_ as a variable name means, "(match anything here, doesn't matter)". It's similar to . in a regular expression -- you're testing a list of multiple values, and

_ as a variable name means, "(match anything here, doesn't matter)". It's similar to . in a regular expression -- you're testing a list of multiple values, and while for your purposes you only care about a specific one of them, you need to verify that it has the right number of values. Much like "..e" will match any three-letter word ending in e, (_, 21, _) will match any three-value list whose second value is 21. ( (_, x, _) will match a three-value list parametrically, and extract the variable x for further use.) That's probably a somewhat confusing explanation, but in practice, idiomatic code using deconstructive pattern matching is usually quite a bit clearer than the alternative, several layers of nested switch statements. It's also incredibly useful. Haskell has the same feature. (I'm pretty sure Erlang does, too.) For one good example of this making complex code relatively simple, check out Chris Okasaki's thesis (Purely Functional Data Structures, http://www.cs.cmu.edu/~rwh/theses/okasaki.pdf http://www.cs.cmu.edu/~rwh/theses/okasaki.pdf), page 79, or the red-black tree balanced-insertion code in the book. (Note: That's SML, which is a different dialect of ML.) In this case: switch on sys.argv (command line args) -- if it's a two-variable array, then ignore the first parameter, put the second in a variable called 'q', and then display the result of executing search(q). Otherwise, raise an exception with that error message.