For the benefit of anyone who might believe e.g. that doing aset.add()
in a loop would have performance competitive with doing aset.update()
, here's an example of how you can test your beliefs quickly before going public:
>\python27\python -mtimeit -s"it=xrange(10000);a=set(xrange(100))" "a.update(it)"
1000 loops, best of 3: 294 usec per loop
>\python27\python -mtimeit -s"it=xrange(10000);a=set(xrange(100))" "for i in it:a.add(i)"
1000 loops, best of 3: 950 usec per loop
>\python27\python -mtimeit -s"it=xrange(10000);a=set(xrange(100))" "a |= set(it)"
1000 loops, best of 3: 458 usec per loop
>\python27\python -mtimeit -s"it=xrange(20000);a=set(xrange(100))" "a.update(it)"
1000 loops, best of 3: 598 usec per loop
>\python27\python -mtimeit -s"it=xrange(20000);a=set(xrange(100))" "for i in it:a.add(i)"
1000 loops, best of 3: 1.89 msec per loop
>\python27\python -mtimeit -s"it=xrange(20000);a=set(xrange(100))" "a |= set(it)"
1000 loops, best of 3: 891 usec per loop
Looks like the cost per item of the loop approach is over THREE times that of the update
approach.
Using |= set()
costs about 1.5x what update
does but half of what adding each individual item in a loop does.