Atomic types and coercion

R has a small set of atomic types and a long list of rules about how they coerce into each other. Memorising the coercion order is the single most useful five-minute investment in your R fluency.

The five atomic types

• numeric (double) — real numbers, the default for any number you type
• integer — whole numbers, explicit with the L suffix: 5L
• character — strings, double or single quotes
• logical — TRUE or FALSE (also written T or F)
• complex — complex numbers, rare in applied work

NA — the explicit missing value

NA is R's representation of missing data. It is contagious: any arithmetic involving NA produces NA. mean(c(1, 2, NA)) is NA, not 1.5. To skip NAs, pass na.rm = TRUE: mean(c(1, 2, NA), na.rm = TRUE) gives 1.5.

x <- c(1, 2, NA, 4, 5)
mean(x)                # NA
mean(x, na.rm = TRUE)  # 3
is.na(x)               # FALSE FALSE TRUE FALSE FALSE
sum(is.na(x))          # 1 — count of missing values

Implicit coercion

When R combines values of different types in a vector, it coerces them all to the most permissive type. The order, from most to least permissive: character > complex > numeric > integer > logical.

c(1, 2, 3, "text")     # all coerced to character
c(1, 2, TRUE, FALSE)   # logical -> numeric: 1 2 1 0

# Explicit coercion
as.numeric("123")      # 123
as.character(456)      # "456"
as.logical(c(0, 1, 2)) # FALSE TRUE TRUE

typeof vs class

typeof() returns the underlying storage type. class() returns the (possibly user-assigned) class attribute. For most everyday values they agree; for objects (data frames, fitted models), class is what the methods dispatch on.