| 2016-04-21 12:55:02 |
Alexander Maretskiy |
description |
See the code below and result of its execution.
Sometimes we have up to 80% difference with real percentile result.
This happens when we have very long list of mixed small and big values.
$ cat percentile_issue.py
from rally.common import streaming_algorithms as st
data = (
list(range(10)),
list(range(10)) * 10,
list(range(10)) * 100,
list(range(10)) * 1000,
list(range(10)) * 10000,
[1, 2, 3, 4, 99999] * 10000,
)
for lst in data:
p = st.PercentileComputation(.95, len(lst))
for i in lst:
p.add(i)
streaming = p.result()
real = st.utils.percentile(lst, .95)
diff = float(abs(real - streaming)) / max(real, streaming) * 100
if diff > 5: # <= 1% is Ok
print "%-8.2f %-8.2f Differs by %.1f%%" % (real, streaming, diff)
else:
print "%-8.2f %.2f" % (real, streaming)
$ python percentile_issue.py
8.55 8.55
9.00 9.00
9.00 9.00
9.00 9.00
9.00 4.50 Differs by 50.0%
99999.00 20001.80 Differs by 80.0% |
See the code below and result of its execution.
Sometimes we have up to 80% difference with real percentile result.
This happens when we have very long list of mixed small and big values.
$ cat percentile_issue.py
from rally.common import streaming_algorithms as st
data = (
list(range(10)),
list(range(10)) * 10,
list(range(10)) * 100,
list(range(10)) * 1000,
list(range(10)) * 10000,
[1, 2, 3, 4, 99999] * 10000,
)
for lst in data:
p = st.PercentileComputation(.95, len(lst))
for i in lst:
p.add(i)
streaming = p.result()
real = st.utils.percentile(lst, .95)
diff = float(abs(real - streaming)) / max(real, streaming) * 100
if diff > 5:
print "%-8.2f %-8.2f Differs by %.1f%%" % (real, streaming, diff)
else:
print "%-8.2f %.2f" % (real, streaming)
$ python percentile_issue.py
8.55 8.55
9.00 9.00
9.00 9.00
9.00 9.00
9.00 4.50 Differs by 50.0%
99999.00 20001.80 Differs by 80.0% |
|
