Dplyr: Allow silent conversion between integer and numeric in if_else() and case_when()

Created on 18 Jan 2017 · 3Comments · Source: tidyverse/dplyr

While I appreciate the type checking done by if_else() and case_when() since it can catch dangerous mistakes, I find the lack of automatic conversion between integer and numeric to be quite cumbersome.

For example, in one survey I am using a variable (here y) is expressed as a number of hours either per day of per week. The time unit is stored in variable x. In order to compute the number of hours per day, one can use if_else to perform the division when needed. Unfortunately, the following code doesn't work when y is of type integer, which is the case in the dBase files I import using read.dbf.

library(dplyr)
df <- data.frame(x=c("Week", "Day"), y=c(1L, 2L))
mutate(df, z = if_else(x == "Week", y/7, y))

# Result:
Error in mutate_impl(.data, dots) : 
  `false` has type 'integer' not 'double'

(In reality, the condition is more complex, so I'd like to use case_when instead of nested if_else.)

Would it be possible to allow conversion of integer columns to numeric? There's no risk of generating incorrect data in that case. In general the difference between numeric and integer is quite hidden in R, so new users are likely to be confused by this error. I'm feeling quite bad to have to explain this to newcomers to perform such a simple operation.

Source