is.numeric within apply vs sapply

Question

In writing R code to identify columns with numeric data, I had some issues and found this question on SO: Is there a vectorized way to use is.numeric?

Within that page I found my answer, but I'm puzzled by the nature of the problem itself. On that page, the following code snip does what I expect:

> dat <- data.frame(v1=1:5,v2=letters[1:5],v3=rnorm(5),v4=c(1,2,'c','d',5))

> sapply(dat,is.numeric)

   v1    v2    v3    v4 
 TRUE FALSE  TRUE FALSE 

But if I change sapply to apply, I get to what caused my problem.

> dat <- data.frame(v1=1:5,v2=letters[1:5],v3=rnorm(5),v4=c(1,2,'c','d',5))

> apply(dat,2,is.numeric)

   v1    v2    v3    v4 
FALSE FALSE FALSE FALSE 

Why do apply and sapply result in different evaluations?

Answer 1

apply coerces everything into the same type. Numbers can become characters but not vice versa. So everything is coerced to character for apply to work.

> apply(dat[,1, drop = F], 2, is.numeric)
  v1 
TRUE 

Answer 2

apply is not advisable for data frames as it will coerce to a matrix. Matrix,as you probably are aware all columns in a matrix must have the same mode(numeric, character, etc.)... by default apply converts the columns into character.