def card(self, timestamp, window=None):
"""
Returns the estimate of the cardinality at 'timestamp' using 'window'
"""
if window is None:
window = self.window
if not 0 < window <= self.window:
raise ValueError('0 < window <= W')
def max_r(l):
return max(l) if l else 0
M = [
max_r([R for ts, R in lpfm
if ts >= (timestamp - window)]) if lpfm else 0
for lpfm in self.LPFM
]
#count number or registers equal to 0
V = M.count(0)
if V > 0:
H = self.m * math.log(self.m / float(V))
return H if H <= get_treshold(self.p) else self._Ep(M)
else:
return self._Ep(M)
def card_wlist(self, timestamp, window_list):
"""
Returns the estimate of the cardinality at 'timestamp' using list of windows
"""
for window in window_list:
if not 0 < window <= self.window:
raise ValueError("0 < window <= W")
tsl = [(timestamp - window, idx) for idx, window in enumerate(window_list)]
tsl.sort()
M_list = [[] for _ in window_list]
# Highly optimized code (PyPy), but may be slow in CPython
for lpfm in self.LPFM:
R_max = 0
_p = len(tsl) - 1
if lpfm:
i = len(lpfm) - 1
while i >= 0:
ts, R = lpfm[i]
while _p >= 0:
_ts, _idx = tsl[_p]
if ts >= _ts:
break
M_list[_idx].append(R_max)
_p -= 1
if _p < 0:
break
R_max = R
i -= 1
for i in xrange(0, _p + 1):
M_list[tsl[i][1]].append(R_max)
res = []
for M in M_list:
# count number or registers equal to 0
V = M.count(0)
if V > 0:
H = self.m * math.log(self.m / float(V))
res.append(H if H <= get_treshold(self.p) else self._Ep(M))
else:
res.append(self._Ep(M))
return res
def card_wlist(self, timestamp, window_list):
"""
Returns the estimate of the cardinality at 'timestamp' using list of windows
"""
for window in window_list:
if not 0 < window <= self.window:
raise ValueError('0 < window <= W')
tsl = [(timestamp - window, idx)
for idx, window in enumerate(window_list)]
tsl.sort()
M_list = [[] for _ in window_list]
# Highly optimized code (PyPy), but may be slow in CPython
for lpfm in self.LPFM:
R_max = 0
_p = len(tsl) - 1
if lpfm:
i = len(lpfm) - 1
while i >= 0:
ts, R = lpfm[i]
while _p >= 0:
_ts, _idx = tsl[_p]
if ts >= _ts: break
M_list[_idx].append(R_max)
_p -= 1
if _p < 0: break
R_max = R
i -= 1
for i in xrange(0, _p + 1):
M_list[tsl[i][1]].append(R_max)
res = []
for M in M_list:
#count number or registers equal to 0
V = M.count(0)
if V > 0:
H = self.m * math.log(self.m / float(V))
res.append(H if H <= get_treshold(self.p) else self._Ep(M))
else:
res.append(self._Ep(M))
return res
def card(self, timestamp, window=None):
"""
Returns the estimate of the cardinality at 'timestamp' using 'window'
"""
if window is None:
window = self.window
if not 0 < window <= self.window:
raise ValueError("0 < window <= W")
def max_r(l):
return max(l) if l else 0
M = [max_r([R for ts, R in lpfm if ts >= (timestamp - window)]) if lpfm else 0 for lpfm in self.LPFM]
# count number or registers equal to 0
V = M.count(0)
if V > 0:
H = self.m * math.log(self.m / float(V))
return H if H <= get_treshold(self.p) else self._Ep(M)
else:
return self._Ep(M)
