def test_quantile():
h = distogram.Distogram(bin_count=3)
h = distogram.update(h, 16, count=4)
h = distogram.update(h, 23, count=3)
h = distogram.update(h, 28, count=5)
assert distogram.quantile(h, 0.5) == approx(23.625)
def test_count_at_not_enough_elements():
h = distogram.Distogram()
h = distogram.update(h, 1)
h = distogram.update(h, 2)
h = distogram.update(h, 3)
assert distogram.count_at(h, 2.5) == 2
def test_count():
h = distogram.Distogram(bin_count=3)
assert distogram.count(h) == 0
h = distogram.update(h, 16, count=4)
assert distogram.count(h) == 4
h = distogram.update(h, 23, count=3)
assert distogram.count(h) == 7
h = distogram.update(h, 28, count=5)
assert distogram.count(h) == 12
def test_count_at():
h = distogram.Distogram(bin_count=3)
print(h)
# fill histogram
h = distogram.update(h, 16, count=4)
h = distogram.update(h, 23, count=3)
h = distogram.update(h, 28, count=5)
print(h)
actual_result = distogram.count_at(h, 25)
assert actual_result == approx(6.859999999)
def test_count_at_right():
h = distogram.Distogram(bin_count=6)
for i in [1, 2, 3, 4, 5, 6, 6.7, 6.1]:
h = distogram.update(h, i)
assert distogram.count_at(h, 6.5) == approx(7.307692307692308)
def test_quantile_not_enough_elemnts():
h = distogram.Distogram(bin_count=10)
for i in [12.3, 5.4, 8.2, 100.53, 23.5, 13.98]:
h = distogram.update(h, i)
assert distogram.quantile(h, 0.5) == approx(13.14)
def test_count_at_left():
h = distogram.Distogram(bin_count=6)
for i in [1, 2, 3, 4, 5, 6, 0.7, 1.1]:
h = distogram.update(h, i)
assert distogram.count_at(h, 0.77) == approx(0.14)
def test_quantile_out_of_bouns():
h = distogram.Distogram()
for i in [1, 2, 3, 4, 5, 6, 6.7, 6.1]:
h = distogram.update(h, i)
assert distogram.quantile(h, -0.2) is None
assert distogram.quantile(h, 10) is None
def test_count_at_out_of_bouns():
h = distogram.Distogram()
for i in [1, 2, 3, 4, 5, 6, 6.7, 6.1]:
h = distogram.update(h, i)
assert distogram.count_at(h, 0.2) is None
assert distogram.count_at(h, 10) is None
def test_update():
h = distogram.Distogram(bin_count=3)
# fill histogram
h = distogram.update(h, 23)
assert h.bins == [(23, 1)]
h = distogram.update(h, 28)
assert h.bins == [(23, 1), (28, 1)]
h = distogram.update(h, 16)
assert h.bins == [(16, 1), (23, 1), (28, 1)]
# update count on existing value
h = distogram.update(h, 23)
assert h.bins == [(16, 1), (23, 2), (28, 1)]
h = distogram.update(h, 28)
assert h.bins == [(16, 1), (23, 2), (28, 2)]
h = distogram.update(h, 16)
assert h.bins == [(16, 2), (23, 2), (28, 2)]
# merge values
h = distogram.update(h, 26)
assert h.bins[0] == (16, 2)
assert h.bins[1] == (23, 2)
assert h.bins[2][0] == approx(27.33333)
assert h.bins[2][1] == 3
def test_count_at_normal():
points = 10000
normal = [random.normalvariate(0.0, 1.0) for _ in range(points)]
h = distogram.Distogram()
for i in normal:
h = distogram.update(h, i)
assert distogram.count_at(h, 0) == approx(points/2, rel=0.05)
def test_bounds():
normal = [random.normalvariate(0.0, 1.0) for _ in range(10000)]
h = distogram.Distogram()
for i in normal:
h = distogram.update(h, i)
dmin, dmax = distogram.bounds(h)
assert dmin == min(normal)
assert dmax == max(normal)
def test_stats():
normal = [random.normalvariate(0.0, 1.0) for _ in range(10000)]
h = distogram.Distogram()
for i in normal:
h = distogram.update(h, i)
assert distogram.mean(h) == approx(np.mean(normal), abs=0.1)
assert distogram.variance(h) == approx(np.var(normal), abs=0.1)
assert distogram.stddev(h) == approx(np.std(normal), abs=0.1)
def test_quantile_on_right():
h = distogram.Distogram(bin_count=6)
data = [12.3, 8.2, 100.53, 23.5, 13.98, 200, 200.2, 200.8, 200.4, 200.1]
for i in data:
h = distogram.update(h, i)
assert distogram.quantile(h, 0.99) == approx(np.quantile(data, 0.99),
rel=0.01)
assert distogram.quantile(h, 0.85) == approx(np.quantile(data, 0.85),
rel=0.01)
def test_normal():
# normal = np.random.normal(0,1, 1000)
normal = [random.normalvariate(0.0, 1.0) for _ in range(10000)]
h = distogram.Distogram(bin_count=64)
for i in normal:
h = distogram.update(h, i)
assert distogram.quantile(h, 0.5) == approx(np.quantile(normal, 0.5),
abs=0.2)
assert distogram.quantile(h, 0.95) == approx(np.quantile(normal, 0.95),
abs=0.2)
def test_quantile_on_left():
h = distogram.Distogram(bin_count=6)
data = [12.3, 5.2, 5.4, 4.9, 5.5, 5.6, 8.2, 30.53, 23.5, 13.98]
for i in data:
h = distogram.update(h, i)
assert distogram.quantile(h, 0.01) == approx(np.quantile(data, 0.01),
rel=0.01)
assert distogram.quantile(h, 0.05) == approx(np.quantile(data, 0.05),
rel=0.05)
assert distogram.quantile(h, 0.25) == approx(np.quantile(data, 0.25),
rel=0.05)
import numpy as np
import distogram
print("generating distribution...")
utterance_count = 10000000
distribution = np.random.normal(size=utterance_count)
#distribution = np.random.uniform(size=utterance_count)
if len(sys.argv) >= 2 and sys.argv[1] == '--enable-np':
print('using numpy types')
else:
print('using python types')
distribution = distribution.tolist()
print("distribution generated")
print("running update bench (5 times)...")
total_time = 0
for i in range(5):
start_time = time.time()
h = distogram.Distogram()
for i in distribution:
h = distogram.update(h, i)
end_time = time.time()
total_time += end_time - start_time
total_time /= 5
print("ran update bench in {} seconds (mean of 5 runs)".format(total_time))
print("req/s: {}".format(utterance_count // total_time))