array_selection: Manipulation of array selections

In DelayedArray: A unified framework for working transparently with on-disk and in-memory array-like datasets

Description

NOTE: The tools documented in this man page are primarily intended for developers or advanced users curious about the internals of the DelayedArray package. End users typically don't need them for their regular use of DelayedArray objects.

An array selection is just an index into an array-like object that contains the information of which array elements are selected. This index can take various forms but 3 special forms are particularly useful and extensively used thoughout the DelayedArray framework: linear index (also referred to as L-index or Lindex), matrix index (also referred to as M-index or Mindex), N-dimensional index (also referred to as N-index or Nindex). See Details section below for more information.

Two utility functions are provided at the moment to convert back and forth between L-indices and M-indices. More will be added in the future to convert between other types of array indices.

Usage

## Convert back and forth between L-indices and M-indices:
Lindex2Mindex(Lindex, dim, use.names=FALSE)
Mindex2Lindex(Mindex, dim, use.names=FALSE, as.integer=FALSE)

Arguments

Lindex	An L-index. See Details section below.
Mindex	An M-index. See Details section below. For convenience, Mindex can also be specified as an integer vector with one element per dimension in the underlying array, in which case it will be treated like a 1-row matrix.
dim	An integer vector containing the dimensions of the underlying array. Note that dim can also be an integer matrix, in which case it must have one row per element in Lindex (or per row in Mindex) and one column per dimension in the underlying array.
use.names	Should the names (or rownames) on the input be propagated to the output?
as.integer	Set to TRUE to force Mindex2Lindex to return the L-index as an integer vector. Dangerous! By default, i.e. when as.integer=FALSE, Mindex2Lindex will return the L-index either as an integer or numeric vector. It will choose the former only if it's safe, that is, only if all the values in the L-index "fit" in the integer type. More precisely: If dim is not a matrix (i.e. is a vector) or if it's a matrix with a single row: Mindex2Lindex returns an integer or numeric vector depending on whether prod(dim) is <= .Machine$integer.max (2^31 - 1) or not. Otherwise Mindex2Lindex returns a numeric vector. Note that with these rules, Mindex2Lindex can return a numeric vector even if an integer vector could have been used. Use as.integer=TRUE only in situations where you know that all the L-index values are going to "fit" in the integer type. Mindex2Lindex will return garbage if they don't.

Details

The 3 special forms of array indices extensively used thoughout the DelayedArray framework:

1. Linear index (or L-index or Lindex): A numeric vector with no NAs where each value is >= 1 and <= the length of the array-like object. When using an L-index to subset an array-like object, the returned value is a vector-like object (i.e. no dimensions) of the same length as the L-index.

   Example:

               a <- array(101:124, 4:2)
               Lindex <- c(7, 2, 24, 2)
               a[Lindex]
             

2. Matrix index (or M-index or Mindex): An integer matrix with one column per dimension in the array-like object and one row per array element in the selection. No NAs. The values in each column must be >= 1 and <= the extent of the array-like object along the corresponding dimension. When using an M-index to subset an array-like object, the returned value is a vector-like object (i.e. no dimensions) of length the number of rows in the M-index.

   Example:

               a <- array(101:124, 4:2)
               Mindex <- rbind(c(3, 2, 1),
                               c(2, 1, 1),
                               c(4, 3, 2),
                               c(2, 1, 1))
               a[Mindex]
             

   Note that this is the type of index returned by base::arrayInd.

3. N-dimensional (or N-index or Nindex): A list with one list element per dimension in the array-like object. Each list element must be a subscript describing the selection along the corresponding dimension of the array-like object. IMPORTANT: A NULL subscript is interpreted as a missing subscript ("missing" like in a[ , , 1:2]), that is, as a subscript that runs along the full extend of the corresponding dimension of the array-like object. This means that before an N-index can be used in a call to [, [<-, [[ or [[<-, the NULL list elements in it must be replaced with objects of class "name". When using an N-index to subset an array-like object, the returned value is another array-like object of dimensions the lengths of the selections along each dimensions.

   Examples:

               a <- array(101:124, 4:2)
   
               ## Normalized N-index:
   
               Nindex <- list(c(1, 4, 1), NULL, 1)
               ## Same as a[c(1, 4, 1), , 1, drop=FALSE]:
               DelayedArray:::subset_by_Nindex(a, Nindex)
   
               Nindex <- list(integer(0), NULL, 1)
               ## Same as a[integer(0), , 1, drop=FALSE]:
               DelayedArray:::subset_by_Nindex(a, Nindex)
   
               ## Non-normalized N-index:
   
               Nindex <- list(-3, NULL, 1)
               Nindex <- DelayedArray:::normalizeNindex(Nindex, a)
               ## Same as a[-3, , 1, drop=FALSE]:
               DelayedArray:::subset_by_Nindex(a, Nindex)
   
               Nindex <- list(IRanges(2, 4), NULL, 1)
               Nindex <- DelayedArray:::normalizeNindex(Nindex, a)
               ## Same as a[2:4, , 1, drop=FALSE]:
               DelayedArray:::subset_by_Nindex(a, Nindex)
   
               dimnames(a)[[1]] <- LETTERS[1:4]
               Nindex <- list(c("D", "B"), NULL, 1)
               Nindex <- DelayedArray:::normalizeNindex(Nindex, a)
               ## Same as a[c("D", "B"), , 1, drop=FALSE]:
               DelayedArray:::subset_by_Nindex(a, Nindex)
             

Value

Lindex2Mindex returns an M-index.

Mindex2Lindex returns an L-index.

See Also

arrayInd in the base package.

Examples

dim <- 4:2
Mindex2Lindex(c(4, 3, 1), dim)
Mindex2Lindex(c(4, 3, 2), dim)

Mindex <- rbind(c(1, 1, 1),
                c(2, 1, 1),
                c(3, 1, 1),
                c(4, 1, 1),
                c(1, 2, 1),
                c(1, 1, 2),
                c(4, 3, 2))

Mindex2Lindex(Mindex, dim)

## With a matrix of dimensions:

dims <- rbind(c(4L, 3L),
              c(5L, 3L),
              c(6L, 3L))

Mindex <- rbind(c(1,  2),
                c(1,  2),
                c(1,  2))

Mindex2Lindex(Mindex, dims)

## Sanity checks:

dim <- c(33:30, 45L, 30L)
stopifnot(Mindex2Lindex(rep(1, 6), dim) == 1)
stopifnot(Mindex2Lindex(dim, dim) == prod(dim))

stopifnot(identical(Mindex2Lindex(arrayInd(1:120, 6:4), 6:4), 1:120))
stopifnot(identical(Mindex2Lindex(arrayInd(840:1, 4:7), 4:7), 840:1))

DelayedArray documentation built on March 25, 2021, 6:01 p.m.