The term "splat operator" comes from Ruby, where the * character (sometimes called the "splat"—see the Jargon File entry) is used to indicate that an entry in an argument list should "soak up" a list of arguments.

CoffeeScript adopted Ruby-style splats very early on (see issue 16), but at Douglas Crockford's suggestion, the syntax was changed from *x to x... a couple of weeks later (see issue 45). Nevertheless, CoffeeScripters still refer to the syntax as the "splat" or "splat operator."

As to what they actually do, splats slice the arguments object in such a way that the splatted argument becomes an array of all "extra" arguments. The most trivial example is

(args...) ->

In this case, args will simply be an array copy of arguments. Splatted arguments can come either before, after, or between standard arguments:

(first, rest...) ->
(rest..., last) ->
(first, rest..., last) ->

In the first two cases, if the function receives 0-1 arguments, rest will be an empty array. In the last case, the function needs to receive more than 2 arguments for rest to be non-empty.

Since JavaScript doesn't allow multiple signatures for functions with the same name (the way C and Java do), splats are a huge time-saver for dealing with varying numbers of arguments.