Remove void "type"; convert statements to expressions

No se nombran archivos, pruebas ni puntos de entrada. Empieza revisando cómo el compilador gestiona los tipos de retorno void y las sentencias, y después define la semántica del lenguaje y las pruebas necesarias para la mutación fluida, los resultados booleanos de procedimientos y las definiciones cuyo valor es una expresión, antes de la implementación.

Perhaps we should remove the use of void as a return type. Objects that perform an operation that modify themselves and don't necessarily have something to return can return themselves (similar to the builder pattern).

Example:

aList = [].add(1).add(2).add(3)
print(aList) // [1, 2, 3]

Procedures that aren't related to a class can always return a boolean indicating success or failure. We can even make the lack of a return be considered a success.

Example:

procedure1() {
    a = 10
    b = 10
    c = a + b
    print(c)
}

procedure1() // prints: '20', returns true

procedure2() {
    number = readInt()
    
     // This if "statement" would also technically be an expression that returns a boolean in this proposal
    if (number % 2 != 0) {
        // This can also be an expression maybe?
        return false
    }
}

procedure2() // < 1 returns false; < 2 returns true

It might encourage chaining procedures together like so:

checkCondition1(args) && checkCondition2(args) && checkCondition3(args) && doGuardedProcedure(args)

This is possible in the current state of LJ, but it might be further encouraged if we default return types to boolean (is this something we want to encourage?).

Presumably function definitions and class definitions would now also be expressions if we were strict about "no statements" which would mean LJ has first-class functions (and potentially lambdas).

function1 = main() { println("Hello, world!" }

function1() // "Hello, world!"
main() // "Hello, world!

function2 = (arg1, arg2) { printf("%s, %s", arg1, arg2) }
function2("cats", "dogs") // "cats, dogs"

function3(otherFunction, arg1, arg2) {
    return otherFunction(arg1, arg2)
} 

function3(function2, "cats", "dogs") // "cats, dogs"

class1 = Dog { ...class stuff... }
dog1 = class1() // Returns a new Dog instance
dog2 = Dog() // Returns a new Dog instance

class2 = { ...class stuff... } // LJ uses '{ }' for sets, so there needs to be some disambiguation here
object1 = class2()

someFactoryFunction(someClass, args) {
    return someClass(args)
}

someFactoryFunction(class1, args) // Returns a new Dog instance

Treating everything like an expression feels more clear. It might seem odd to beginners that some funny words "return" stuff while other funny words "do" stuff. Some beginners have trouble figuring out when to use print() and when to use a return for example.