The apply Family of Functions

The apply family — sapply(), lapply(), apply(), mapply(), and vapply() — provides a functional alternative to explicit for loops by applying a function to every element of a vector, list, or margin of a matrix, and returning the results collected into a single object. Choosing the right member of the family mostly comes down to what shape of input you have (a vector/list versus a matrix/data frame) and what shape of output you want (a list, a simplified vector, or a guaranteed type).

lapply(X, FUN) always returns a list, with one element per input element, preserving names if X is named, which makes it the predictable, type-safe workhorse of the family; sapply(X, FUN) calls lapply() internally and then attempts to simplify the result into a vector or matrix when every element has the same length, but silently falls back to returning a plain list when the results aren't uniform, which is convenient interactively but risky inside production code because the output type can vary depending on the data. lapply(mylist, function(x) x^2) on a list of numeric vectors returns a list where each element has been squared.

apply(X, MARGIN, FUN) is specifically for matrices and data frames, applying FUN across rows when MARGIN = 1 or across columns when MARGIN = 2 (and both simultaneously with MARGIN = c(1, 2)), which makes it the natural tool for computing something like a row-wise total or a column-wise mean across a whole table. Because apply() coerces its input to a matrix first, applying it to a data frame with mixed column types (numeric and character together) will silently coerce everything to character, which is a common source of subtle bugs, so apply() is best reserved for genuinely homogeneous, all-numeric matrices.

vapply(X, FUN, FUN.VALUE) is the type-safe cousin of sapply(): it requires you to specify FUN.VALUE, a template describing the exact type and length FUN should return for each element (like numeric(1) for a single number), and it throws an informative error immediately if any call produces something that doesn't match that template, which makes it much safer to use inside functions or packages where an unexpected input could otherwise cause sapply() to return a differently-shaped result. mapply(FUN, X, Y) is the multivariate version, iterating over two or more vectors in parallel and passing corresponding elements of each to FUN simultaneously, similar to calling FUN(X[1], Y[1]), FUN(X[2], Y[2]), and so on.