Does if_else always evaluate both 'true' and 'false' regardless of 'condition'?

[jwilliman]

It appears to me that if_else evaluates both the true and false arguments, and then returns only the one that matches the condition argument. This differs from ifelse which appears to only evaluate the true or false argument that matches the condition argument.

This is a problem if if_else is being used to avoid evaluating an expression that is know to fail in certain situations. Is this necessary to ensure "that true and false are the same type."? What would be a tidyverse way of handling this?

x <- list("a", 1, 2, "b")

sapply(x, function(x) ifelse(is.numeric(x), x + 1, NaN)) 
#> [1] NaN   2   3 NaN
sapply(x, function(x) dplyr::if_else(is.numeric(x), x + 1, NaN))
#> Error in x + 1: non-numeric argument to binary operator

^{Created on 2020-06-19 by the reprex package (v0.3.0)}

[hongooi73]

This differs from ifelse which appears to only evaluate the true or false argument that matches the condition argument.

Eh?

> ifelse(c(FALSE, TRUE), 1:2, c(3, stop()))
Error in ifelse(c(FALSE, TRUE), 1:2, c(3, stop())) : 

What does happen is that if condition is only TRUE (or only FALSE), then the result vector that isn't needed won't be evaluated. But you can't rely on this, of course.

[lionel-]

if_else() does not have short-circuiting semantics. Use if () if you need short-circuiting. It is necessary to compute both sides of the condition to figure out the return type. The code you have reported is a type error. In my opinion it is also a bit hard to understand because it relies on a strange mix of scalar predicate and recycling.