Xor macro in clojure, with tests

xor.clj

(ns xor 
  (:use clojure.test))

(defmacro xor
  "Evaluates exprs one at a time, from left to right.  If only one form returns
  a logical true value (neither nil nor false), returns true.  If more than one
  value returns logical true or no value returns logical true, retuns a logical
  false value.  As soon as two logically true forms are encountered, no
  remaining expression is evaluated.  (xor) returns nil."
  ([] nil)
  ([f & r]
     `(loop [t# false f# '[~f ~@r]]
        (if-not (seq f#) t#
                (let [fv# (eval (first f#))]
                  (cond
                   (and t# fv#) false
                   (and (not t#) fv#) (recur true (rest f#))
                   :else (recur t# (rest f#))))))))

(def test-output (ref []))
(defn- doit
  "Returns val and creates the side effect of printing num."
  [num val]
  (dosync
   (alter test-output conj num))
  val)

(defn get-test-output
  "Returns test-output content and clears it"
  []
  (let [t @test-output]
    (dosync (ref-set test-output []))
    t))

(deftest nullary
  (is (not (xor))))

(deftest unary
  (is (get-test-output))
  (is (xor (doit 1 true)))
  (is (= [1] (get-test-output)))
  (is (not (xor (doit 1 nil))))
  (is (= [1] (get-test-output))))

(deftest binary  
  (is (not (xor (doit 1 nil) (doit 2 nil))))
  (is (= [1 2] (get-test-output)))
  (is (not (xor (doit 1 true) (doit 2 true))))
  (is (= [1 2] (get-test-output)))
  (is (xor (doit 1 nil) (doit 2 true)))
  (is (= [1 2] (get-test-output)))
  (is (xor (doit 1 true) (doit 2 nil)))
  (is (= [1 2] (get-test-output))))

(deftest ternary
  (is (not (xor (doit 1 nil) (doit 2 nil) (doit 3 nil))))
  (is (= [1 2 3] (get-test-output)))
  (is (xor (doit 1 nil) (doit 2 nil) (doit 3 true)))
  (is (= [1 2 3] (get-test-output)))
  (is (xor (doit 1 nil) (doit 2 true) (doit 3 nil)))
  (is (= [1 2 3] (get-test-output)))
  (is (not (xor (doit 1 nil) (doit 2 true) (doit 3 true))))
  (is (= [1 2 3] (get-test-output)))
  (is (xor (doit 1 true) (doit 2 nil) (doit 3 nil)))
  (is (= [1 2 3] (get-test-output)))
  (is (not (xor (doit 1 true) (doit 2 nil) (doit 3 true))))
  (is (= [1 2 3] (get-test-output)))
  (is (not (xor (doit 1 true) (doit 2 true) (doit 3 nil))))
  (is (= [1 2] (get-test-output)))
  (is (not (xor (doit 1 true) (doit 2 true) (doit 3 true))))
  (is (= [1 2] (get-test-output))))

(deftest n-ary  
  (is (not (xor (doit 1 nil) (doit 2 nil) (doit 3 nil)
                (doit 4 nil) (doit 5 nil) (doit 6 nil))))
  (is (= [1 2 3 4 5 6] (get-test-output)))
  (is (not (xor (doit 1 nil) (doit 2 true) (doit 3 nil)
                (doit 4 true) (doit 5 nil) (doit 6 nil))))
  (is (= [1 2 3 4] (get-test-output)))
  (is (xor (doit 1 true) (doit 2 nil) (doit 3 nil)
           (doit 4 nil) (doit 5 nil) (doit 6 nil)))
  (is (= [1 2 3 4 5 6] (get-test-output)))
  (is (not (xor (doit 1 true) (doit 2 true) (doit 3 true)
                (doit 4 true) (doit 5 true) (doit 6 true))))
  (is (= [1 2] (get-test-output)))
  (is (not (xor (doit 1 nil) (doit 2 nil) (doit 3 nil)
                (doit 4 nil) (doit 5 true) (doit 6 true))))
  (is (= [1 2 3 4 5 6] (get-test-output))))


(run-tests)

XOR has to return true if an odd number of its args are true, not if there's only one truthy argument.

XOR has to return true if an odd number of its args are true, not if there's only one truthy argument.

That is one definition of how xor should work. Another (this) implementation is the result of the XOR is true if and only if 1 input is true. Neither is wrong it just depends on how it is implemented.

XOR has to return true if an odd number of its args are true, not if there's only one truthy argument.

Wikipedia seems to side with you: https://en.wikipedia.org/wiki/Exclusive_or