monad.clj

(ns monad-explore.core
  (:use clojure.algo.monads))

(defmacro let?
  "Almost the same as let. If you add the :ensure keyword paired with
  some predicate as a var in the let form, let? will not continue
  unless the predicate evaluates to true. (The predicate will have
  access to all bindings above.)"
  [bindings & body]
  (let [[bind [kwd pred & more]] (split-with (complement #{:ensure}) bindings)]
    `(let [~@bind]
       ~@(cond (and kwd more) [`(when ~pred (let? [~@more] ~@body))]
               kwd            [`(when ~pred ~@body)]
               :else body))))

(let? [a 2
       :ensure (= a 2)
       b (+ a 2)
       :ensure (= b 3)]
      (* a b))
;;=> nil

(let? [a 2
       :ensure (= a 2)
       b (+ a 2)
       :ensure (= b 4)]
      (* a b))
;;=> 8

;; I know it's the identity monad, I just wanted to get nil when guards failed vs. undefined.
(defmonad skip-m
  [m-result identity
   m-bind (fn [mv mf] (mf mv))
   m-zero nil])

(domonad skip-m
         [a  2
          :when (= a 2)
          b  (+ a 2)
          :when (= b 3)]
         (* a b))
;;=> nil

(domonad skip-m
         [a  2
          :when (= a 2)
          b  (+ a 2)
          :when (= b 4)]
         (* a b))
;;=> 8

;; OR, without defining anything of my own at all....

(domonad maybe-m
         [a  2
          :when (= a 2)
          b  (+ a 2)
          :when (= b 3)]
         (* a b))
;;=> nil

(domonad maybe-m
         [a  2
          :when (= a 2)
          b  (+ a 2)
          :when (= b 4)]
         (* a b))
;;=> 8

I think I might be blind: I can't see any difference between the form at line 16 and the form at line 23. I'm guessing that first b was supposed to be 3.

Whoops, you're absolutely right. That's what I get for not copy-pasting directly out of the repl. Thanks for the note, the gist should now be correct.

You could use maybe-m instead of defining your own skip-m.

Pretty cool how a little experience with monads leads to elegant code, isn't it.

@jduey, absolutely. I came back after leaving them alone for a few months and everything clicked.

@khinsen, nice. I don't know why I didn't assume that m-zero for maybe-m was nil.