Nick Coghlan wrote:
skip@pobox.com wrote:
Nick> $ ./python -m timeit "'' % ()" Nick> 1000000 loops, best of 3: 0.389 usec per loop
vs.
Nick> $ ./python -m timeit "'%s' % 'nothing'" Nick> 10000000 loops, best of 3: 0.0736 usec per loop
I think you need to use a tuple for the second case to make it comparable to the first.
It doesn't actually make that big a difference - I'm guessing a Py_INCREF shortcut ends up getting used either way:
$ ./python -m timeit "'%s' % 'nothing'" 10000000 loops, best of 3: 0.0848 usec per loop $ ./python -m timeit "'%s' % 'nothing'," 10000000 loops, best of 3: 0.133 usec per loop
Technically there you aren't using a tuple to provide the substitution argument, you are doing the substitution from a single string then putting the result in a tuple:
'%s' % 'nothing', ('nothing',)
On the basis of evidence like this: sholden@lifeboy ~ $ python -m timeit "'%s' % 'nothing'" 10000000 loops, best of 3: 0.0705 usec per loop sholden@lifeboy ~ $ python -m timeit "'%s' % ('nothing',)" 1000000 loops, best of 3: 0.691 usec per loop I'd say not using a tuple was the way to go if ever you needed to optimize the code for speed.
$ ./python -m timeit "'' % ()" 1000000 loops, best of 3: 0.513 usec per loop
If you want an even stranger result, it appears to take the interpreter longer to substitute nothing from a tuple than it does to substitute the empty string from a string: sholden@lifeboy ~ $ python -m timeit "'' % ()" 1000000 loops, best of 3: 0.454 usec per loop sholden@lifeboy ~ $ python -m timeit "'%s' % ''" 10000000 loops, best of 3: 0.0715 usec per loop
(times are a bit variable at this very moment since I have a few different apps open)
Me too. Timings are not benchmarks, I am not a lawyer, etc. Our machines seem to be of comparable speed. regards Steve -- Steve Holden +1 571 484 6266 +1 800 494 3119 Holden Web LLC http://www.holdenweb.com/