i like object oriented programming. fight me
@monorail it's good
most object-oriented languages are disappointing (why do they all use classes rather than prototypes) but the concept be good
in your typical classical oo model, you have objects containing attributes, and those objects have a class which contains the methods you can perform on the object
it starts getting complicated when you think through more advanced features: do static attributes also live on the class? does that make it an object? where do static methods live that differentiates them from instance methods? what if you want to put custom methods on a particular object? where do traits fit into the hierarchy? etc.
ruby is classical and addresses those needs, but it does so by giving every object a hidden additional class called an "eigenclass". static and object-specific methods live in secret eigenclasses. it's weird
you can avoid that complexity by eliminating classes entirely. instead, you just have objects, and they have attributes. you don't need a separate structure to hold methods, because an attribute can trivially just contain a method. each object can be designated one or more prototypes, which simply means that if you can't find an attribute on an object, you check if the prototypes have it. that's it
base classes? prototypes. traits? also prototypes. statics? just live on a "class" object instead of an "instance" object. easy.
for instance: you're guaranteed to be able to pass constructors around as parameters, so you don't need to define factories that more-or-less exist to wrap up constructors in objects? some classical languages do treat classes as objects anyway and also achieve this benefit, but many don't
even without multiple inheritance (which is sadly something js lacks), traits still work out fine since you can treat them as a pile of object-specific methods. obviously it's nicer if you can just use the trait itself as an additional prototype ofc, but the flexibility of prototypal oo means it's still possible without that option
overall there's really no functional disadvantage to using a prototypal model, it's conceptually simpler, and the flexibility is often valuable, so
re: tangent and bleh bit negative re prototype discussion
a lambda is another term for an anonymous function, since the idea of anonymous functions originated in lambda calculus. many languages, python included, actually use the keyword
lambda to denote one
@monorail @riking @IceWolf @LottieVixen
i think part of the problem is that although nearly every language that supports it talks about closure as though it's some kind of in-language object (having a "closure" type, talking about how it supports "closures", etc.), it's really not one
closure is a property that functions have in most languages: when defined inside another function, they automatically "close over" the variables from that local scope and retain access to the scope even after the containing function has returned
the thing that's called "a closure" is usually an anonymous function. in the python example, it's a named function, because python's anonymous functions are deliberately very limited. either way, it's really a function
@IceWolf @monorail @00dani @LottieVixen Closures are anonymous[*] functions that capture their environment. But often you only care about the capturing, not the anonymity (in which case you can use a regular function in languages where regular functions also capture); or you only care about the anonymity, not the capturing.
@monorail @Felthry @IceWolf @LottieVixen
partial application is the ability to pass only some of the arguments to a function and get a new function that just needs the remaining arguments. currying is a particular technique for achieving partial application by having your function take just one argument at a time
you can also do partial application in other ways, depending on how arguments work in your language? for instance, if you have keyword arguments, you might do things a bit differently. i think python's got a
partial function you can call that allows you to partially apply both positional and keyword args
the cute thing holly was doing was indeed an example of (manual) currying though, yes
@00dani this is necro'ing an old thread but im just trying to understand this in ways that i get it.
is this kind of how lisp handles "generics" and "methods" via specializers on objects?
@monorail i view oop as orthogonal to imperative/functional since they can totally coexist
especially since an object is literally just a function with a dispatcher
fight fight mrawr >:3
i made a language which is both object-oriented and not object-oriented 8>
everything is like static functions (in Java/Python terminology), there is no a.foo() only foo(a) !
but some static functions try looking inside their first parameter if it's a "special hashmap" (an "Object") and if it has a the static function's full (long) name mapped to a value, it executes that value.
if it does something otherwise, that's the "default implementation"! :D
that way, there's no worry about whether a function like "crossProduct" on "vectors" should be implemented as a utility function later on or built into the vector class/interface! :D
you can add a utility function like say, "determinant" for "matrix" after the fact, and it will work for original MatrixImpl implementations like any static utility function, but if you make it do that thing I said, then if someone makes a NewMatrixImpl class, they can override your static function! :D
then later I learned this is probably the same as the Common Lisp Object System that's older than I am! XDD
but I'm still happy I figured it out :3
The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!