py: runningmax with double deque and rampup ok?

I have implemented this runningMax class:

class RunningMax():

    def __init__(self,window):
        self.window = window
        self.values = deque()
        self.maxima = deque()
        return

    def append(self,x):
        self.values.append(x)
        i = 0
        while (len(self.maxima) > 0) and (x > self.maxima[-1]) and (i < min(len(self.values),self.window)):
            # pop all values at back that are smaller than new value
            self.maxima.pop()
            i += 1
        self.maxima.append(x)
        return

    def pop(self):
        full = len(self.values) == self.window
        if self.maxima[0]==self.values[0] and full:
            # if a value is about to leave the sliding window, kick it from maxima
            max = self.maxima.popleft()
        else:
            max = self.maxima[0]
        if full:
            self.values.popleft()
        return max

def compute_max(a,window):
    rm = RunningMax(window)
    maxs = []
    for i,ai in enumerate(a):
        rm.append(ai)
        maxs.append(rm.pop())
    return maxs

The output of compute_max(a,window) compares exactly to:

pd.Series(a).rolling(window,min_periods=1).max()

So, its implemented correctly. Just would like to ask if there is any potential for improvement here? Anything that could be written more optimally basically.

Thanks!