case_when() evaluating all RHS expressions except only the ones matching LHS

[frederikziebell]

According to one example in the function documentation

# case_when() evaluates all RHS expressions, and then constructs its
# result by extracting the selected (via the LHS expressions) parts.
# In particular NaN are produced in this case:
y <- seq(-2, 2, by = .5)
case_when(
  y >= 0 ~ sqrt(y),
  TRUE   ~ y
)

Wouldn't it be better to only evaluate the RHS for the instances where the LHS is TRUE and then paste these evaluations together?

If there are five cases evenly distributed across the vector, this means five times more computations. Yes, vectorization is fast, but it's faster to not do computations one is not interested in anyway.

[hadley]

This is currently needed for type stability.