On Tue, Jul 21, 2015 at 1:35 PM, Tim Peters <tim.peters@gmail.com> wrote:
[Alexander Belopolsky]
One thing I know for a fact is that the use of % formatting instead of .format makes a significant difference in my applications. This is not surprising given these timings:
$ python3 -mtimeit "'%d' % 2" 100000000 loops, best of 3: 0.00966 usec per loop $ python3 -mtimeit "'{}'.format(2)" 1000000 loops, best of 3: 0.216 usec per loop
Well, be sure to check what you're actually timing. Here under Python 3.4.3:
from dis import dis def f(): return "%d" % 2 dis(f) 2 0 LOAD_CONST 3 ('2') 3 RETURN_VALUE
That is, the peephole optimizer got rid of "%d" % 2 entirely, replacing it with the string constant "2". So, in all, it's more surprising that it takes so long to load a constant ;-)
Interesting that the same optimization can't be done on the .format() version - it's not as if anyone can monkey-patch str so it does something different, is it? To defeat the optimization, I tried this: rosuav@sikorsky:~$ python3 -mtimeit -s "x=2" "'%d' % 2" 100000000 loops, best of 3: 0.0156 usec per loop rosuav@sikorsky:~$ python3 -mtimeit -s "x=2" "'%d' % x" 10000000 loops, best of 3: 0.162 usec per loop rosuav@sikorsky:~$ python3 -mtimeit -s "x=2" "'{}'.format(2)" 1000000 loops, best of 3: 0.225 usec per loop rosuav@sikorsky:~$ python3 -mtimeit -s "x=2" "'{}'.format(x)" 1000000 loops, best of 3: 0.29 usec per loop The difference is still there, but it's become a lot less dramatic - about two to one. I think that's the honest difference between them, and that's not usually going to be enough to make any sort of significant difference. ChrisA