kotlisp

A home-made lisp project built on my parser combinator library, parsekt, and based loosely on the neat book at http://www.buildyourownlisp.com .

You must git clone parsekt and run mvn install in order to compile and run this project.

To run the REPL from the command line run

mvn exec:java -Dexec.mainClass="net.raboof.kotlisp.MainKt"`

Or run Main.kt in your IDE.

language

This language is a lisp-1 with lexical scoping. It follows some of the interesting quirks in the Build Your Own Lisp book including q-expressions. Instead of implementing quote and macros, there is a quoted lisp construct built into the language denoted by curly braces instead of parenthesis. Other than that, the language is more similar to scheme than common lisp.

Features

Evaluation Rules

As in many lisps, both functions and data are represented as lists surrounded by parenthesis.

When a list is evaluated, each item in the list is evaluated first. The first item in the list must evaluate to a valid function. The rest of the already-evaluated items in the list are passed to the function.

A function is created with the backslash () operator). Here is a function of two arguments:

(\ {arg1 arg2} {body})

Defining a function and giving it a name at the same time looks like:

{fun {name arg1 arg2} {body})

Notice how the arguments and body are not normal lists. They are surrounded by curly braces instead of parenthesis. This changes the evaluation rules. Instead of evaluating {arg1 arg2}, the evaluator leaves it alone; it evaluates to itself. This is called a quoted list or q-expression. Other lisps implement this feature by preceding a normal list with a single-quote character or by using macros.

q-expressions

q-expressions allow for some interesting features without implementing a quasi-quote or macro system. It isn't as powerful but it is conceptually simple and avoids the need to implement an expand phase in the evaluator.

Instead of writing out an argument list, we could instead assemble our argument list {arg1 arg2} with code.

(\ {arg1 arg2} {body})

; equivalent to the above:
(\ (map (\ {n} {symbol (concat "arg" n)}) {"1" "2"}) {body})

This concatenates "arg" with "1" and "2", and converts the strings to symbol names suitable for use in an argument list.

If {body} were replaced with {print arg2}, then this function would print its second argument. The symbol arg2 is defined inside the function despite it not appearing elsewhere in the code.

List manipulation

A list of some of the core list manipulation functions and their results:

(first {a b c}) => a

(head {a b c}) => {a}

(rest {a b c}) => {b c}

(cons a {b c}) => {a b c}

(list a b c) => {a b c}

(join {a} {b c}) => {a b c}

(last {a b c}) => c

(nth 1 {a b c}) => b

Convert a q-expression to an s-expression (i.e. unquote):

(eval {a b c}) => (a b c)

More:

(nil? {}) => #t

(list? {x}) => #t

Logic

As in scheme, true is #t and false is #f.

(if {cond} {q-expr} {q-expr}) - if condition with an optional else branch

(and expr*) - and condition which short-circuits on the first false expression

(or expr*) - or condition which short-circuits on the first true expression

(eq arg1 arg2) - check equality. Expressions are unfolded and equality is checked for each element

(boolean? expr*) - returns #t if all given expressions are booleans

Environments

There is a lexical environment and a dynamic environment.

When a function is invoked, its parameters are bound in a new environment and are only visible to code written in the function including any inner function definitions. This means that lexical environments have a hierarchy based on function definitions nested in the code. In addition to definining lexical symbols with function parameter names, the = function can be used to define a symbol in the current scope.

def puts a symbol at the top of the lexical environment. This is called the "global" environment. All of the built-in functions exist there.

I'm still figuring out how I want dynamic environments to work. For now, the dynamic environment can be accessed and modified using the dynamic and dynamic-def functions.

dynamic-fun function is also defined for convenience.

IO

read-line - read a line from standard input

print - print to standard output

error - print a formatted error to standard output

load - evaluate a file and return the result of the last expression

Calling into java

Java interoperability is achieved with a few functions and some clever marshalling.

• (ctor className params...): Constructs an object with any number of parameters
• (. object methodName params...): invokes a method on an object
• (field-get object propertyName): get an object field's value
• (field-set object propertyName value): set an object field's value
• (int number): coerce a long to a java int. This may go away with better marshalling.

The Lisp types are automatically translated into equivalent Java types. That means that, for example, you can call a java method on a string literal without having to construct a "java" string first:

(. "foo" "concat" "bar")
=> foobar

Constructing a java Socket is a more interesting example:

(fun {make-socket host port} {ctor "java.net.Socket" host (int port)})
(make-socket "example.net" 80)

Examples

Fibonacci:

(fun {fib n} {
  if {or (eq n 0) (eq n 1)}
    {1}
    {+ (fib (- n 1)) (fib (- n 2))}
})