Next, the book discusses the main Lisp data structure: the list. It discusses how Lisp assumes that in any list, the first item in that list is a command of some sort, and to change that behavior, you place a single quote, ', before the list. Clojure has the same expectation, so that if you enter this at a REPL:

(foo bar baz)

you will get the following error,

CompilerException java.lang.RuntimeException: Unable to resolve symbol: foo in this context...

indicating that it tried to execute that list as a function call, but failed to find a function called “foo”. Just as in Lisp, you can quote the list by prefixing it with a single quote. Thus, this

'(foo bar baz)

is now treated as data, instead of a function call. Similarly, this is a function call

(expt 2 3)

is a call to the expt function, while this

'(expt 2 3)

is a list with three elements.

Barski then goes into a length discussion about the list and its structure, specifically the cons cell. In Lisp, a cons cell is a pair of items, and its list is implemented as a linked list, with the second part of each cons cell linking to the next cons cell in the list. Clojure does not have a cons cell, and so does not implement its lists in terms of them. Lisp uses cons cells individually as simple pairs, too, and you can simulate this use by using a 2-element vector, like so

(def pair [:a :b])

Beginning on page 38, Barski starts discussing the various functions that operate on lists. I will provide the Clojure equivalent functions, where possible.

The cons Function

In Lisp, you create cons cells using the cons function. The result is a cons cell with the two arguments set as the left and right portions of the cell, or just a left-side, if the right is nil. You can simulate this, using the vector function

(vector 'chicken 'cat)

or as a literal vector

['chicken 'cat]

Clojure does have a cons function, but it functions somewhat differently than its Lisp counterpart. It prepends its first argument onto its second argument, which must be a list-like thing, returning the result as a list. So this in Lisp

(cons 'chicken 'cat)

can’t be done using Clojure’s cons function, since the second argument is not a list-like thing. If you try, you will get this error

IllegalArgumentException Don't know how to create ISeq from: clojure.lang.Symbol clojure.lang.RT.seqFrom (RT.java:505)

If you need to create a sequence of a chicken and a cat, use one of these:

(vector 'chicken 'cat)
['chicken 'cat]
'(chicken cat)

Obviously, the first two create vectors, while the last one create a list.

In Lisp, the empty list and nil are equivalent, but not so in Clojure. However, these two examples of consing a value and and empty list, and a value and nil, behave essentially the same, returning a list of just the first element.

(cons 'chicken 'nil) ; (chicken)
(cons 'chicken ())   ; (chicken)

The next three examples function effectively the same, though remember that what you are getting back is not a linked list of cons cells, as you would in Lisp.

(cons 'pork '(beef chicken))           ; (pork beef chicken)
(cons 'beef (cons 'chicken '()))       ; (beef chicken)
(cons 'pork 
      (cons 'beef (cons 'chicken ()))) ; (pork beef chicken)

The car and cdr Functions

Lisp has the basic functions car and cdr for accessing the first element of a lisp, and the remaining elements of a list, respectively. These names are directly related to the original hardware on which Lisp ran. Clojure does not have functions with these names, but it does provide its own functions that give the same results.

Instead of car, Clojure gives us first, which actually is a better name for what it does. It works like this

(first '(pork beef chicken)) ; pork

Instead of cdr, Clojure actually has two functions, next and rest. With a listy thing of 2+ elements, both these function behave the same, returning everything but the first element

(rest '(pork beef chicken)) ; (beef chicken)
(next '(port beef chicken)) ; (beef chicken)

They differ, however, on what happens if there’s nothing after the first element, or the entire list is the empty list. next will return nil in both these cases, while rest will return an empty list. As I said, in Lisp, these are the same thing, but in Clojure, they are very different. The empty list is truthy, while nil is not

(when nil :truthy) ; nil
(when '() :truthy) ; :truthy

The c*r Functions

Lisp defines combinations of car and cdr to allow you to get the second item, the third, the second item from the remainder of the remainder of a list, etc. Most CL implementations provide these functions up to four levels deep. Here’s a partial list to illustrate

(cadr '(pork beef chicken))          ; BEEF
(cdar '((peas carrots tomatoes)
        (pork beef chicken)))        ; (CARROTS TOMATOES)
(cddr '((peas carrots tomatoes)
        (pork beef chicken) duck))   ; (DUCK)
(caddr '((peas carrots tomatoes)
         (pork beef chicken) duck))  ; DUCK
(cddar '((peas carrots tomatoes)
         (pork beef chicken) duck))  ; (TOMATOES)
(cadadr '((peas carrots tomatoes)
          (pork beef chicken) duck)) ; BEEF

And let’s be honest, keeping all those ‘a’s and ‘d’s straight can be pretty confusing (at least, it is to me).

Since Clojure doesn’t have car and cdr, it also lacks these functions. It does have one analog function, and that is for the CL function cadr, which provides the car of the cdr of a list, also known as the second item. Thus, Clojure has second that does the same thing,

(second '(pork beef chicken)) ; beef

but that’s it for built-ins. You can roll your own versions of these CL functions, but from what I’ve read, this is considered a code smell. Even though I don’t think you should try to implement these functions, let me show you how you might do it, if you wanted to.

Instead of explaining each one, and its possible implementation in Clojure, I’m just going to include the code for all of them in one, big blob.

(defn cdar
  [lst]
  (rest (first lst)))

(cdar '((peas carrots tomatoes)
        (pork beef chicken)))       ; (carrots tomatoes)

(defn cddr
  [lst]
  (rest (rest lst)))

(cddr '((peas carrots tomatoes)
        (pork beef chicken) duck))  ; (duck)

(defn caddr
  [lst]
  (first (rest (rest lst))))

(caddr '((peas carrots tomatoes)
         (pork beef chicken) duck)) ; duck

(defn cddar
  [lst]
  (rest (rest (first lst))))

(cddar '((peas carrots tomatoes)
         (pork beef chicken) duck)) ; (tomatoes)

(defn cadadr
  [lst]
  (first (rest (first (rest lst)))))

(cadadr '((peas carrots tomatoes)
          (pork beef chicken) duck)) ; beef

So, if you shouldn’t create analogous functions, how do you get at the elements you need? One way would be to use Clojure’s destructuring. I’m not going to go into a full explanation of how destructuring works, but suffice it to say that it’s a way to assign elements of a listy thing to individual locals.

Here’s how you could achieve the same results as cdar

(let [[fst] '((peas carrots tomatoes) (pork beef chicken))]
  (rest fst)) ; (carrots tomatoes) ; instead of cdar

As you can see, this code assigns the first element '(peas carrots tomatoes) to the local fst. It then calls rest on that, which returns the remainder of that list.

Here, then, is a code dump of destructuring implementation of the rest of the c*r functions that I listed earlier.

(let [[fst] '((peas carrots tomatoes)
              (pork beef chicken))]
  (rest fst)) ; (carrots tomatoes) ; instead of cdar

(let [[_ & rst] '((peas carrots tomatoes)
                  (pork beef chicken) duck)]
  (rest rst)) ; (duck) ; instead of cddr

(let [[_ & rst] '((peas carrots tomatoes)
                  (pork beef chicken) duck)
      tail (rest rst)]
  (first tail)) ; duck ; instead of caddr

(let [[fst] '((peas carrots tomatoes)
              (pork beef chicken) duck)
      tail (rest fst)]
  (rest tail)) ; (tomatoes) ; instead of cddar

(let [[_ & rst] '((peas carrots tomatoes)
                  (pork beef chicken) duck)
      fst (first rst)
      tail (rest fst)]
  (first tail)) ; beef ; instead of cadadr

That’s It, For Now

OK, that brings us to the end of chapter 3. I think this may be the longest post so far. I will move on to chapter 4, and will be back soon with the next installment.

As always, full code is available at https://github.com/joeygibson/lolclojure.

Defining Functions

In Lisp, you use defun, but in Clojure, it’s defn. Here’s a square function in Lisp.

(defun square (n) 
  (* n n))

And here’s the same function in Clojure. Notice that the function arguments are enclosed in square brackets (it’s a vector), instead of parens.

(defn square
  "Returns the square of the passed-in number"
  [n]
  (* n n))

That string is known as the docstring. It stays with the function, and is available in the REPL by running the doc function, like this (doc square). Lisp also supports docstrings in functions, but it comes after the argument list, instead of before. While docstrings are optional, I highly encourage you to include them. They can span multiple lines, and since they stay with your function, they are useful from the REPL.

Equality

In Lisp, there are may functions for determining equality, and you have to choose the right one for any given circumstance. Among these are eq, equal, equalp, and a few others. In Clojure, there’s just =. If you’re coming from Java, you know = by itself is assignment, not an equality check. For that, you have to use ==, but even that only computes reference equality, and is not always what you want. In Clojure, = does everything you want, in every circumstance. It is your friend.

Exponentiation

Starting on page 34, there are a few examples using the expt function, which raises its first argument to the power of its second. This is a built-in function in CL, but Clojure doesn’t have one. You could use Math.pow from Java, but this only works with doubles, and once the numbers get really large, it switches to scientific notation.

(Math/pow 2 2)       ; 4.0
(Math/pow 2 3)       ; 8.0
(Math/pow 2.0 3)     ; 8.0
(Math/pow 2 10)      ; 1024.0
(Math/pow 53N 53)    ; 2.4356848165022712E91

(In case you haven’t seen it, appending an N to a number literal causes the number to be of type clojure.lang.BigInt. Appending an M makes it a java.math.BigDecimal.)

You can write your own exponentiation function that gives better results than using the one from Java. Here are two different ways to write it. Both versions are tail-recursive, which means they won’t exhaust the stack, but the first uses a nested function, while the second is recursive on a loop. Here’s the nested function version

(defn expt
  "Raise x to the nth power"
  [x n]
  (letfn [(rexpt [n acc]
                 (if (zero? n)
                   acc
                   (recur (dec n) (* acc x))))]
    (rexpt n 1)))

Notice the letfn that contains a local function called rexpt. This function does all the work, and is called as the last line of the main function. It takes a parameter to be used as an accumulator, and this is returned once the exponent is decremented to zero. This nested function is also a closure, because the value of x is referenced directly. We don’t need to change it like we do n, so we just use its name.

Now, here is the version that uses loop. While CL has a loop macro, Clojure’s loop is completely different. All it does is provide a recursion point. This means that when you use the recur function later, execution will jump back to where the loop call is, instead of back to the beginning of the function. The locals declared in the loop’s vector are rebound with the values specified by the recur call. I think this version is easier to understand than the first one.

(defn expt
  "Raise x to the nth power"
  [x n]
  (loop [n n
         acc 1]
    (if (zero? n)
      acc
      (recur (dec n) (* acc x)))))

Notice that the code inside the loop is identical to that in the rexpt local function from the previous example. It’s just not wrapped inside another function. Also of note is in the let we assign n to n. This is a common technique, and will result in a local called n being assigned the value of the passed-in n. The local n can then be decremented with each recursion, without affecting the outer n.

Both of these function provide identical results.

(expt 2 2)       ; 4
(expt 2 3)       ; 8
(expt 2.0 3)     ; 8.0
(expt 2 10)      ; 1024
(expt 53N 53)    ; 24356848165022712132477606520104725518533453128685640844505130879576720609150223301256150373

Notice that passing in an integer results in an integer. Passing in a double results in a double. And passing in a BigInt results in a very large number (Hint: scroll horizontally… it goes on for a while).

Printing Things

CL uses (princ), (prin1), (print), etc., to output things to the console. In Clojure, you use (print) and (println).

(print "He yelled \"Stop that thief!\" from the busy street.") ; no newline
(println "He yelled \"Stop that thief!\" from the busy street.") ; newline

(print) outputs the string, but does not append a newline. (println) appends a newline, as you would expect, given its name.

To Be Continued…

This has gotten very long, so I will stop now, and continue in another post. Stay tuned for Chapter Three, Part Two.

This Lisp code

(let ((a 5)
	  (b 6))
  (+ a b))

looks like this in Clojure

(let [a 5
      b 6]
  (+ a b))

I think the Clojure way is a little easier to read.

The biggest difference between the two is when it comes to local functions. CL has flet for defining local functions, and labels for defining local functions that need to be able to call each other (or call themselves recursively). Here’s an example of each

;; A single local function
(flet ((f (n)
		  (+ n 10)))
  (f 5))

;; Two local functions that don't need to call each other
(flet ((f (n)
		  (+ n 10))
	   (g (n)
		  (- n 3)))
  (g (f 5)))

Here’s the equivalent code in Clojure. Note how the square brackets make things stand out a little bit more.

(letfn [(f [n]
           (+ n 10))]
  (f 5))

(letfn [(f [n]
           (+ n 10))
        (g [n]
           (- n 3))]
  (g (f 5)))

If the functions need to reference each other, in CL you have to use labels, instead of flet. Here’s how that looks (the only difference is the form used; the arguments remain the same)

(labels ((a (n)
			(+ n 5))
		 (b (n)
			(+ (a n) 6)))
  (b 10))

In Clojure, you don’t need to use anything other than letfn, because it already supports the recursive nature that labels provides

(letfn [(a [n]
           (+ n 5))
        (b [n]
           (+ (a n) 6))]
  (b 10))

Finally, if you have local functions and other local bindings you need to establish, you can use a let, but no recursion is supported. This is sort of like CL’s flet but you can also use it for binding locals that are not functions

(let [a (fn [n]
          (+ n 5))
      b (fn [n]
          (+ (a n) 6))
      c 10]
  (b c))

I think the way Clojure uses square brackets in certain places that CL uses parentheses makes the code easier to read, overall.