This function makes it possible to control the ordering of window functions in R that don't have a specific ordering parameter. When translated to SQL it will modify the order clause of the OVER function.

order_by(order_by, call)

Arguments

order_by

a vector to order_by

call

a function call to a window function, where the first argument is the vector being operated on

Details

This function works by changing the call to instead call with_order() with the appropriate arguments.

Examples

order_by(10:1, cumsum(1:10))
#>  [1] 55 54 52 49 45 40 34 27 19 10
x <- 10:1
y <- 1:10
order_by(x, cumsum(y))
#>  [1] 55 54 52 49 45 40 34 27 19 10

df <- data.frame(year = 2000:2005, value = (0:5) ^ 2)
scrambled <- df[sample(nrow(df)), ]

wrong <- mutate(scrambled, running = cumsum(value))
arrange(wrong, year)
#>   year value running
#> 1 2000     0       0
#> 2 2001     1      55
#> 3 2002     4      54
#> 4 2003     9      50
#> 5 2004    16      16
#> 6 2005    25      41

right <- mutate(scrambled, running = order_by(year, cumsum(value)))
arrange(right, year)
#>   year value running
#> 1 2000     0       0
#> 2 2001     1       1
#> 3 2002     4       5
#> 4 2003     9      14
#> 5 2004    16      30
#> 6 2005    25      55