Nick Coghlan wrote: > C:\>python -m timeit -s "floats = map(float, range(1000))" "ints = > map(int, floa ts)" > 1000 loops, best of 3: 481 usec per loop > > C:\>python -m timeit -s "floats = map(float, range(1000))" "ints = [int(x) > for x in floats]" > 1000 loops, best of 3: 721 usec per loop > > C:\>python -m timeit -s "floats = map(float, range(1000))" "ints = []" > "for x in floats: ints.append(int(x))" > 1000 loops, best of 3: 992 usec per loop > > For builtin functions, map is usually the fastest option (and, IMO, the > most readable). List comprehensions should be preferred to map + lambda, > though. >
>From the "Evil Coder's Guide to Fast Code": $ py24 -m timeit -s "floats = map(float, range(1000))" "ints = map(int, floats)" 1000 loops, best of 3: 442 usec per loop $ py24 -m timeit -s "floats = map(float, range(1000))" -s"from itertools import starmap, izip" "ints = list(starmap(int, izip(floats)))" 1000 loops, best of 3: 343 usec per loop Raymond Hettinger must be doing something smart here that should be ported to the map() builtin. Peter -- http://mail.python.org/mailman/listinfo/python-list
