[Python-Dev] optimization required: .format() is much slower than %

[Simon Cross]

On Tue, May 27, 2008 at 12:43 AM, Christian Heimes <lists at cheimes.de> wrote:
> Ten minutes ago I raised a concern about speed differences between the
> old style % formatting and the new .format() code. Some quick
> benchmarking from Benjamin and me showed, that it's even worse than I
> expected.

My own tests show a less dire difference:

$ ./python -m timeit "'some text with %s'.format('nothing')"
1000000 loops, best of 3: 0.578 usec per loop
$ ./python -m timeit "'some text with %s' % 'nothing'"
1000000 loops, best of 3: 0.472 usec per loop

> $ ./python -m timeit "'%s'.format('nothing')"
> 100000 loops, best of 3: 2.63 usec per loop
> $ ./python -m timeit "'%s' % 'nothing'"
> 10000000 loops, best of 3: 0.188 usec per loop

It struct me as odd that this one case shows such a big difference
while the others show less of one.

My tests show that the old-style % formatting is much faster when the
final string is 20 characters or less:

$ ./python -m timeit "'....|....|....|...%s' % '12'"
10000000 loops, best of 3: 0.0764 usec per loop
$ ./python -m timeit "'....|....|....|...%s' % '123'"
1000000 loops, best of 3: 0.481 usec per loop

A read through stringobject.c didn't give me any clue as to why 20 is
the magic number. %d shows an even bigger jump:

$ ./python -m timeit "'....|....|....|...%d' % 12"
10000000 loops, best of 3: 0.0764 usec per loop
$ ./python -m timeit "'....|....|....|...%d' % 123"
1000000 loops, best of 3: 1.28 usec per loop

Schiavo
Simon