Choosing the right column based on a vector of column names - r

Question

I'm trying to pull values from columns based on the values in a vector. I'm not sure I have the right words to describe the problem, but the code should help.
This feels related to coalesce maybe not?
library(tidyverse)
# Starting table
dat <-
tibble(
A = 1:10,
B = 31:40,
C = 101:110,
value = c("A", "C", "B", "A", "B", "C", "C", "B", "A", "A")
)
I want:
dat %>%
mutate(
output = c(1, 102, 33, 4, 35, 106, 107, 38, 9, 10)
)
I could do
dat %>%
mutate(
output =
case_when(value == "A" ~ A,
value == "B" ~ B,
value == "C" ~ C)
)
but my real application has many values and I want to take advantage of value having the matching info
Is there a function that does:
dat %>%
mutate(output = grab_the_right_column(value))
Thanks!

Answer 1

The rowwise approach would be less efficient, but it is compact within the tidyverse approaches to get the column value based on the column name for each row.
library(dplyr)
dat %>%
rowwise %>%
mutate(output = get(value)) %>%
ungroup
-output
# A tibble: 10 x 5
# A B C value output
# <int> <int> <int> <chr> <int>
# 1 1 31 101 A 1
# 2 2 32 102 C 102
# 3 3 33 103 B 33
# 4 4 34 104 A 4
# 5 5 35 105 B 35
# 6 6 36 106 C 106
# 7 7 37 107 C 107
# 8 8 38 108 B 38
# 9 9 39 109 A 9
#10 10 40 110 A 10
These type of issues are more efficient with a row/column indexing approach from base R. Create a matrix of row sequence and the matching index of columns with the 'value' column and the column names to extract the element
dat$output <- as.data.frame(dat)[,1:3][cbind(seq_len(nrow(dat)), match(dat$value, names(dat)[1:3]))]

Answer 2

You can also use purrr and pmap():
library(dplyr)
library(purrr)
dat%>%mutate(output=
pmap(., ~{
v1<-c(...)
v1[names(v1)==v1[['value']]]
}
)%>%
as.numeric()%>%
unlist)
# A tibble: 10 x 5
A B C value output
<int> <int> <int> <chr> <dbl>
1 1 31 101 A 1
2 2 32 102 C 102
3 3 33 103 B 33
4 4 34 104 A 4
5 5 35 105 B 35
6 6 36 106 C 106
7 7 37 107 C 107
8 8 38 108 B 38
9 9 39 109 A 9
10 10 40 110 A 10