> range(10, 20, 3) == range(10, 21, 3)
True

In Python 2, the only advantage the generator irange() might have is that it doesn't suffer from the limitation to non-long integers that xrange() is subjected to:

>>> import sys
>>> xrange(sys.maxint)
xrange(9223372036854775807)
>>> xrange(sys.maxint + 1)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
OverflowError: Python int too large to convert to C long

回答2:

irange provides an generator which does not loads entire result into memory. Say you have range(1, 1000000) in case a list of 1000000 numbers would be loaded into memory, whereas in case of xrange(1, 1000000), just one number would be in memory at a time.