(learn '(LISP))

(learn (LISP))

(why? (LISP))

LISP is the language of the gods. It is the ultimate hacker's language. That comes from its tinkering history (how many LISP dialects are really out there?), its elegant syntax and design, and finally, its open and flexible macro system.

Historically, LISP was the second (after Fortran) earliest high-level programming language. It was a pioneer in design, thought, and implementation of high-level programming languages, years before tools and technologies like C, bash, Unix, and networks like ARPANET were 10+ years out from being developed.

On syntax and design, LISP's native data structure is a list, and LISP source code is written as a list. This concept is called homoiconicity, which means that the program structure is similar to its syntax.

If a language is homoiconic, the source code has the same structure as its abstract syntax tree, which allows the code in the language to be accessed and transformed as data. LISP is expressive, has symmetry to it, and is beautiful in many ways.

On macros, LISP is wide open and flexible, because its macros are pre-processed
and returned as forms to the compiler, they are not values to be evaluated by
the compiler.

With this, you can write code that will write your own code. You can define your own syntax, write your own domain specific language, or implement another language using LISP (for example, a Python interpreter implemented as macros in LISP).

You can create any sort of programming paradigm you like and include it in your programs. If you can imagine it, you can do it. It enforces no structure (except (make (sure (to (balance (your parens!)))))).

(history (and background))

LISP is one of the oldest high-level programming languages in use today. Described by neckbeards as having mystical origins, it was created by John McCarty in the mid 1950s while at MIT. LISP's development was influenced by Alonzo Church's lambda calculus, developed in the 1930s, which is a formal mathematical logic using function abstraction and application using binding and substitution.

It was first implemented on an IBM 704. It soon made its way to a PDP-1 (note, Unix was first run on a PDP-7 by Ken Thompson and Dennis Ritchie about 10 years later) soon after by Steve Russell, who read McCarty's paper. McCarty was suprised by Russell's work and didn't realize eval could be implemented in machine code.

LISP's names comes from LISt Processor. In LISP, code is data and data is code. Unlike other languages, source code is either data or code.

A current popular LISP used today is called Clojure, which is a functional and dynamically typed language that compiles to Java byte code and thus can be turned into a .jar file and run everywhere Java can.

Compiling source code into an executable jar is as easy as: lein uberjar and then running the jar in the Java runtime with: java -jar my_jar.jar.

Note: Clojure is different from another LISP dialect called Clozure CL, a common LISP implementation. Additionally, note that Clojure is also different from the programming concept known as a closure, which is a technique in programming to bind variables for use in higher-order functions. Note that also you can implement many closures in Clojure or Clozure CL.

(basic (concepts))

In LISP, all source code is represented by symbolic expressors, S-expressions, or nested tree-structured lists. A list looks like this: (). A list takes two forms: a form to be evaluated and a data form.

The data form is with a quote prefixing the list, '(), or '(2 4 6). If the list is to be evaluated, the list starts with a functional call, and then with arguments to the function. A function f that takes two arguments would be called like so: (f arg1 arg2).

For example, to cast and concatonate the integer 2600 with the string " is the ultimate hacker magazine", you'd call the str function on those two datatypes. Do the following in a LISP REPL or LISP source code:

=> (str 2600 " is the ultimate hacker magazine")
"2600 is the ultimate hacker magazine"

To add 2 to 4, call the + function on those two datatypes:

=> (+ 2 4)
6

To print something to standard out:

=> (println "Clojure is a mystical language")
"Clojure is a mystical language"

To create a variable, you use the reserved word (which is actually a macro behind the scense) def along with the name of the variable and then the variable value. For example, (def a-lisp-dialect "Clojure"). You can then use it like so: (println a-lisp-dialect).

To create a function, use the reserved word (also a macro) defn along with arguments given in a vector, and then the function definition. Clojure, as a functional language, omits the return keyword and every function definition uses an implicit return. For example,

(defn my-squarer [num]
  (* num 2))

(cons (car (cdr (and recursion))))

In the early days, two assembly language macros for the IBM 704 became well-known functions for operating on lists: car (Contents of the Address part of the Register number) and cdr (Contents of the Decrement part of the Register number).

Recall that the basic datastructure in LISP is a list, (). This is also referred to as a cons cell, made of two items: a value and a reference pointing to another cons cell.

Take the list '(2 4 8). This can alternatively be written as '(2 (4 (8 nil))) using the cons cell paradigm.

The car is the first element of a list. For '(2 4 8), 2 is the car. The cdr is the rest of the list, except for the first element. If the rest of the list is empty, the cdr is nil. In the above list, the cdr is '(4 8)

Let's expand on these to build a recursive function. Since LISP has a functional programming mystique (more on that soon), if we want to operate on this collection, we would want to use recursion. We also want to use functions to help us as we recurse.

Suppose we are given a list, '(2 4 8), and a task, to write a recursive function that will sum the items in the collection.

A way to write a recursive function on this list would be to check if the list is empty, and if not, pop the first element off, add that element to our accumulator, and then recur.

(defn recursive-sum
  [list acc]
  (if (empty? list)
    acc
    (recur (recursive-sum (cdr list) (+ acc (car list))))))

(functional (programming))

Functional programming and LISP often go together when one sings the praises of one of those concepts. Functional programming is the concept that the most of your program can be represented in functions, and that you can trust those functions to perform the action you expect.

This is in contrast to imperative programming, where you tell the computer what to do and the program changes state. Functional programmers dislike mutable (changing) state, and value "pure" functions, or functions that, given the same arguments, always have the same return value. An example of these might be the following:

(imperative (form (in javascript)))

var sum_of_array_items = function(arr){
  var sum = 0;
  for (var i=0; i<= arr.length - 1; i++){
    sum+=arr[i];
  }
  return sum;
}

Program state is represented by position inside the for loop, as well as the temporary variables i and sum.

(functional (form (in clojure)))

(defn sum-of-list-items [my-list]
  (reduce + my-list))

Similar to the above, this function takes a collection (can be a () or [] datatype, similar to an array in JavaScript or list in Python), and calls reduce with the + function on every element of the collection. Since reduce takes a function (+), a collection, and an optional accumulator, and this returns the sum of the list above.

Another way to think about calling reduce on a collection with a + operator would look like: take a list: '(2 4 8) but place a + between every element in the collection, so: (2 + 4 + 8) (or in LISP, (+ 2 (+ 4 (+ 8 nil))).

(lisp-1 (vs lisp-2))

Diehards in the LISP community may debate LISP-1 vs LISP-2. LISP-2 treats functions as values, so in order to make a function call, you must prefix the function call with a special funcall operate, or else the function is treated as data. LISP-1 is a bit more conductive to functional programming by assuming that unless quoted, the list is a list to be evaluated, and the function exists in the first position of the list.

An example of a LISP-1:

(sort > '(5 2 6 3 1 4))

And the same example in LISP-2 form:

(sort #’> '(5 2 6 3 1 4))

Note the difference between the #' prefix for the > function call.

(scheme (vs common-lisp))

One of the side-effects of a language of a language as old as LISP is there are many, many different implementations. In the 1980s, an effort was made to standardize, with a specification called Common LISP. Common LISP (CL) focuses on practicality, so it is easier to get projects started and write less code in CL. However, purists disagree with the tradeoff of practicality over form.

An example of Common LISP for computing a factorial (a factorial is the product of all the integers below it, ie factorial of 5 is 5 * 4 * 3 * 2 * 1):

(defun factorial (n)
  (if (= n 0)
      1
      (* n (factorial (- n 1)))))

Scheme, in contrast, is the most beautiful representation of LISP. If programming were art, it would be represented as Scheme LISP. It is one of the few languages that support tail-call optimization (write recursive functions, which usually has poor space and time complexity in Big O notation, for iterative space and time complexity — the best of both worlds — elegant source code with fast performance and small footprint on the stack and heap).

An example in Scheme of a recursive and tail call optimized function for computing a factorial:

(define (factorial n)
  (fact-iter 1 n))
(define (fact-iter product n)
  (if (< n 2)
      product
      (fact-iter (* product n)
                 (- n 1))))

(macros)

Macros are one of the more interesting features of LISP, which allow you to transform LISP code. With it, you can change the language, implement your own features, or even write a new programming language entirely.

During the macro-expansion phase, the LISP expression will be passed into the macro function, which can do arbitrary computation at macro-expansion time, the
result of which will be LISP code. The LISP code is then passed to the interpreter or compiler, which is then executed at run time.

LISP macros result in unrestricted string rewriting, which is Turing Complete. LISP is also Turning Complete. Therefore, with macros, you can write code that will write your code for you.

Let's implement.

Clojure doesn't have a for loop like many programming languages do. Clojure is too functional for that, and would prefer, for example, you apply a function to the elements of the collection instead, i.e. (map #(* % %) '(2 4 8)) to multiply a number by itself (note #() is itself a macro for %the Clojure lambda function which looks like: (fn [args])).

However, we can %write a macro for-loop that will pre-process all of our calls to for-loop %and turn them into regular Clojure code.

(defmacro for-loop [[sym init check change :as params] & steps]
 `(loop [~sym ~init value# nil]
    (if ~check
      (let [new-value# (do ~@steps)]
        (recur ~change new-value#))
      value#)))

Use like so:

    (for-loop [i 0 (< i 10) (inc i)]
      (println i))

(fun (facts))

Earmuffs: In Clojure, variables are declared with a def statement. If the variable is intended to be used globally, it gets earmuffs, or @ surrounding it. So a local variable: (def my-favorite-language "Clojure") vs. a global var: (def @my-favorite-language@ "Clojure").

LISP Machine: The 1980s had a burst of activity for LISP, as it was the favored artificial intelligence language. Most computers (still, to this day) use a von Nuemann architecture of a central processing unit (CPU) that fetches data from memory via a bus to a memory register in the CPU, executes an instruction, and ultimately writes data back to memory via the bus.

In this architecture, the bottleneck is the bus since the CPU must "waste" clock cycles fetching and retrieving data from the bus. AI programs in the 60s and 70s required a considerable amount of processor time and memory space.

As the integrated circuit technology shrank the size and cost of computers, and the memory needs of AI programs exceeded current computers, researchers tried a new approach: a computer specifically designed to run large AI programs and tailored to the semantics of the LISP language.

(lisp (implementations))

• Common Lisp
• Scheme
• Armed Bear
• Clozure
• Steel Bank CL
• Emacs LISP
• Racket
• Hy
• Clojure
• LFE

(if (conclusion?) (learn LISP!))

LISP is indeed the language of the gods! It is a language that is written well. Learning LISP will change the way you think about programming. Now, go learn you a LISP for the greater good!

(References)

• http://www-formal.stanford.edu/jmc/history/lisp/lisp.html
• http://www.paulgraham.com/lisp.html
• https://en.wikipedia.org/wiki/Lisp_(programming_language)
• https://en.wikipedia.org/wiki/Scheme_(programming_language)
• https://en.wikipedia.org/wiki/Tail_call
• https://en.wikipedia.org/wiki/Common_Lisp
• https://en.wikipedia.org/wiki/Lisp_machine
• http://stackoverflow.com/questions/4578574/what-is-the-difference-between-lisp-1-and-lisp-2
• http://stackoverflow.com/questions/9981943/how-to-implement-a-for-loop-in-clojure
• http://wiki.c2.com/?LispMacro
• http://stackoverflow.com/questions/1986961/how-is-the-var-name-naming-convention-used-in-clojure
• https://en.wikipedia.org/wiki/Homoiconicity
• https://www.bloomberg.com/graphics/2015-paul-ford-what-is-code/