mapblock: Apply a function over blocks of a vector
In lambda.tools: Tools for Modeling Data with Functional Programming

Description Arguments Value Usage Details Author(s) Examples

This form of map operates on non-overlapping adjacent blocks of a data structure.

`x`	Any indexable data structure
`block`	The block size used to map over
`fn`	A function applied to a block
`...`	Optional arguments to pass to sapply

A vector containing the result of fn applied to each block

mapblock(x, window, fn, ...)

This function is useful primarily in the two-dimensional form. The use case is when a number of rotation matrices should be applied to a set of points. By collecting all the rotation matrices into a larger matrix, it is easy to produce a map process along the sub-matrices in a way that doesn't require managing indices.

Unlike maprange, mapblock doesn't have a do.pad option. Typical usage scenarios begin by constructing a matrix block that is compatible with some other data structure. Hence given a matrix A with dimensions m x n and a window of length m, it is possible to construct a k x m block matrix B composed of smaller m x m sub-matrices such that each iteration of mapblock operates on a 1 x m vector against an m x m sub-matrix. The point is that by construction the dimensions must be compatible, so padding after the fact becomes unnecessary.

The 1D version is provided for completeness and is equivalent to a 2D map, except on the edge cases.

Brian Lee Yung Rowe

# Apply multiple rotation matrices to a set of points
a <- matrix(sample(12, 20, replace=TRUE), nrow=2)
theta <- 2 * pi * sample(360,4, replace=TRUE) / 360
b <- fold(theta, function(d,acc)
  cbind(acc,matrix(c(cos(d),sin(d),-sin(d),cos(d)), nrow=2)), c())
z <- mapblock(b, 2, function(m) m %*% a, simplify=FALSE)

# The 1D version is equivalent to a 2D map
x <- 1:24
mapblock(x, 4, sum) == map(matrix(x,nrow=4), sum)