The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.

narray Usage Examples

Michael Schubert

2022-10-02

Loading the package

This package provides consistent utility functions for array programming with arbitrary dimensions (summary below).

We recommend to load this package in its own namespace to not shadow base R functions using box or import.

# example referencing the package namespace
# do not load the package with 'library(...)' here
narray::stack(...)

Stacking and splitting

stack() is like cbind/rbind, but along arbitrary axes, and taking care of (1) names along each dimension and (2) padding partial matching arrays.

A = matrix(1:4, nrow=2, ncol=2, dimnames=list(c('a','b'),c('x','y')))
B = matrix(5:6, nrow=2, ncol=1, dimnames=list(c('b','a'),'z'))

C = stack(A, B, along=2)
C
#>   x y z
#> a 1 3 6
#> b 2 4 5

D = stack(m=A, n=C, along=3) # we can also introduce new dimensions
D
#> , , m
#> 
#>   x y  z
#> a 1 3 NA
#> b 2 4 NA
#> 
#> , , n
#> 
#>   x y z
#> a 1 3 6
#> b 2 4 5

split() splits an array along a given axis; can do each element or defined subsets.

split(C, along=2, subsets=c('s1','s1','s2'))
#> $s1
#>   x y
#> a 1 3
#> b 2 4
#> 
#> $s2
#>   z
#> a 6
#> b 5

Mapping functions on arrays

Like apply, but not reordering array dimensions and allowing to specify subsets that the function should be applied on. The function must either return a vector of the same length as the input (returns matrix of same dimension) or of length 1 (drops current dimension or returns subsets).

map(C, along=2, function(x) x*2) # return same length vector
#>   x y  z
#> a 2 6 12
#> b 4 8 10
map(C, along=2, mean, subsets=c('s1', 's1', 's2')) # summarize each subset to scalar
#>   s1 s2
#> a  2  6
#> b  3  5

We can also index multiple arrays using the lambda function. If the result is a scalar we will get back an array, and an index with result column otherwise.

dot = function(x, y) sum(x * y)
lambda(~ dot(A, B), along=c(A=1, B=2))
#>    B
#> A    z
#>   a 23
#>   b 34
lambda(~ dot(A, B), along=c(A=1, B=2), simplify=FALSE)
#>   A B result
#> 1 a z     23
#> 2 b z     34

Intersecting

Takes a number of arrays, intersects their names along a given dimension, and returns sub-arrays that match in their names; intersect_list takes a list of arrays and returns a list of subsets.

E = matrix(1:6, nrow=3, dimnames=list(c('a','b','d'), c('x','y')))
F = matrix(7:9, nrow=3, dimnames=list(c('b','a','c'), 'z'))

intersect(E, F, along=1)
E
#>   x y
#> a 1 4
#> b 2 5
F
#>   z
#> a 8
#> b 7

Converting to and from data.frames

construct() takes a data frame and a formula specifying dependent (values) and independent (axes) of the resulting array.

DF = data.frame(k1=base::rep(letters[1:3],2),
                k2=base::rep(letters[24:25],3), v=1:6)[-6,]
construct(v ~ k1 + k2, data=DF)
#>    k2
#> k1  x  y
#>   a 1  4
#>   b 5  2
#>   c 3 NA

Masks from factors and lists

Takes either a factor or a list of vectors and creates a binary matrix specifying whether each element is present.

G = list(a='e1', b=c('e1','e2'), c='e2')
mask(G)
#>      e1    e2
#> a  TRUE FALSE
#> b  TRUE  TRUE
#> c FALSE  TRUE

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.