def test_empty_merge_full():
x1 = VarianceTracker()
vals = list(range(10))
for v in vals:
x1.update(v)
var = x1.variance()
mean = x1.mean
count = x1.count
x2 = VarianceTracker().merge(x1)
assert var == x2.variance()
assert mean == x2.mean
assert count == x2.count
assert x1.variance() == x2.variance()
assert x1.mean == x2.mean
assert x1.count == x2.count
def test_update():
vals = [1.0, 2.0, 3]
tracker = VarianceTracker()
for v in vals:
tracker.update(v)
assert tracker.mean == pytest.approx(2.0, 1e-6)
assert tracker.count == len(vals)
# Note: this is a population variance estimate (normalized by N-1)
assert tracker.variance() == pytest.approx(1.0, 1e-6)
def test_protobuf_roundtrip_matches():
x1 = VarianceTracker()
for i in range(10):
x1.update(i)
proto = x1.to_protobuf()
roundtrip = VarianceTracker.from_protobuf(proto)
assert x1.count == roundtrip.count
assert x1.variance() == roundtrip.variance()
assert x1.mean == roundtrip.mean
def test_merge_bigger():
x1 = VarianceTracker()
for i in range(10):
x1.update(i)
assert x1.variance() == pytest.approx(9.1667, 1e-4)
assert x1.count == 10
assert x1.mean == pytest.approx(4.5, 1e-7)
merged = x1.merge(x1.copy())
assert merged.variance() == pytest.approx(8.684, 1e-4)
assert merged.count == 20
assert merged.mean == pytest.approx(4.5, 1e-7)
def test_copy():
x1 = VarianceTracker()
for i in range(10):
x1.update(i)
var = x1.variance()
count = x1.count
mean = x1.mean
x2 = x1.copy()
assert x2.variance() == var
assert x2.count == count
assert x2.mean == mean
for i in range(3):
x1.update(i)
assert x1.variance() != var
assert x1.count == count + 3
assert x1.mean != mean
assert x2.variance() == var
assert x2.mean == mean
assert x2.count == count
def test_single_value():
tracker = VarianceTracker()
tracker.update(1.0)
assert tracker.count == 1
assert tracker.variance() == 0