One Function to rule them all

Fork this

https://github.com/iloveponies/one-function-to-rule-them-all

Return of the recursion

Often you want to combine elements of a collection, like calculate the sum or product of a list of numbers, or concatenate a list of strings. That is, we want to make a transformation like this:

    (cons 1 (cons 2 (cons 3 nil)))
;=> (+    1 (+    2 (+    3 0)))

Or like this:

    (cons 1 (cons 2 (cons 3 nil)))
;=> (*    1 (*    2 (*    3 0)))

Our tool for this job was recursion:

(defn sum [a-seq]
  (if (empty? a-seq)
    0
    (+ (first a-seq) (sum (rest a-seq)))))

To make this more efficient, we made this tail recursive using recur and an accumulator:

(defn sum [a-seq]
  (let [sum-helper (fn [acc a-seq]
                     (if (empty? a-seq)
                       acc
                       (recur (+ acc (first a-seq))
                              (rest a-seq))))]
    (sum-helper 0 a-seq)))

This iterative sum would evaluate like this:

;function evaluation                                        accumulator value
(sum              (cons 1 (cons 2 (cons 3 (cons 4 nil))))) ;
;=> (sum-helper 0 (cons 1 (cons 2 (cons 3 (cons 4 nil))))) ; 0
;=> (sum-helper 1 (cons 2 (cons 3 (cons 4 nil))))          ; (+ 0 1) => 1
;=> (sum-helper 3 (cons 3 (cons 4 nil)))                   ; (+ 1 2) => 3
;=> (sum-helper 6 (cons 4 nil))                            ; (+ 3 3) => 6
;=> (sum-helper 10 nil)                                    ; (+ 6 4) => 10
;=> 10

Now lets say that we would like to compute the product of a list of numbers. It would go like this:

(defn product [a-seq]
  (let [product-helper (fn [acc a-seq]
                         (if (empty? a-seq)
                           acc
                           (recur (* acc (first a-seq))
                                  (rest a-seq))))]
    (product-helper 1 a-seq)))

The only two things that changed was the function + that was replaced with * and the initial value 0 that was replaced by 1. So one starts to wonder if there is a pattern behind this all. And it turns out that there is. A function called reduce.

(reduce combining-function initial-accumulator-value a-sequence) takes:

  • a function that is used to combine the current accumulator value and the current element of the parameter sequence
  • an initial value for the accumulator
  • a sequence

So what do we get if we re-define our sum and product with reduce?

(defn sum [a-seq]
  (reduce + 0 a-seq))
(defn product [a-seq]
  (reduce * 1 a-seq))

Let’s see how this sum would evaluate:

(sum            (cons 4 (cons 7 (cons 2 nil))))
;=> (reduce + 0 (cons 4 (cons 7 (cons 2 nil))))
;=> (reduce + (+ 0 4) (cons 7 (cons 2 nil)))    ; accumulator: 0, element: 4
;=> (reduce + 4 (cons 7 (cons 2 nil)))          ; (+ 0 4) => 4
;=> (reduce + (+ 4 7) (cons 2 nil))             ; accumulator: 4, element: 7
;=> (reduce + 11 (cons 2 nil))                  ; (+ 4 7) => 11
;=> (reduce + (+ 11 2) nil)                     ; accumulator: 11, element: 2
;=> (reduce + 13 nil)                           ; (+ 11 2) => 13
;=> 13                                          ; accumulator: 13

So reduce managed to abstract out just the two things that where different in sum and product. The initial value and the operation that composes the values into a one value. Everything else seems to be handled by reduce. So how does it work?

(defn reduce [f initial a-seq]
  (if (empty? a-seq)
    initial
    (recur f
           (f initial (first a-seq))
           (rest a-seq))))

Lets go through an evaluation of reduce:

    (reduce + 0                   [1 2 3])
;=> (reduce + (+ 0 1)             [2 3])
;=> (reduce + (+ (+ 0 1) 2)       [3])
;=> (reduce + (+ (+ (+ 0 1) 2) 3) [])
;=> (+ (+ (+ 0 1) 2) 3)
;=> 6

Time to put the good tool to use.

Exercise 1

Write the function (concat-elements a-seq) that takes a sequence of sequences and concatenates them together with concat.

Don’t use apply to implement this function.

(concat-elements [])            ;=> ()
(concat-elements [[1 2]])       ;=> (1 2)
(concat-elements [[1 2] [3 4]]) ;=> (1 2 3 4)

Two Sides of a Coin

One can call reduce in two different ways:

(reduce combinator-function initial-accumulator-value input-sequence)

or

(reduce combinator-function                           input-sequence)

If input-sequence is not empty, then the second form works like this:

    (reduce f (cons elem rest-of-seq))
;=> (reduce f elem rest-of-seq)

that is, it uses the first element of the parameter sequence as the initial accumulator value.

And if input-sequence is empty, then:

  • The first form just returns initial-accumulator-value
  • The second form returns (combinator-function), that is, it calls combinator-function with zero parameters.

Let’s try this version without initial value:

(defn seq-min [a-seq]
  (reduce min a-seq))

(seq-min [1])       => 1
(seq-min [5 3 2 6]) => 2
(seq-min [])
  => java.lang.IllegalArgumentException:
     Wrong number of args (0) passed to: core$min (NO_SOURCE_FILE:0)

Now the call with empty sequence results in an exception, since min is not defined for 0 arguments. Since seq-min used reduce without an initial value, it called (min) when it was given an empty sequence.

Lets look at an evaluation of a reduce call without an initial value.

(seq-min [5 3 2 6])
;=> (reduce min           [5 3 2 6])
;=> (reduce min 5         [3 2 6]) ; Use the first element as the initial value
;=> (reduce min (min 5 3) [2 6])
;=> (reduce min 3         [2 6])
;=> (reduce min (min 3 2) [6])
;=> (reduce min 2         [6])
;=> (reduce min (min 2 6) [])
;=> (reduce min 2         [])
;=> 2

Exercise 2

Write the function (str-cat a-seq) that takes a sequence of strings and catenates them with one space character between each.

(str-cat ["I" "am" "Legend"])  ;=> "I am Legend"
(str-cat ["I" "am" "back"])    ;=> "I am back"
(str-cat ["more" " " "space"]) ;=> "more   space"
(str-cat [])                   ;=> ""
You probably want to handle the special case of empty parameter outside reduce.

Exercise 3

Write the function (my-interpose x a-seq) that places x between every element of a-seq.

Keep in mind how conj works for vectors.

(my-interpose 0 [1 2 3])               ;=> (1 0 2 0 3)
(my-interpose "," ["I" "me" "myself"]) ;=> ("I" "," "me" "," "myself")
(my-interpose :a [1])                  ;=> (1)
(my-interpose :a [])                   ;=> ()

Let’s look at another example. We implemented the function count-elem in Recursion, which counts the occurrences of an element in a sequence. Let’s reimplement that function with reduce:

(defn count-elem [elem a-seq]
  (let [counter (fn [count e]
                  (if (= e elem)
                    (inc count)
                    count))]
    (reduce counter 0 a-seq)))
(count-elem :D [13 "\o/" :D :$ :D [:D] :< "~^._.^~"])
;=> (reduce counter 0                 [13 "\o/" :D :$ :D [:D] :< "~^._.^~"])
;=> (reduce counter (counter 0 13)    ["\o/" :D :$ :D [:D] :< "~^._.^~"])
;=> (reduce counter 0                 ["\o/" :D :$ :D [:D] :< "~^._.^~"])
;=> (reduce counter (counter 0 "\o/") [:D :$ :D [:D] :< "~^._.^~"])
;=> (reduce counter 0                 [:D :$ :D [:D] :< "~^._.^~"])
;=> (reduce counter (counter 0 :D)    [:$ :D [:D] :< "~^._.^~"])
;=> (reduce counter 1                 [:$ :D [:D] :< "~^._.^~"])
;=> (reduce counter (counter 1 :$)    [:D [:D] :< "~^._.^~"])
;=> (reduce counter 1                 [:D [:D] :< "~^._.^~"])
;=> (reduce counter (counter 1 :D)    [[:D] :< "~^._.^~"])
;=> (reduce counter 2                 [[:D] :< "~^._.^~"])
;=> (reduce counter (counter 2 [:D])  [:< "~^._.^~"])
;=> (reduce counter 2                 [:< "~^._.^~"])
;=> (reduce counter (counter 2 :<)    ["~^._.^~"])
;=> (reduce counter 2                 ["~^._.^~"])
;=> (reduce counter (counter 2 "~^._.^~") [])
;=> (reduce counter 2 [])
;=> 2

Exercise 4

Write the function (my-count a-seq) that returns the length of a sequence.

Do not use count in your implementation.

(my-count [])      ;=> 0
(my-count [1 2 3]) ;=> 3
(my-count [1])     ;=> 1

Exercise 5

Write the function (my-reverse a-seq) that reverses a sequence.

(my-reverse [1 2 3]) ;=> (3 2 1)
(my-reverse [1 2])   ;=> (2 1)
(my-reverse [])      ;=> ()

Exercise 6

Write the function (min-max-element a-seq) that returns the maximal and minimal elements of a-seq in a vertor like [min max].

(min-max-element [2 7 3 15 4]) ;=> [2 15]
(min-max-element [1 2 3 4])    ;=> [1 4]
(min-max-element [1])          ;=> [1 1]

Exercise 7

Write the function (insert sorted-seq n) that adds the number n into a sorted sequence of number. The ordering of the sequence must be preserved.

You don’t need to use reduce for this, and you probably don’t want to.

(insert [] 2)      ;=> (2)
(insert [1 3 4] 2) ;=> (1 2 3 4)
(insert [1] 2)     ;=> (1 2)

Now implement (insertion-sort a-seq) using reduce and the function insert.

(insertion-sort [2 5 3 1]) ;=> (1 2 3 5)
(insertion-sort [1 2])     ;=> (1 2)

Exercise 8

Write the fuction (parity a-seq) that picks into a set those elements of a-seq that occur odd number of time.

(parity [:a :b :c])    ;=> #{:a :b :c}
(parity [:a :a :b :b]) ;=> #{}
(parity [1 2 3 1])     ;=> #{2 3}

Var…argghhhh!

As you may have noticed, some functions in Clojure can take variable amount of parameters.

(+ 1 1)       ;=> 2
(+ 1 1 1 1 1) ;=> 5

Not surprisingly, one can define such functions themselve.

(defn one-or-two
  ([x] 1)
  ([x y] 2))

This function has two separate definitions. Each definition is separated from the others by enclosing it in parentheses. Each of the definitions takes a different amount of parameters. This is called arity overloading. In English, the right definition is chosen based on how many parameters were given to the function. This function can be called with one or two parameters.

(one-or-two :a)  ;=> 1
(one-or-two 5 6) ;=> 2
(one-or-two 1 2 3)
;=> ArityException
;     Wrong number of args (3) passed to: user$one-or-two
;     clojure.lang.AFn.throwArity (AFn.java:437)

Exercise 9

Write the function minus that takes one or two parameters.

  • If given a one parameter \(x\), it returns \(-x\).
  • If given to parameters \(x\) and \(y\), it returns \(x - y\).
(minus 2)   ;=> -2
(minus 4 3) ;=> 1

But what if you don’t know the amount of parameters given before hand? Lets look at a definition of max:

(defn max
  ([x] x)                         ; one parameter
  ([x y] (if (> x y) x y))        ; two parameters
  ([x y & more]                   ; more than two parameters
    (reduce max (max x y) more)))

First of all, the function has three separate definitions. If we call the function with one or two parameters, the first and second definitions are used, respectively.

The third definition of max is using variable arguments. It means that the function can be called with any amount of parameters. The first two parameters are given names x and y and the rest of the parameters are grouped into a sequence and given a name more. Lets look at an evaluation:

(max 1 2 3 4)
;=> (reduce max (max 1 2) '(3 4))
;=> (reduce max 2         '(3 4))
;=> (reduce max (max 2 3) '(4))
;=> (reduce max 3         '(4))
;=> (reduce max (max 3 4) '())
;=> (reduce max 4         '())
;=> 4

Exercise 10

Write the function count-params that accepts any number of parameters and returns how many it was called with. You need only a one definition for this.

(count-params)            ;=> 0
(count-params :a)         ;=> 1
(count-params :a 1 :b :c) ;=> 4

Exercise 11

Write the function my-* that takes any number of parameters.

  • If no parameters are given, return 1
  • If one parameter \(x\) is given, return \(x\).
  • If two parameters \(x\) and \(y\) are given, return \(xy\).
  • If more than two parameters \(x\), \(y\), \(\dots\) are given, return their product \(x \cdot y \cdots\).

You are free to use *, but not apply.

(my-*)           ;=> 1
(my-* 4 3)       ;=> 12
(my-* 1 2 3 4 5) ;=> 120

Exercise 12

Remember the function pred-and that you implemented in Predicates? Write a new definition for it that works for any amount of parameters.

  • If no parameters are given, return a predicate that always returns true.
  • If only one predicate p is given, return p.
  • If two predicates are given, return a predicate that returns true if both of them return true and false otherwise.
  • If more than two predicates are given, return a predicate that returns true only if all of the predicates return true and false otherwise.
(filter (pred-and) [1 0 -2])                    ;=> (1 0 -2)
(filter (pred-and pos? odd?) [1 2 -4 0 6 7 -3]) ;=> (1 7)
(filter (pred-and number? integer? pos? even?)
        [1 0 -2 :a 7 "a" 2])                    ;=> (0 2)

Encore

The next exercise is a bit trickier.

Exercise 13

3 points

Write the function my-map that works just like standard map. It takes one or more sequences and a function f that takes as many parameters as there are sequences.

(my-map inc [1 2 3 4])                  ;=> (2 3 4 5)
(my-map + [1 1 1] [1 1 1] [1 1 1])      ;=> (3 3 3)
(my-map vector [1 2 3] [1 2 3] [1 2 3]) ;=> ((1 1 1) (2 2 2) (3 3 3))

Sudoku →