Subset MLX array

MLX subsetting is like base R with a few differences:

Usage

# S3 method for class 'mlx'
x[...] <- value

# S3 method for class 'mlx'
x[..., drop = FALSE]

x: An mlx array, or an R array/matrix/vector that will be converted via as_mlx().
...: Indices for each dimension. Provide one per axis; omitted indices select the full extent. Logical indices recycle to the dimension length.
value: Value to assign, typically an mlx or R array. Values recycle in R's column-major order; their length must divide the selection size.
drop: Should dimensions be dropped? (default: FALSE)

The subsetted MLX object.

drop = FALSE by default.
Indices containing NA, zero, or out-of-bounds positions give an error.
Numeric indices truncate toward zero, as in R. Per-axis numeric indices must be undimensioned vectors; numeric matrices are coordinate indices.
Logical per-axis indices on matrices and arrays must not exceed the axis length.
Explicit NULL indices select no elements; omitted indices select the full axis.
Single indices on a 2D or higher array are only allowed for assignment. For example, if x is a matrix, x[x < 0] <- 0 is OK but subset <- x[x < 0] is not. Use mlx_flatten() explicitly for subsetting.
There is one exception: as in R, a single numeric matrix index selects individual elements. The number of columns must match the rank of x; each row gives coordinates for one element. The return value from subsetting is a flat mlx vector.
mlx vectors, logical masks, and matrices behave the same as their R equivalents.
Duplicate vector assignments like x[c(1,1)] <- 2:3 are undefined behaviour. Coordinate matrix assignments reject duplicate rows.
Character indices match against the relevant axis dimnames.

x <- mlx_matrix(1:9, 3, 3)
x[1, ]
#> mlx array [1 x 3]
#>   dtype: float32
#>   values:
#>      [,1] [,2] [,3]
#> [1,]    1    4    7