[Tutor] Deal a Random Deck - Challenge

[Zak Arntson]

Here's an interesting problem a coworker (former mentor) uses when hiring:
 * Write an algorithm to deal out a set of "cards"

It's pretty straightforward in C, and here's a port of the C code in Python:
###
import random

def deck(size):
    d = range(size)
    for i in range(size):
        r = random.randrange(i, size)
        temp = d[i]
        d[i] = d[r]
        d[r] = d[i]

    return d
###

You can replace the swap code, but weirdly it's slower to process.
###
import random

def deck2(size):
    d = range(size)
    for i in range(size):
        r = random.randrange(i, size)
        d[i], d[r] = d[r], d[i]

    return d
###

Then my coworker turned around with this solution, which turns out to be
the fastest of the three:
###
import random

def deck3(size):
    d = [(random.random(), i) for i in range(size)]
    d.sort()
    return [d[i][1] for i in range(size)]
###

So my challenge, then: Can anyone come up with an even faster solution
than option 3?

As an aside, I tried this with deck(100000) through the profiler and go
cumtimes as follows: deck = 5.126, deck2 = 5.693, deck3 = 3.327

---
Zak Arntson
www.harlekin-maus.com - Games - Lots of 'em