class BaseFunction {
static #allowInstantiation = false;
constructor(...args) {
if (!BaseFunction.#allowInstantiation) {
throw new Error(
"Why are you trying to use 'new'? Classes are so 2015! Use our fancy 'run' method instead!"
);
}
for (const [name, validator] of this.parameters()) {
this[name] = validator(args.shift());
}
}
parameters() {
return [];
}
body() {
return undefined;
}
static run(...args) {
BaseFunction.#allowInstantiation = true;
const instance = new this(...args);
BaseFunction.#allowInstantiation = false;
return instance.body();
}
}
class Add extends BaseFunction {
parameters() {
return [
["a", (x) => Number(x)],
["b", (x) => Number(x)],
];
}
body() {
return this.a + this.b;
}
}
console.log(Add.run(5, 3)); // 8
This should be programmer horror
validators
is a shitty name for something that actually does type conversion.JS disgusts me
A true FP programmer would make it
apply
instead ofrun
…Ahem, map…
And, of course, everything is a lazy list even if the functions can’t handle more than one element in each list.
Dont look at C++ with std:: function
I’m pretty sure this post is designed to kill the soul. I am made slightly worse for witnessing this abortion of an implementation and I will never be quite the same again.
OP, what’s your address? I have a “present” for you
I think that’s called a functor.
It’s just Java
Hence, Clojure. It’s not just functions that implement IFn… as the string of “cannot cast to clojure.lang.IFn” errors that I get because I couldn’t be bothered to validate my data’s shape is eager to inform me.
Yep, some code examples from the official documentation. This:
printPersons( roster, (Person p) -> p.getGender() == Person.Sex.MALE && p.getAge() >= 18 && p.getAge() <= 25 );
…is syntactic sugar for this:
interface CheckPerson { boolean test(Person p); } printPersons( roster, new CheckPerson() { public boolean test(Person p) { return p.getGender() == Person.Sex.MALE && p.getAge() >= 18 && p.getAge() <= 25; } } );
…which is syntactic sugar for this:
interface CheckPerson { boolean test(Person p); } class CheckPersonEligibleForSelectiveService implements CheckPerson { public boolean test(Person p) { return p.gender == Person.Sex.MALE && p.getAge() >= 18 && p.getAge() <= 25; } } printPersons(roster, new CheckPersonEligibleForSelectiveService());
The
printPersons
function looks like this:public static void printPersons(List<Person> roster, CheckPerson tester) { for (Person p : roster) { if (tester.test(p)) { p.printPerson(); } } }
Basically, if you accept a parameter that implements an interface with only one method (
CheckPerson
), then your caller can provide you an object like that by using the lambda syntax from the first example.They had to retrofit lambdas into the language, and they sure chose the one hammer that the language has.
Source: https://docs.oracle.com/javase/tutorial/java/javaOO/lambdaexpressions.html
That’s not quite right. In bytecode, lambdas are significantly more efficient than anonymous class instances. So while the lambda implementation is semantically equivalent, characterizing it like you have is reductive and a bit misleading.
Golang also does this, but it’s not classes.
How so?
Golang uses modules, not classes. Each of which may have its own main file.
Huh? Main file? Do you mean main package? A module can contain an arbitrary number of main packages but I don’t see how that has anything to do with this post. Also are you saying modules are equivalent to classes? That may be the strangest take I’ve ever heard about Go.
I meant main function. Oops
Amazing, lol