Advanced Index Specification with Keywords — aaa11_squarebrackets

Instead of the usual indexing types, formulas can be used to specify indices in vectors, arrays, or data.frames that can evaluate keywords.

The following keywords are available:

.M: the given margin/dimension; 0 if not relevant.
.Nms: the (dim)names at the given margin.
.N: the size of a given dimension (if .M is not 0) or else the length of x.
.I: equal to seq_len(.N).
.bi(...): a function to specify bilateral indices. See the Details section below.
.ptrn(ptrn, start, end): a function specifying a pattern of indices.
It is defined as (start:end)[ptrn], where start and end are numeric scalars and ptrn is a logical vector.
By default, start = 1L and end = .N.

These keywords can be used inside formulas to specify more advanced indices.

Here are a few examples of advanced indexing in arrays:


x <- matrix(1:20, 5, 4)
dimnames(x) <- list(month.abb[1:5], month.abb[1:4])
print(x)
#>     Jan Feb Mar Apr
#> Jan   1   6  11  16
#> Feb   2   7  12  17
#> Mar   3   8  13  18
#> Apr   4   9  14  19
#> May   5  10  15  20


# select first half of rows, select first & last column:
tt_x(x, ~ 1:round(.N/2), ~ .bi(1, -1))
#>     Jan Apr
#> Jan   1  16
#> Feb   2  17


# extract where rownames contain "r" & where colnames DO NOT contain "r":
ss_x(x, ~ grep("r", .Nms), use = c(1, -2))
#>     Jan Feb
#> Mar   3   8
#> Apr   4   9


# reverse order of all dimensions:
ss_x(x, ~ .bi(-.I))
#>     Apr Mar Feb Jan
#> May  20  15  10   5
#> Apr  19  14   9   4
#> Mar  18  13   8   3
#> Feb  17  12   7   2
#> Jan  16  11   6   1


# select every other element for each dimension:
ss_x(x, ~ .ptrn(c(TRUE, FALSE)))
#>     Jan Mar
#> Jan   1  11
#> Mar   3  13
#> May   5  15


# select different range from each dimension:
x <- array(1:1000, 5:3)
start <- c(3, 1, 2)
ss_x(x, ~ (start[.M]):.N)
#> , , 1
#>
#>      [,1] [,2] [,3] [,4]
#> [1,]   23   28   33   38
#> [2,]   24   29   34   39
#> [3,]   25   30   35   40
#>
#> , , 2
#>
#>      [,1] [,2] [,3] [,4]
#> [1,]   43   48   53   58
#> [2,]   44   49   54   59
#> [3,]   45   50   55   60

Details

Bilateral Indices
The .bi(...) function is used to specify bilateral indices.
One can specify both positive and negative numbers.
Positive numbers (for example ~ .bi(1:10)) specify indices work like normally.
Negative numbers specify indices from the end;
i.e. ~ .bi(-2) would be equivalent to length(x) - 1 if using an ii_ method, or dim(x)[.M] - 1 if using a ss_ or tt_ method.

The usage of .bi(...) is similar to the c(...) function, in that multiple vectors can be concatenated to one.
And one can specify both negative and positive numbers together.
For example: ~ .bi(-10:-1, 10:1).

One can also use the other keywords inside .bi().
For example, ~ .bi(-.I) can be used to reverse a vector, or reverse a dimension of an array or data.frame.