我已经测试了几种方法(包括使用shadow wranger建议的truth()函数)。

%timeit  (not a) ^  (not b)   # 47 ns
%timeit  (not a) != (not b)   # 44.7 ns
%timeit truth(a) != truth(b)  # 116 ns
%timeit  bool(a) != bool(b)   # 190 ns

Exclusive Or的定义如下

def xor( a, b ):
    return (a or b) and not (a and b)

这里建议的一些实现在某些情况下会导致操作数的重复求值，这可能会导致意想不到的副作用，因此必须避免。

也就是说，一个返回True或False的xor实现相当简单;如果可能的话，返回其中一个操作数的方法要棘手得多，因为对于应该选择哪个操作数没有共识，特别是当有两个以上的操作数时。例如，xor(None， -1， []， True)是否应该返回None，[]或False?我敢打赌，每个答案对某些人来说都是最直观的。

对于True-或false -结果，有多达五种可能的选择:返回第一个操作数(如果它在value中匹配最终结果，否则布尔值)，返回第一个匹配(如果至少存在一个，否则布尔值)，返回最后一个操作数(如果…Else…)，返回最后一次匹配(if…Else…)，或者总是返回boolean。总共是5 ** 2 = 25种xor口味。

def xor(*operands, falsechoice = -2, truechoice = -2):
  """A single-evaluation, multi-operand, full-choice xor implementation
  falsechoice, truechoice: 0 = always bool, +/-1 = first/last operand, +/-2 = first/last match"""
  if not operands:
    raise TypeError('at least one operand expected')
  choices = [falsechoice, truechoice]
  matches = {}
  result = False
  first = True
  value = choice = None
  # avoid using index or slice since operands may be an infinite iterator
  for operand in operands:
    # evaluate each operand once only so as to avoid unintended side effects
    value = bool(operand)
    # the actual xor operation
    result ^= value
    # choice for the current operand, which may or may not match end result
    choice = choices[value]
    # if choice is last match;
    # or last operand and the current operand, in case it is last, matches result;
    # or first operand and the current operand is indeed first;
    # or first match and there hasn't been a match so far
    if choice < -1 or (choice == -1 and value == result) or (choice == 1 and first) or (choice > 1 and value not in matches):
      # store the current operand
      matches[value] = operand
    # next operand will no longer be first
    first = False
  # if choice for result is last operand, but they mismatch
  if (choices[result] == -1) and (result != value):
    return result
  else:
    # return the stored matching operand, if existing, else result as bool
    return matches.get(result, result)

testcases = [
  (-1, None, True, {None: None}, [], 'a'),
  (None, -1, {None: None}, 'a', []),
  (None, -1, True, {None: None}, 'a', []),
  (-1, None, {None: None}, [], 'a')]
choices = {-2: 'last match', -1: 'last operand', 0: 'always bool', 1: 'first operand', 2: 'first match'}
for c in testcases:
  print(c)
  for f in sorted(choices.keys()):
    for t in sorted(choices.keys()):
      x = xor(*c, falsechoice = f, truechoice = t)
      print('f: %d (%s)\tt: %d (%s)\tx: %s' % (f, choices[f], t, choices[t], x))
  print()

Python逻辑或:A或B:如果bool(A)为True则返回A，否则返回B Python逻辑和:A和B:如果bool(A)为False则返回A，否则返回B

为了保持这种思维方式，我的逻辑xor定义将是:

def logical_xor(a, b):
    if bool(a) == bool(b):
        return False
    else:
        return a or b

这样它就可以返回a, b或False:

>>> logical_xor('this', 'that')
False
>>> logical_xor('', '')
False
>>> logical_xor('this', '')
'this'
>>> logical_xor('', 'that')
'that'

有时我发现自己使用1和0而不是布尔值True和False。在这种情况下，xor可以定义为

z = (x + y) % 2

它有如下的真值表:

     x
   |0|1|
  -+-+-+
  0|0|1|
y -+-+-+
  1|1|0|
  -+-+-+

这是映射-缩减泛化的实现。注意，这相当于functools。Reduce (lambda x, y: x != y, map(bool, orands))。

def xor(*orands):
    return bool(sum(bool(x) for x in orands) % 2)

如果你在寻找一个单一热点探测器，这是一个概括。这种概括可能适用于英语中exclusive-or的用法(例如:“花一美元，你可以买一杯果汁、咖啡或茶”)，但这与典型的操作顺序不符。E.g.xor_1hot (1 1 1) = = 0 ! = 1 = = xor_1hot (xor_1hot(1, 1), 1)。

def xor_1hot(*orands):
    return sum(bool(x) for x in orands) == 1

你可以用

# test
from itertools import product
n = 3
total_true = 0
for inputs in product((False, True), repeat=n):
    y = xor(*inputs)
    total_true += int(y)
    print(f"{''.join(str(int(b)) for b in inputs)}|{y}")
print('Total True:', total_true)

单热检测器输出:

000 |假 001 |真 010 |真 011 |假 100 |真 101 |假 110 |假 111 |假 总数正确:3

使用映射-规约模式输出:

000 |假 001 |真 010 |真 011 |假 100 |真 101 |假 110 |假 111 |真 总数:4