Rowwise while loop over dataframe? - r

Question

I have a function calculating something based on input variables, but I want to change the input variables depending on the result of the function. I feel that it is a very trivial question, but I have not found a working solution yet.
I have a dataset similar to this:
v1<-sample(5:12, 10, replace=T)
v2<-rep(100, 10)
v3<-rep(1,10)
v4<-sample(1:4, 10, replace=T)
t1<-sample(10:30, 10, replace=T)
And a function which uses the variables v1 to v4 to calculate something:
fun<-function(v1, v2, v3, v4){
#does complicated things
result<-((v2/(v1*v4))*v3
return(result)
}
out<-fun(v1, v2, v3, v4)
df<-data.frame(v1, v2, v3, v4, t1, out)
I need t1 < out. For non-vectorized data this seems to work:
while (out < t1){
v3= v3 + 1
out<- fun(v1, v2, v3, v4)
}
Ideally the last value for v3 is stored somehow.
v1 v1 v2 v3 v4 t1 out out_new v3_new
1 6 100 1 2 15 8.333333 16 2
2 12 100 1 2 17 4.166667 20 5
3 5 100 1 1 12 20.000000 20 1
4 10 100 1 4 26 2.500000 . .
5 8 100 1 2 15 6.250000 . .
6 6 100 1 3 18 5.555556
7 10 100 1 4 20 2.500000
8 11 100 1 2 12 4.545455
9 12 100 1 3 28 2.777778
10 6 100 1 2 25 8.333333
The output I am looking for is a dataframe looking like this:
What I am looking for is a vectorized solution that works on a large (100.000 rows) dataframe/tibble. I have tried different approaches with dplyr (mutate with case_when, ifelse) but have failed to come up with a working solution.

Answer 1

There is a way to achieve this using dplyr::rowwise():
library(tidyverse)
v1<-sample(5:12, 10, replace=T)
v2<-rep(100, 10)
v3<-rep(1,10)
v4<-sample(1:4, 10, replace=T)
t1<-sample(10:30, 10, replace=T)
fun<-function(v1, v2, v3, v4){
#does complicated things
result<-(v1/v2)*v4*v3
return(result)
}
reallyWeirdFunction = function(v1, v2, v3, v4, t){
out<- fun(v1, v2, v3, v4)
while (out < t){
v3= v3 + 1
out<- fun(v1, v2, v3, v4)
}
out
}
df<-data.frame(v1, v2, v3, v4, t1)
df %>%
rowwise() %>%
mutate(
out=fun(v1, v2, v3, v4),
out2=reallyWeirdFunction(v1, v2, v3, v4, t1)
)
#> # A tibble: 10 x 7
#> # Rowwise:
#> v1 v2 v3 v4 t1 out out2
#> <int> <dbl> <dbl> <int> <int> <dbl> <dbl>
#> 1 7 100 1 3 17 0.21 17.0
#> 2 11 100 1 2 24 0.22 24.2
#> 3 12 100 1 1 11 0.12 11.0
#> 4 10 100 1 4 15 0.4 15.2
#> 5 10 100 1 2 22 0.2 22
#> 6 9 100 1 1 16 0.09 16.0
#> 7 5 100 1 2 24 0.1 24
#> 8 12 100 1 2 23 0.24 23.0
#> 9 8 100 1 3 30 0.24 30
#> 10 7 100 1 2 14 0.14 14.
Created on 2020-08-28 by the reprex package (v0.3.0)
However, this is usually a very bad idea to use while loops in R, and rowwise() can be very slow in large datasets. You should definitely try to find a better algorithm than incrementing v3 like you are doing. I could not find one for your example though.

Answer 2

With mapply() and some other definitions you can do in base R:
set.seed(42)
df <- data.frame(v1=sample(5:12, 10, replace=T), v2=rep(100, 10), v3=rep(1,10),
v4=sample(1:4, 10, replace=T), t1=sample(10:30, 10, replace=T))
fun<-function(v1, v2, v3, v4) {
#does complicated things
result <- ((v2/(v1*v4))*v3)
return(result)
}
fun2 <- function(v1, v2, v3, v4, t) {
out <- fun(v1, v2, v3, v4)
while (out < t){
v3 <- v3 + 1
out <- fun(v1, v2, v3, v4)
}
return(list(v3new=v3, out=out))
}
cbind(df, t(mapply(fun2, df$v1, df$v2, df$v3, df$v4, df$t1)))
> cbind(df, t(mapply(fun2, df$v1, df$v2, df$v3, df$v4, df$t1)))
# v1 v2 v3 v4 t1 v3new out
# 1 5 100 1 3 24 4 26.66667
# 2 9 100 1 4 12 5 13.88889
# 3 5 100 1 3 18 3 20
# 4 5 100 1 4 13 3 15
# 5 6 100 1 1 14 1 16.66667
# 6 8 100 1 1 22 2 25
# 7 6 100 1 2 14 2 16.66667
# 8 6 100 1 4 29 7 29.16667
# 9 5 100 1 2 11 2 20
# 10 12 100 1 2 17 5 20.83333