andrewcb / genki

a composable sequence/iterator abstraction library for Swift

Genki — a composable sequence/iterator abstraction library (in progress)

Genki is a library providing abstractions over Swift sequences, allowing various common algorithms and operations to be implemented by composing them. In many cases, Genki's algorithms will be faster than using map/filter/reduce, due to not allocating and disposing intermediate structures.

What's currently included:

• SubsamplingSequence — a Sequence that wraps another Sequence and produces a sequence of values computed by performing an arbitrary operation on each set of a fixed number of its elements. Each span is represented by a bucket value, which conforms to the SubsamplingBucket protocol and consumes elements to produce a sequence of Buckets, each of which contains the result for its set of samples.

For example: if one has a sequence of Double values and wishes to find the mean value of every 3, values, the code could look like:

func meanOfEvery3(source: [Double]) -> [Double] {
    return SubsamplingSequence<[Double],MeanBucket<Double>>(source: source, span: 3).map { $0.mean }
}

SubsamplingSequence is a generic type, which takes two type parameters: the first is the type of the source sequence, and the second is a bucket type; a type which implements the SubsamplingBucket protocol and performs the actual computation on the samples in each span.

A few Bucket typeds are provided as is: these are all parametrised to the sequence element type, and are:

• MeanBucket — calculate the mean (average) of each span of samples. To do this, instantiate SubsamplingSequence with a MeanBucket<Element> of the type your sequence is comprised of (this must currently be Float, Double or Int), and for each MeanBucket value yielded, take its .mean value.

• MinMaxBucket — find the minimum and maximum values in each span; Element must be Comparable. Each returned bucket value has two values: .min and .max, which contain the relevant element values.

About

Genki is written by Andrew Bulhak and is licenced under the MIT License.