Cumulative Sum

Cumulative sum, optionally along a single dimension.

Usage

nv_cumsum(operand, dim = NULL, nan_rm = FALSE)

Arguments

operand

(arrayish)
Operand.

dim

(integer(1) | NULL)
Dimension along which to accumulate. If NULL (default), the input is first flattened to a 1-D array, like base::cumsum().

nan_rm

(logical(1))
How to handle NaN values in floating-point inputs. If FALSE (default), NaN propagates forward from its first occurrence. If TRUE, NaN is treated as the identity element of the cumulative op (0 for sum, 1 for prod, -Inf / +Inf for max / min) and contributes nothing to the running value.

Value

arrayish
Has the same shape and data type as the input.

Relation to base R

Both nv_cumsum() (with dim = NULL) and base::cumsum() flatten a multi-dimensional input to 1-D before accumulating, but the flatten order differs: anvl arrays are row-major (C order), so the flattened sequence iterates the last dim fastest, whereas base R uses column-major (Fortran) order. The two agree on 1-D inputs.

See also

prim_cumsum() for the underlying primitive.

Examples

x <- nv_matrix(1:6, nrow = 2)
nv_cumsum(x)              # row-major flatten, then accumulate
#> AnvlArray
#>   1
#>   4
#>   9
#>  11
#>  15
#>  21
#> [ CPUi32{6} ] 
nv_cumsum(x, dim = 1L)    # accumulate along rows
#> AnvlArray
#>   1  3  5
#>   3  7 11
#> [ CPUi32{2,3} ] 
nv_cumsum(nv_array(c(1, NaN, 3)))                # NaN propagates
#> AnvlArray
#>    1
#>  nan
#>  nan
#> [ CPUf32{3} ] 
nv_cumsum(nv_array(c(1, NaN, 3)), nan_rm = TRUE) # NaN treated as 0
#> AnvlArray
#>  1
#>  1
#>  4
#> [ CPUf32{3} ]