Link Search Menu Expand Document

Map with Window function

Table of Contents


There are several cases that you may need to change every single element of the Window<T> using some function. Mainly it is used for merging the stream of windowed sample with a window function. The map operation between two windows is limited to the same size windows, so it is not that simple to generate correctly configured stream.

That operation implementation doesn’t do anything you wouldn’t be able to do on your own by implementing it directly, though it is always nice to have a better API.

Basically what this operation does, if you have source windowed stream w, and the function f with corresponding to the size of the window length, it multiplies corresponding values:

(w[0], w[1], w[2], w[3]) \
                           => (w[0]*f[0], w[1]*f[1], w[2]*f[2], w[3]*f[3])
(f[0], f[1], f[2], f[3]) /

For convenience it is implemented for Sample type, but it can be used with any type of the stream. The functions work as an extension of BeanStream<Window<Sample>> and return the same type of the stream.

Stream of Sample type

To multiply the source windowed stream with window function you need to use .windowFunction() on the stream which was already windowed. It gets either as a parameter lambda function { (i, n) -> sampleOf(...) } or a class Fn<Pair<Int, Int>, Sample>, what are the the differences and limitations of both approaches please follow functions documentation.

The arguments of the generation function are:

  1. The index of the sample in the window (Int)
  2. The overall number of samples in the window (Int)
// via lambda
        .windowFunction { (i, n) ->
            val halfN = n / 2.0
            sampleOf(1.0 - abs((i - halfN) / halfN))

// via class definition
class TriangularFn: Fn<Pair<Int, Int>, Sample>() {
    override fun apply(argument: Pair<Int, Int>): Sample {
        val (i, n) = argument
        val halfN = n / 2.0
        return sampleOf(1.0 - abs((i - halfN) / halfN))


Or there is a few predefined window functions:

Any of them is called similarly to .windowFunction():


// or 

// or 

// or 

Stream of any type

While it is very convenient to work with Sample, any other type allows to do the same. Though the API is a little more cumbersome – in addition to generation function you need to specify the multiplication function, both of each are regular WaveBeans functions.

Here is an example of applying the window function over the stream of Long:

input { (i, _) -> i }     // the type of the stream here is BeanStream<Long>
    .window(5) { 0L }     // windowing transforms to the BeanStream<Window<Long>> 
    .windowFunction(      // we'll double each elements inside the window
        func = { 2 },
        multiplyFn = { (a, b) -> a * b }

Additional convenience functions

For convenience all window functions are accessible when you add the suffix Func to the method call:

  • To calculate specific value on index i out of overall size n, i.e. hammingFunc(i, n).
  • To calculate the whole window of size n as a Kotlin sequence, i.e. hammingFunc(n).

All functions can be used to create a SampleVector of size n:

sampleVectorOf(n, ::hammingFunc)

Which later on can be used in arithmetic operation with another vectors.

Low Level API

Low Level API of the operation is the function implementation which gets two parameters: generation function and multiplication function.

Generation function

The generation function is intended to generate the window based on the source window size and has the following arguments:

  • Input type is Pair<Int, Int>, the first is the current index of the value, the second is the overall number of samples in the window.
  • Output type is T, is the value of the window on the specified index.

Multiply function

The multiply function defines how tow multiply two values coming from the stream and the window:

  • The input type is Pair<T, T>, which is a pair of sample to multiply, the first is coming from the stream, the second is coming from the window function.
  • The output type if T which is the result of multiplication.

Example of functions (working with Sample type):

val windowFunction: Fn<Pair<Int, Int>, Sample> = Fn.wrap { (i, n) ->
    // triangular window function
    val halfN = n / 2.0
    sampleOf(1.0 - abs((i - halfN) / halfN))

val multiplyFn: Fn<Pair<Sample, Sample>, Sample> = Fn.wrap { (a, b) ->
    a * b

Thus the usage of them is as simple as calling it via .map() operation:

    .map(MapWindowFn(windowFunction, multiplyFn))