[sltkr]

Nitpicking, but this statement doesn't work with large numbers (which xrange() is supposed to handle correctly):

    self._len = int(ceil(float(stop - start) / step))

Python supports arbitrary-length integers; you can't just cast those to (fixed-length) floating point numbers without losing precision. It's better to use integer division here, for example:

    self._len = (stop - start)//step + bool((stop - start)%step)

(A variant like (stop - start + step - 1)//step works only for positive numbers; I guess it could work if you put it in the earlier if-clause and put a corresponding assignment with +1 at the end in the other branch.)

[dcrosta]

Great catch. If you want to submit a pull request on github you can get the credit -- github.com/dcrosta/xrange

[graue]

A related nitpick: since Python's integers are arbitrary size, technically operations like - and / are not constant time, so your claim of an implementation "with constant-time and constant-space operations" is not true.

However, I understand if that complaint is just too nitpicky for you to want to mention it. Great post btw.

[sltkr]

Thanks for the consideration, but I don't really use github, so you go ahead and take the credit yourself.

[dcrosta]

OK, thanks. Updated in github and on the blog.

[dcrosta]

Huh -- CPython 2.x doesn't let you create an xrange with values past 263-1. CPython 3.x does.

[sltkr]

Even 63-bit integers aren't (all) representable in IEEE double floating point values (that Python uses) which have a 53 bits mantissa.

For example, int(float(10¹⁸ - 1)) != 10¹⁸ - 1, but xrange(1, 10¹⁸) is perfectly valid (even in Python 2).

edit: how do I type two consecutive asterisks on Hacker News? Backslash doesn't seem to work as an escape character.

[sp332]

If you put two spaces at the beginning of a line, you'll get a monospaced "literal" mode.

  For example, int(float(10**18 - 1)) != 10**18 - 1,
  but xrange(1, 10**18) is perfectly valid (even in Python 2).

[heretohelp]

\\\\

Edit: failure. That's two stymied people :(

[ori_b]

You can also cast to long, which doesn't lose precision.