def _welford():
"""Welford's method of calculating the running variance.
Consume values and return running estimates of (n, M2) where:
n = number of data points seen so far
M2 = the second moment about the mean
= sum( (x-m)**2 ) where m = mean of the x seen so far.
"""
# Note: for better results, use this on the residues (x - m) instead
# of the raw x values, where m equals the mean of the data.
M2_partials = []
x = (yield None)
m = x # First estimate of the mean is the first value.
n = 1
while True:
delta = x - m
m += delta/n # Update the mean.
stats.add_partial(delta*(x-m), M2_partials) # Update the 2nd moment.
M2 = _sum(M2_partials)
assert M2 >= 0.0
x = (yield (n, M2))
n += 1
def value(self):
return _sum(self.partials)
def sum(data, start=0):
"""sum(data [, start]) -> value
Return a high-precision sum of the given numeric data. If optional
argument ``start`` is given, it is added to the total. If ``data`` is
empty, ``start`` (defaulting to 0) is returned.
Examples
--------
>>> sum([3, 2.25, 4.5, -0.5, 1.0], 0.75)
11.0
Float sums are calculated using high-precision floating point arithmetic
that can avoid some sources of round-off error:
>>> sum([1e50, 1, -1e50] * 1000) # Built-in sum returns zero.
1000.0
Fractions and Decimals are also supported:
>>> from fractions import Fraction as F
>>> sum([F(2, 3), F(7, 5), F(1, 4), F(5, 6)])
Fraction(63, 20)
Decimal sums honour the context:
>>> import decimal
>>> D = decimal.Decimal
>>> data = [D("0.1375"), D("0.2108"), D("0.3061"), D("0.0419")]
>>> sum(data)
Decimal('0.6963')
>>> with decimal.localcontext(
... decimal.Context(prec=2, rounding=decimal.ROUND_DOWN)):
... sum(data)
Decimal('0.68')
Limitations
-----------
``sum`` supports mixed arithmetic with the following limitations:
- mixing Fractions and Decimals raises TypeError;
- mixing floats with either Fractions or Decimals coerces to float,
which may lose precision;
- complex numbers are not supported.
These limitations may change without notice in future versions.
"""
if not isinstance(start, numbers.Number):
raise TypeError('sum only accepts numbers')
total = start
data = iter(data)
x = None
if not isinstance(total, float):
# Non-float sum. If we find a float, we exit this loop and continue
# with the float code below. Until that happens, we keep adding.
for x in data:
if isinstance(x, float):
# Convert running total to a float. See comment below for
# why we do it this way.
total = type(total).__float__(total)
break
total += x
else:
# No break, so we're done.
return total
# High-precision float sum.
assert isinstance(total, float)
partials = []
add_partial(total, partials)
if x is not None:
add_partial(x, partials)
for x in data:
try:
# Don't call float() directly, as that converts strings and we
# don't want that. Also, like all dunder methods, we should call
# __float__ on the class, not the instance.
x = type(x).__float__(x)
except OverflowError:
x = float('inf') if x > 0 else float('-inf')
add_partial(x, partials)
return _sum(partials)
