RunningCalcs 0.4

Introduction

Instances of the RunnincCalc classes in this library can be fed one input value at a time. This allows running several calculations in a single pass over an iterator. This isn’t possible with the built-in variants of most calculations, such as max() and heapq.nlargest().

RunningCalc instances can be fed values directly, for example:

mean_rc, stddev_rc = RunningMean(), RunningStdDev()
for x in values:
    mean_rc.feed(x)
    stddev_rc.feed(x)
mean, stddev = mean_rc.value, stddev_rc.value

Additionally, the apply_in_parallel() function is supplied, which makes performing several calculations in parallel easy (and fast!). For example:

mean, stddev = apply_in_parallel([RunningMean(), RunningStdDev()], values)
five_smallest, five_largest = apply_in_parallel([RunningNSmallest(5), RunningNLargest(5)], values)

Optimizations

In addition to the basic feed() method, some RunningCalc classes also implement an optimized feedMultiple() method, which accepts a sequence of values to be processed. This allows values to be processed in chunks, allowing for faster processing in many cases.

The apply_in_parallel() function automatically splits the given iterable of input values into chunks (chunk size can be controlled via the chunk_size keyword argument). Therefore using apply_in_parallel() is both fast and easy.

Writing Your Own RunningCalc Class

1. sub-class RunningCalc

2. implement the __init__() and feed() methods

3. make the calculation output value accessible via the value attribute

4. optionally implement an optimized feedMultiple() method Note: the RunningCalc base class includes a default naive implementation of feedMultiple()