CODEWITHSUNDEEP
rconditional-statementsdplyr
Phil
Phil

Reputation: 1179

Mutate by condition with many columns with each one a different setting

I have looking for but not found how make a simple if for many columns in dplyr.

I have this code (it works):

library(dplyr)
library(magrittr)
data("PlantGrowth")
PlantGrowth %>% mutate (
  a=if_else(group=="ctrl", weight*2, weight*100),
  b=if_else(group=="ctrl", weight*1,5, weight/100),
  c=if_else(group=="ctrl", weight*4, weight*100),
  d=if_else(group=="ctrl", weight*5, weight/1000)
)

And I would like to not repeat the condition. Something like that:

PlantGrowth %>% mutate_if_foo (
  group=="ctrl",{
   a=weight*2,
   b=weight*1,5,
   c=weight*4,
   d=weight*5
  }
)%>% mutate_if_foo (
  group!="ctrl",{
   a=weight*100,
   b=weight/100),
   c=weight*100),
   d=weight/1000)
  }
)

I've found many answers on mutate_if,mutate_all, mutate_at , case_when but they don't answer at my question.

Please with dplyr / tidyverse.

Thanks in advance

EDIT

I've tried, from @Rohit_das idea about functions.

mtcars %>% ( function(df) { 
  if (df$am==1){
    df%>% mutate(
      a=df$mpg*3,
      b=df$cyl*10) 
   }else{ 
     df%>% mutate(
      a=df$disp*300,
      d=df$cyl*1000) 
   }
})

but I have Warning message: 

In if (df$am == 1) { : 
the condition has length > 1 
and only the first element will be used

Upvotes: 3

Views: 132

Answers (3)

Phil
Phil

Reputation: 1179

I think I've found an answer. I tested on mtcars. I didn't test yet on my real code.

Comment please if I you think I am wrong in the concept.

The conditions of the filters have to be exclusives else I will take duplicate lines.

        library(dplyr)
        library(magrittr)
        library(tibble) # only if necessary to preserve rownames
        mtcars %>% ( function(df) { 
                rbind(
                    (df 
                     %>% tibble::rownames_to_column(.) %>%tibble::rowid_to_column(.)  # to preserve rownames
                     %>%dplyr::filter(am==1) 
                     %>%dplyr::mutate(
                        a=mpg*3,
                        b=cyl*10,d=NA)),
                    (df 
                     %>% tibble::rownames_to_column(.) %>%tibble::rowid_to_column(.)  # to preserve rownames
                     %>%dplyr::filter(am!=1) 
                     %>%dplyr::mutate(
                      a=disp*3,
                      d=cyl*100,b=NA))
                )
        })     %>%arrange(rowid)

Upvotes: 0

davsjob
davsjob

Reputation: 1950

I think I found a neat solution with purrr. It takes a data frame of inputs and then dynamically names new columns a:d with new inputs for each column. First column will use x = 2, y = 100 and z = "a" and then the next row, and so on. The cool thing with functional programming like this is that it is very easy to scale up.

library(tidyverse)

iterate <- tibble(x = c(2, 1.5, 4, 5),
                  y = c(100, 1/100, 100, 1/1000),
                  z = c("a", "b", "c", "d"))

fun <- function(x, y, z) {
  PlantGrowth %>% 
    mutate(!!z := if_else(group == "ctrl", weight * x, weight * y)) %>% 
    select(3)
}

PlantGrowth %>% 
  bind_cols(
    pmap_dfc(iterate, fun)
    ) %>% 
  as_tibble

Which gives you the same df:

# A tibble: 30 x 6
   weight group     a     b     c     d
    <dbl> <fct> <dbl> <dbl> <dbl> <dbl>
 1   4.17 ctrl   8.34  6.26  16.7  20.8
 2   5.58 ctrl  11.2   8.37  22.3  27.9
 3   5.18 ctrl  10.4   7.77  20.7  25.9
 4   6.11 ctrl  12.2   9.17  24.4  30.6
 5   4.5  ctrl   9     6.75  18    22.5

Upvotes: 0

Rohit Das
Rohit Das

Reputation: 2032

Not sure I understand the issue here. If you just want to reduce the verbosity of the code then just create a custom function 

customif = function(x,y) { 
   if_else(group=="ctrl", weight*x, weight*y)
}

then you can call this function in your mutate as 

PlantGrowth %>% mutate (
  a=customif(2,100),
  b=customif(1,5, 1/100),
  c=customif(4, 100),
  d=customif(5, 1/1000)
)

Upvotes: 1

Related Questions