data.table fifelse giving wrong warning?

Question

I found warning difference using fifelse from data.table library:

set.seed(123)
df <- data.table(ID = rep(1:10,each = 2),x = sample(c(1,NA),20,replace = T))

test1 <- df[,fifelse(any(!is.na(x)),max(x,na.rm = T),as.numeric(NA)),by = ID]

produces a warning:

Warning messages 1: In max(x, na.rm = T) : no non-missing arguments to max; returning -Inf 2: In max(x, na.rm = T) : no non-missing arguments to max; returning -Inf

while:

test2 <- df[,ifelse(any(!is.na(x)),max(x,na.rm = T),as.numeric(NA)),by = ID]

don't. And the two results are identical:

identical(test1,test2)
[1] TRUE

And there is no -Inf in the result. What does this mean ?

Just a guess. The error message is coming from the max() function. I would infer from this, that with fifelse, the function max() is running for all the groups, even when the condition any(!is.na(x)) is FALSE, while with ifelse it does run only when it is TRUE? — djourd1

akrun akrun · Accepted Answer · 2021-01-21T21:23:36

It may be better to use if/else as the input is of length 1

df[, if(any(!is.na(x))) max(x, na.rm = TRUE) else NA_real_, by = ID]

data.table fifelse giving wrong warning?

1 Answers