0

From a character column in a tibble, I'm trying to extract numbers and convert to numeric type using ifelse() to apply as.numeric() on values that do not contain letters (or "+").

While this example works, it returns a warning message:

library(dplyr)

tibble(a = c("0", "0.1", "abc")) %>% 
  mutate(b = ifelse(!grepl("[a-zA-Z\\+]+", a), as.numeric(a), NA))

# A tibble: 3 × 2
  a         b
  <chr> <dbl>
1 0       0  
2 0.1     0.1
3 abc    NA  
Warning message:
Problem while computing `b = ifelse(grepl("[a-zA-Z\\+]+", a), NA, as.numeric(a))`.
ℹ NAs introduced by coercion

Please help me understand why my script is still trying to convert the character value to numeric, and how I can fix this.

Gregor Thomas
  • 136,190
  • 20
  • 167
  • 294
Paul
  • 2,877
  • 1
  • 12
  • 28

1 Answers1

1

As explained in this strongly related question: Does ifelse really calculate both of its vectors every time?, ifelse (generally) calculates both responses for every entry, and then puts the appropriate ones together into the result vector. So even though as.numeric("abc") isn't in your results, it still tries to calculate it.

Which means that there probably isn't much point in using ifelse() here. Might as well simplify to b = as.numeric(a), or if you want to avoid the warnings b = suppressWarnings(as.numeric(a)).

Gregor Thomas
  • 136,190
  • 20
  • 167
  • 294
  • Thanks Gregor, interesting read. I had tried case_when(), which also returned the warning. Work around: use the ifelse() to copy the contents of a to b if the test is met, other wise b is NA. Then apply 'as.numeric()' to b. – Paul Feb 01 '23 at 22:08
  • Sure, that works too (though it seems equivalent and less efficient than `suppressWarnings()`) – Gregor Thomas Feb 02 '23 at 14:30