生成器的 max() 是在构建类似列表的对象还是工作效率更高?

Is max() of a generator building a list-like object or is it working more efficient?

假设我有一个目录,其中有名称为 'filename_1''filename_2' 等的文件名,并且有一个生成器 models_paths,我用它来查找最新的数字:

mypath = 'my/path/filename'
models_paths = Path(mypath).parent.glob(Path(mypath).name + '*')
number_newest = max(int(str(file_path).split('_')[-1]) for file_path in models_paths)

我想知道max是在构建一个类似列表的数据结构还是在使用像

这样的算法
number_newest = None
for file_path in models_paths:
    number_current = int(str(file_path).split('_')[-1])
    number_newest = number_current if number_newest is None else max(number_current, number_newest)

换句话说:如果我写

我会失去处理效率 and/or 内存效率吗
mypath = 'my/path/filename'
models_paths = Path(mypath).parent.glob(Path(mypath).name + '*')
models_paths = list(models_paths)
number_newest = max(int(str(file_path).split('_')[-1]) for file_path in models_paths)

?

max 不构建列表。

这可以在这个例子中用自定义对象清楚地证明:

class Thing:
    
    def __init__(self, x):
        self.x = x
        print(f'creating {x}')
        
    def __lt__(self, other):
        return self.x < other.x

    def __del__(self):
        print(f'destroying {self.x}')

    def __str__(self):
        return f'<{self.x}>'
        

print(max(Thing(i) for i in range(5)))

给出:

creating 0
creating 1
destroying 0
creating 2
destroying 1
creating 3
destroying 2
creating 4
destroying 3
<4>
destroying 4

如您所见,一旦确定不再是具有最大值的对象,就会对每个对象调用 __del__ 方法。如果将它们附加到列表中,情况就不会如此。

对比:

print(max([Thing(i) for i in range(5)]))

给出:

creating 0
creating 1
creating 2
creating 3
creating 4
destroying 3
destroying 2
destroying 1
destroying 0
<4>
destroying 4

您可以编写一个(效率较低的)等效函数并证明它做同样的事情:

def mymax(things):
    empty = True
    for thing in things:
        if empty or (thing > maximum):  # parentheses for clarity only
            maximum = thing
            empty = False
    if empty:
        raise ValueError
    return maximum