SparseArray-Arith-methods: 'Arith' operations on SparseArray objects

In Bioconductor/SparseArray: High-performance sparse data representation and manipulation in R

'Arith' operations on SparseArray objects

Description

SparseArray derivatives support operations from the Arith group, with some restrictions. See ?S4groupGeneric in the methods package for more information about the Arith group generic.

IMPORTANT NOTES:

• Only SVT_SparseArray objects are supported at the moment. Support for COO_SparseArray objects might be added in the future.

• SVT_SparseArray of type() "complex" don't support Arith operations at the moment.

Details

Two forms of 'Arith' operations are supported:

1. Between an SVT_SparseArray object svt and a single value y:

       svt op y
       y op svt

   The Arith operations that support this form are: *, /, ^, %%,%/%. Note that, except for * (for which both svt * y and y * svt are supported), single value y must be on the right e.g. svt ^ 3.

2. Between two SVT_SparseArray objects svt1 and svt2 of same dimensions (a.k.a. conformable arrays):

       svt1 op svt2

   The Arith operations that support this form are: +, -, *.

Value

A SparseArray derivative of the same dimensions as the input object(s).

See Also

• S4groupGeneric in the methods package.

• SparseArray objects.

• Ordinary array objects in base R.

Examples

## ---------------------------------------------------------------------
## Basic examples
## ---------------------------------------------------------------------

svt1 <- SVT_SparseArray(dim=c(15, 6), type="integer")
svt1[cbind(1:15, 2)] <- 100:114
svt1[cbind(1:15, 5)] <- -(114:100)
svt1

svt1 * -0.01

svt1 * 10   # result is of type "double"
svt1 * 10L  # result is of type "integer"

svt1 / 10L

svt1 ^ 3.5

svt1 %% 5L
svt1 %/% 5L

svt2 <- SVT_SparseArray(dim=dim(svt1), type="double")
svt2[c(2, 6, 12:17, 22:33, 55, 59:62, 90)] <- runif(26)
svt2

svt1 + svt2
svt1 - svt2
svt1 * svt2

svt2 * (0.1 * svt1 - svt2 ^ 2) + svt1 / sum(svt2)

## Sanity checks:
m1 <- as.matrix(svt1)
m2 <- as.matrix(svt2)
stopifnot(
  identical(as.matrix(svt1 * -0.01), m1 * -0.01),
  identical(as.matrix(svt1 * 10), m1 * 10),
  identical(as.matrix(svt1 * 10L), m1 * 10L),
  identical(as.matrix(svt1 / 10L), m1 / 10L),
  identical(as.matrix(svt1 ^ 3.5), m1 ^ 3.5),
  identical(as.matrix(svt1 %% 5L), m1 %% 5L),
  identical(as.matrix(svt1 %/% 5L), m1 %/% 5L),
  identical(as.matrix(svt1 + svt2), m1 + m2),
  identical(as.matrix(svt1 - svt2), m1 - m2),
  identical(as.matrix(svt1 * svt2), m1 * m2),
  all.equal(as.matrix(svt2 * (0.1 * svt1 - svt2 ^ 2) + svt1 / sum(svt2)),
            m2 * (0.1 * m1 - m2 ^ 2) + m1 / sum(m2))
)

## ---------------------------------------------------------------------
## An example combining operations from the 'Arith', 'Compare',
## and 'Logic' groups
## ---------------------------------------------------------------------

m3 <- matrix(0L, nrow=15, ncol=6)
m3[c(2, 6, 12:17, 22:33, 55, 59:62, 90)] <- 101:126
svt3 <- SparseArray(m3)

## Can be 5x or 10x faster than with a dgCMatrix object on a big
## SVT_SparseMatrix object!
svt4 <- (svt3^1.5 + svt3) %% 100 - 0.2 * svt3 > 0
svt4

## Sanity check:
m4 <- (m3^1.5 + m3) %% 100 - 0.2 * m3 > 0
stopifnot(identical(as.matrix(svt4), m4))

Bioconductor/SparseArray documentation built on Sept. 19, 2024, 11:36 a.m.