Use starts_with inside map instead of explicitly naming

Question

I want to apply a function inside a tibble but I don't want to explicitly Name the columns. E.G.

library(tidyverse)
library(tidyselect)
test = tibble(var1 = c("la", "le", "lu"), var2 = c("ma", "me", "mu"), var3 = c("fi", "fa", "fu"), dummy=1)
with_funct = test %>% mutate(blub = pmap_chr(list(var1, var2, var3), paste, sep='+'))

I get the expected result:

# A tibble: 3 x 5
  var1  var2  var3  dummy blub    
  <chr> <chr> <chr> <dbl> <chr>   
  1 la    ma    fi        1 la+ma+fi
  2 le    me    fa        1 le+me+fa
  3 lu    mu    fu        1 lu+mu+fu

But instead of writing list(var1, var2, var3) I'd prefer to use starts_with("var") but this does not work out.

So if I use

with_funct = test %>% mutate(blub = pmap_chr(starts_with("var"), paste, sep='+'))

I get an

"Error: No tidyselect variables were registered"

I'd appreciate any help.

score 2 · Accepted Answer · answered Mar 27 '19 at 15:57

You could use select() inside pmap_chr() to work on just the columns that starts with "var". I use the dot to refer to the dataset being used in mutate().

One reasons this works is because pmap() works rowwise across a tibble. The way I'm using this the columns are used in the function (paste() in your case) in the order they appear in the dataset.

test %>% 
    mutate(blub = pmap_chr(select(., starts_with("var")), paste, sep='+'))

# A tibble: 3 x 5
  var1  var2  var3  dummy blub    
  <chr> <chr> <chr> <dbl> <chr>   
1 la    ma    fi        1 la+ma+fi
2 le    me    fa        1 le+me+fa
3 lu    mu    fu        1 lu+mu+fu

score 1 · Answer 2 · answered Mar 27 '19 at 15:57

1

Try this:

with_funct2 = test %>% mutate(blub = pmap_chr(test %>% select(starts_with("var")), paste, sep='+'))

Hope it helps

answered Mar 27 '19 at 15:57

FALL Gora

481
3
8

2

This works fine here, but if there is any grouping or modification to the data before getting to *this* `mutate`, then this will be wrong: it'll be using the original (unmodified `test`) data. Better to use (as aosmith shows) the `.` placeholder instead, as it will include data *as it exists at this point in time in the pipe*. – r2evans Mar 27 '19 at 15:59
1

agreed! more flexible – FALL Gora Mar 27 '19 at 16:01
your Version works, but I cannot just change "test" to "." that throws an error. – MichaelA Mar 27 '19 at 16:12

Use starts_with inside map instead of explicitly naming

2 Answers2