{{cd2>这样使用。在from itertools import dropwhile
np.fromiter(dropwhile(lambda x: x, ar[::-1]), dtype=bool)[::-1]
编辑
这是更快的方法。(只需使用itertools.takewhile)
^{pr2}$
时间:ar = np.array([True, False, True, True, True])
#mine
%timeit ar[:-sum(1 for i in takewhile(lambda x: x, reversed(ar)))]
1000000 loops, best of 3: 1.84 us per loop
#mine
%timeit np.fromiter(dropwhile(lambda x: x, ar[::-1]), dtype=bool)[::-1]
100000 loops, best of 3: 2.93 us per loop
#@Jaime
%timeit ar[np.bitwise_or.accumulate(~ar[::-1])[::-1]]
100000 loops, best of 3: 3.63 us per loop
#@askewchan
%timeit ar[:len(ar)-np.argmin(ar[::-1])]
100000 loops, best of 3: 6.24 us per loop
#@xbb
%timeit ar[:len(ar)-np.where(ar[::-1]==False)[0][0]] if np.where(ar[::-1]==False)[0].size>0 else ar[:0]
100000 loops, best of 3: 7.61 us per loop
附言:
这不是什么神奇的功能。在def no_magic_func(ar):
for i in xrange(ar.size-1, -1, -1):
if not ar[i]:
return ar[:i+1]
return ar[0:0]
时间:ar = np.array([True, False, True, True, True])
%timeit no_magic_func(ar)
1000000 loops, best of 3: 954 ns per loop