Hacker News Reader

Comment by akkartik | Hacker News Reader

Ah, you're right. It's a list but not a cons.

  $ sbcl
  * (symbolp nil)
  T
  * (listp nil)
  T
  * (consp nil)
  NIL

Is there a boolean predicate? My Common Lisp is rusty and I don't see one at https://www.cs.cmu.edu/Groups/AI/html/cltl/clm/node73.html#S....

View on HN

Also, the type of NIL in Common Lisp is NULL. There is a NIL type, but it is empty: there exists no value which is an instance of the NIL type. NIL is a subtype of every type (including itself), and the type T is a supertype of every type, including itself. This situation with NIL and T on opposite ends of the "type spindle" is quite clever and subtly better than making NIL a type which contains NIL.

Thanks for the pointer! I'd never seen http://www.lispworks.com/documentation/lw70/CLHS/Body/t_null... before, in spite of poring over the CLHS for many hours over almost a decade.

Can you suggest any example Common Lisp programs that print the _types_ NULL or NIL or T? Or show off the benefits of this spindle arrangement? I can intuitively see the benefits of this approach, but an example makes things much more memorable.

The nice thing about T is that it's the first letter of True and of Type. E.g.

  ;; catch-all method for any type
  (defmethod foo ((arg t)) ...)

  ;;
  (defmethod foo (arg ...) ...) ;; alternative spelling: omit class

NIL is the empty list and that is mnemonic for "empty type". For instance, the intersection of two mutually exclusive types is the empty type. Nothing can be both a cons and an atom, so the type denoted by the type expression (and cons atom) is empty; and that is nil. The set manipulation which underlies types needs an empty set, and nil fill the need for that empty set (pun intended).

NULL is useful in writing methods that capture NIL (the only instance of NULL). That may be recognized as the "null object pattern":

  (defmethod foo ((arg null))
    ;; (foo nil) called
    )

  (defmethod foo ((arg number))
    ;; (foo <number>) called
    )

  (defmethod foo ((arg t))
    ;; (foo <anything else>) called
    )

> Is there a boolean predicate?

Nope, because everything is a Boolean: NIL is false, and everything else is true.

This is not quite precisely true. Everything is a generalized boolean, but not a boolean. Only T and NIL are of type BOOLEAN, which you can check with the function:

    (defun booleanp (x) (TYPEP x 'BOOLEAN))

You're right, of course. Still, 'everything is a boolean' is close enough for government work:-)

Reply on news.ycombinator.com