CODEWITHSUNDEEP
rdplyrtidyverse
littleworth
littleworth

Reputation: 5169

How to apply function row-by-row into a data frame using dplyr without hardcoding the column names

I have the following data frame:

dat <- structure(list(setosa = c(50L, 0L, 0L), versicolor = c(0L, 11L, 
39L), virginica = c(0L, 36L, 14L)), .Names = c("setosa", "versicolor", 
"virginica"), row.names = c("1", "2", "3"), class = "data.frame")

dat
#>   setosa versicolor virginica
#> 1     50          0         0
#> 2      0         11        36
#> 3      0         39        14

This is the current code I use to calculate the score by hard-coding the column name into it:

library(dplyrj)
dat %>% 
  rowwise() %>% 
  # here I hard code the column names into the score
  mutate(score = max(c(setosa,versicolor, virginica)/ sum(c(setosa, versicolor, virginica))))

Which produces:

# A tibble: 3 x 4
  setosa versicolor virginica score
   <int>      <int>     <int> <dbl>
1     50          0         0 1.00 
2      0         11        36 0.766
3      0         39        14 0.736

What I want to do is to calculate each score, but without hard coded the column name.

How can achieve that?

Upvotes: 2

Views: 89

Answers (2)

Romain Francois
Romain Francois

Reputation: 17642

With the unquote splice operator !!!, you can do: 

> library(tidyverse)
> psum <- function(...) reduce(list(...), `+` )
> mutate( dat, core = pmax(!!!syms(names(dat))) / psum(!!!syms(names(dat))) )
  setosa versicolor virginica      core
1     50          0         0 1.0000000
2      0         11        36 0.7659574
3      0         39        14 0.7358491

That works by generating the call for you, i.e.

> rlang::qq_show( mutate( dat, core = pmax(!!!syms(names(dat))) / psum(!!!syms(names(dat))) ) )
mutate(dat, core = pmax(setosa, versicolor, virginica) / 
                   psum(setosa, versicolor, virginica)
)

Upvotes: 1

akrun
akrun

Reputation: 887941

A concise base R option would be

dat$score <- do.call(pmax, dat)/rowSums(dat)

In tidyverse we can do

library(tidyverse)
dat %>% 
    mutate(score = do.call(pmax, .)/reduce(., `+`))
#   setosa versicolor virginica     score
#1     50          0         0 1.0000000
#2      0         11        36 0.7659574
#3      0         39        14 0.7358491

Upvotes: 3

Related Questions