create data frame containing rows that add up to 100 - r

Question

This is my first stab at this:
library(dplyr)
step_size <- 5
grid <- expand.grid(
x1 = seq(0, 100, step_size)
, x2 = seq(0, 100, step_size)
, x3 = seq(0, 100, step_size)
)
grid$sum = grid$x1 + grid$x2 + grid$x3
grid$x1 <- (grid$x1 / grid$sum) * 100
grid$x2 <- (grid$x2 / grid$sum) * 100
grid$x3 <- (grid$x3 / grid$sum) * 100
grid$sum <- grid$x1 + grid$x2 + grid$x3
nrow(grid)
result <- distinct(grid) %>% filter(!is.na(sum))
head(result, 20)
nrow(result)
Basically, I want to create a data frame that contains as many rows as possible that add up to 100 and are uniformly distributed.
is there an easier better approach in R? thanks!

Answer 1

Using data.table...
library(data.table)
grid <- expand.grid(
x1 = seq(0, 100)
, x2 = seq(0, 100)
, x3 = seq(0, 100)
)
setDT(grid)
res <- grid[grid[, rowSums(.SD) == 100], ]
res[, summation := rowSums(.SD)]
Result:
> res[, unique(summation)]
[1] 100
This can also be done in base but data.table is faster:
library(data.table)
grid <- expand.grid(
x1 = seq(0, 100)
, x2 = seq(0, 100)
, x3 = seq(0, 100)
)
grid2 <- expand.grid(
x1 = seq(0, 100)
, x2 = seq(0, 100)
, x3 = seq(0, 100)
)
setDT(grid)
microbenchmark::microbenchmark(
data.table = {
res <- grid[grid[, rowSums(.SD) == 100], ]
},
base = {
res2 <- grid2[rowSums(grid2) == 100, ]
}
)
Unit: milliseconds
expr min lq mean median uq max neval cld
data.table 59.41157 89.6700 109.0462 107.7415 124.2675 183.9730 100 a
base 65.70521 109.6471 154.1312 125.4238 156.9168 611.0169 100 b

Answer 2

Here's a simple function. You can specify how many rows/columns you want, and what each row sums to.
func <- function(cols = 3, rows = 10, rowTotal = 100) {
dt1 <- replicate(n = cols, runif(n = rows))
dt1 <- data.frame(apply(X = dt1, MARGIN = 2, FUN = function(x) x / rowSums(dt1) * rowTotal))
return(dt1)
}
rowSums(func()) # default values (3 cols, 10 rows, each row sums to 100)
rowSums(func(cols = 5, rows = 10, rowTotal = 50)) # 5 cols, 10 rows, row sums to 50)