itertools.compress
(2.7 / 3.1中的新功能)很好地支持了像这样的用例,尤其是当与结合使用时itertools.cycle
:
from itertools import cycle, compress
seq = range(100)
criteria = cycle([True]*10 + [False]*20) # Use whatever pattern you like
>>> list(compress(seq, criteria))
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99]
Python 2.7计时(相对于Sven的显式列表理解):
$ ./python -m timeit -s "a = range(100)" "[x for start in range(0, len(a), 30) for x in a[start:start+10]]"
100000 loops, best of 3: 4.96 usec per loop
$ ./python -m timeit -s "from itertools import cycle, compress" -s "a = range(100)" -s "criteria = cycle([True]*10 + [False]*20)" "list(compress(a, criteria))"
100000 loops, best of 3: 4.76 usec per loop
Python 3.2计时(也相对于Sven的显式列表理解):
$ ./python -m timeit -s "a = range(100)" "[x for start in range(0, len(a), 30) for x in a[start:start+10]]"
100000 loops, best of 3: 7.41 usec per loop
$ ./python -m timeit -s "from itertools import cycle, compress" -s "a = range(100)" -s "criteria = cycle([True]*10 + [False]*20)" "list(compress(a, criteria))"
100000 loops, best of 3: 4.78 usec per loop
可以看出,相对于2.7中的内联列表理解而言,它没有太大的区别,但是通过避免隐式嵌套范围的开销,在3.2中有很大帮助。
如果目标是遍历结果序列而不是将其转化为完全实现的列表,则在2.7中也可以看到类似的区别:
$ ./python -m timeit -s "a = range(100)" "for x in (x for start in range(0, len(a), 30) for x in a[start:start+10]): pass"
100000 loops, best of 3: 6.82 usec per loop
$ ./python -m timeit -s "from itertools import cycle, compress" -s "a = range(100)" -s "criteria = cycle([True]*10 + [False]*20)" "for x in compress(a, criteria): pass"
100000 loops, best of 3: 3.61 usec per loop
对于特别长的模式,可以用类似chain(repeat(True, 10), repeat(False, 20))
这样的表达式替换模式表达式中的列表,这样就不必在内存中完全创建它。