The Nineteenth Byte

Rydwolf Programs

19:00

ah

@Ginger The where module doesn't make it any cleaner

Ginger

this syntax doesn't feel right to me tho

~~import 2.2 of module~~

import module (^2.2.0)

that feels good

mousetail

@Ginger Metadata for sure

Ginger

ooh, you know what I just realised? it's possible, using this system, for the standard library to be rolling-release

mousetail

You want to specify the version on one place so it's easy to upgrade a dependency

Rydwolf Programs

Oh that's a good point. But maybe if you specify a version in one file, it'd be used as the default for the rest?

Ginger

19:05

no no yall are overthinking this

each file can specify the versions it wants and then the packager resolves which version(s) to use

Rydwolf Programs

So you could do import image_recognizer (^2.2.0) in main.rbt and then goat_tail_recognizer.rbt could just import image_recognizer

Ginger

if you're worried about redundant ones I'm thinking up a solution

Rydwolf Programs

@Ginger Yeah but that means if you want to change it for the whole project, you need to change it in every file

That's a bad thing

Ginger

I also realized: since modules can be upgraded to the latest available version from any available source, including potentially untrusted bundles, Rabbit will probably have to include a system for signing modules, but idk how that would work and not be a pain in the ass

hyper-neutrino

I don't think it's a good idea to have different files in the same project request different versions of a dependency

mousetail

19:06

^

Ginger

why?

Rydwolf Programs

If you tell it to use an untrusted bundle you're explicitly giving it permission to use an untrusted source

Signing doesn't seem necessary, and how would it get the public key anyway

Ginger

well yes but what's stopping me from hiding a h4x0red version of a standard library module in a bundle

probably good design for the packager

hyper-neutrino

a. you will get weird bugs trying to transfer data between parts of the project that might not match each other and b. if you refactor by moving any code to another file you might forget to check if dependencies align and get weird bugs from that

Rydwolf Programs

@Ginger Nothing, which is why you don't install your packages from random bundles

mousetail

19:08

Version constraints can get quite complex

Rydwolf Programs

A bundle's main module would be fine to install, assuming you trust the bundle file, but the dependencies within that bundle should all be untrusted

Unless you explicitly permit it to trust those

mousetail

For reference, a project I work on at my work has many in-house dependencies that need to be updated every day

Ginger

idea: Rabbit releases can have codenames; modules can specify what codename they want to use and that would be the default version for all stdlib modules so you don't have to specify what version of threading you want every time you import it

mousetail

If that was more chan changing a single file it would be incredibly tedious

Ginger

okay, metadata

actually I'd probably end up doing that anyway lol

Rydwolf Programs

19:11

IMO either should be allowed

Ginger

bleugh no

one or the other

and I like one

Rydwolf Programs

Simple single-file scripts shouldn't need a whole metadata file

Ginger

@RydwolfPrograms Rabbit is not a language that is made for single-file scripts

mousetail

Disagree with ^^

Ginger

actually, lemme just clarify this right now: Rabbit will not work with single scripts; if you want to make a Rabbit program it has to be a module

I'm sure some people won't like that, but that's how I'm doing it

if you want to make scripts use Python

single-file modules, however, are fine

Jacob

19:13

can you just run a module like a script?

Ginger

what do you mean?

Rydwolf Programs

@Ginger What is the difference between a module and a single file script?

Ginger

Modules have metadata files

Rydwolf Programs

Wouldn't a single file just be a single file module?

Jacob

like rabbit myprogram.rabbit which runs it, no?

Rydwolf Programs

19:14

Oh well if you already have a gross metadata file then put your dependencies in that

Problem solved

hyper-neutrino

will modules be installed per workspace like node or system-wide like python

Ginger

@Jacob nope, you'd have to do rabbit mymodule (assuming you're in the same directory as mymodule or it's in the global library)

@hyper-neutrino you can choose

Rydwolf Programs

@hyper-neutrino I think node does it user wide now

Ginger

there's a global module library, but you can tell Rabbit to also use another library somewhere else, like in a dev environment

Jacob

@Ginger okay, but you can run it, pretty much the same as a script just in a folder instead

hyper-neutrino

19:15

I know that npm i at least still does it in the deepest folder including a package.json though you can do -g to make it user-wide

Ginger

@Jacob yes, assuming you also have a valid metadata file in the module's directory

oh, and just to clarify: a "library" is a directory containing any number of modules and some metadata files

Jacob

could you tell me more about rabbit

also please put those little cute emojis in the readme 🐰

Ginger

so in a hypothetical future where Rabbit is commonly used, each user would probably have their own library, there would be a global library for all users (probably containing the stdlib), and the OS might have its own private library for internal stuff

@Jacob uh, most of the "planning" I've done is in the form of long chat monlogues

but if you have specific questions I can answer them

Jacob

is it fp?

Ginger

hmmmmmmmmmm

it's Pythonic

Jacob

19:19

ah i see

well i suppose that guarantees its mainstream success in the near future

Ginger

I came up with it to fix a bunch of issues I had with Python's packaging system (and which I have written several rants about)

Jacob

i thought you were just making a new package manager for python

pypackage right

Ginger

yeah, that failed because of how sucky Python packaging is

so I'm making my own language because why not I can do whatever I want

Jacob

darn i was rooting for u

Ginger

well now you can root for Rabbit!

Jacob

19:22

only if the log messages have emojis

if they don't i'm out

Rydwolf Programs

@Ginger I would...strongly hesitate to use that word for it :p

> Readability counts.

> Beautiful is better than ugly.

Ginger

It's Python-diagonal

Rydwolf Programs

OOP will never be pretty

Ginger

~~ok boomer~~

Rydwolf Programs

Especially not verbosity-oriented-programming of the sort Rabbit seems to be going with

Ginger

19:33

it's not object-oriented programming, it's physical-construction-philosopy-inspired software development

Rydwolf Programs

Anyone who voluntarily opts for OOP in 2023 without taking it in a fairly unique direction is making an objectively wrong choice

Ginger

rimshot.wav

Rydwolf Programs

OOP's just bad at achieving what it sets out to

Ginger

you're the only person I know who doesn't like it, and I personally have never had any problems with it

Rydwolf Programs

And the way that most langs do it, with all the verbosity and poor choices Java set a precent of, makes it even worse

hyper-neutrino

19:35

"OOP bad" isn't exactly an uncommon stance TBH

Rydwolf Programs

Yeah lol

Ginger

well good thing you can help me make fewer poor choices

Rydwolf Programs

It's kinda too late

Ginger

I like objects, I like inheritance

it all makes sense to my brain

I can understand it easily

Rydwolf Programs

You can have both of those things without going overboard

Ginger

19:36

hm

okay, how about this: what changes would you make to Rabbit's current syntax?

Rydwolf Programs

Thinking about things in terms of physical objects is nice...*sometimes*

Ginger

I'll be quiet until you say you're done

Rydwolf Programs

First I would remove classes and all related stuff

And go with structs that can have methods, preferably with impl

And traits, which do everything inheritance does but far better

I'd examine every aspect of the syntax with minimalism and visual cleanness in mind

It takes ten minutes to update your parser now and ten years to do it a decade from now

I'd make sure it's very easy to iterate on existing code, but also easy to plan things out from the start

Which is higher level than the syntax, but the syntax should reflect those values by being quick to visually parse or modify

And you get a couple weirdness points; you can afford to do a couple of nonstandard things with the syntax. Do more than two or three and you'll alienate people, but do nothing new and you're failing to take advantage of the opportunity

That's pretty much it

Ginger

alright

first: I don't like struct/impl

Rydwolf Programs

OOP with classes and stuff tends to be very big and very verbose for little good reason

@Ginger Syntax-wise? Or do you not like traits?

Ginger

19:42

Traits are cool, but I don't like the syntax

Rydwolf Programs

You kinda need impl for traits to work, since you need to be able to define methods on a class after it's been defined elsewhere

Ginger

it feels weird to me to have the structure and functionality of something be in two separate blocks

I like having them in one place, where they feel like a cohesive whole

Seggan

@RydwolfPrograms its not objectively wrong

Rydwolf Programs

Separating the data fields from the functions does have some advantages tho, aside from being able to add more later. It makes it easier to remember that the data is really all that's being stored (assuming you're not doing it JS-style where methods are lambdas in properties).

Ginger

were there a way to do traits but have structs and impls be one thing I'd love to do that

@RydwolfPrograms wdym "all that's being stored"?

Rydwolf Programs

19:45

Like with Rust Iterators, I keep catching myself thinking of the iterator as some sort of special thing

When it's really just a couple data fields, and some functions which I can apply to those data fields

Ginger

I like thinking of them as a special thing

why? I don't know

Rydwolf Programs

Same here

The separation of data and functions helps counter that

Ginger

but having a "thing", a solid object that I can interact with, makes more sense to my brain then a weird lump of data and functions that I can apply to it

it feels more cohesive

Rydwolf Programs

Because sometimes that's useful, but sometimes looking at it as a weird lump is more useful

But also...the function isn't stored in the object, it's something you do to the object. They're not as much a single object as we train ourselves to think

And while you can think of it as "Dog is barking", 90% of the time in actual code it makes more sense to think of it as "I am barking Dog"

Ginger

maybe it's that having two "things" which only work with each other and are specifically made for each other feels weird to me because I'm like "why have them be separate? they're made for each other so they should be one thing"

Rydwolf Programs

19:50

Well, the impl is made for the struct, the reverse may not be true

You can add traits to structs you didn't define

Which is part of what makes them so cool

Ginger

yeah, I don't like that

Rydwolf Programs

Why?

It allows for all sorts of incredibly useful things

Ginger

why would I be modifying someone else's code? it's not my code

it feels weird to me

Rydwolf Programs

You're not modifying it, you're making your code apply to it

If you make a trait for things that bark, of course it should apply to std::Dog

Ginger

yeah I guess, but it feels wrong to me

Rydwolf Programs

19:52

(worth noting you can't impl a trait you didn't define onto a struct you also didn't define; you have to define one of those two)

Ginger

I'll do it if it's the best thing to do, but I just don't like it

Rydwolf Programs

Traits aren't just for adding functionality, they're also for classification

There's traits in Rust's stdlib with no methods

That's part of why it makes sense to allow implementing them on outside things

Ginger

hmmmmmmmmmmmmmmmmmmm

how would this system work with something like GObject, which is very OOP-based?

Rydwolf Programs

I can't actually think of an instance of a library in Rust that adds methods to types from another library

hyper-neutrino

do you plan to have operator overloading

Ginger

19:54

because I am not doing anything that I can't make work with GObject

@hyper-neutrino yes

Rydwolf Programs

@Ginger Well let's say there's a Button class you'd normally inherit from to make a custom button type

hyper-neutrino

would this (adding traits to others' structs / adding/changing methods) be able to be used to add op overloading to native types? like for example making list addition vectorize instead of concatenate

Ginger

@hyper-neutrino AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

NOOOOOOOOO WHY WOULD YOU DO THAT

you can't just change how a critical part of the language works for no reason and then expect all the other code to be fine with it!

Seggan

^

hyper-neutrino

oh I don't expect other code to be fine with it

Jacob

19:56

me omw to change how a critical part of the language works for no reason and then expect all the other code to be fine with it!

hyper-neutrino

I did this for a library file for python for AoC cuz I wanted to be able to add stuff like using << to rotate arrays

user

@hyper-neutrino Wouldn't it be better to have separate operators for that? ++ for concat and + for add?

Traits can't be used to do that, ignore hyper

Rydwolf Programs

@hyper-neutrino No

hyper-neutrino

you shouldn't leave out good features just because people can abuse them in bad ways

you would not take a hammer out of a toolbox just because it can be used as a blunt weapon

Rydwolf Programs

Usually you can only add a trait to an outside struct if you defined the trait

user

19:57

People can't abuse them anyway

^^

hyper-neutrino

hmm okay

Rydwolf Programs

Like in Rust, you can't add the Sub trait to String

Since you didn't define either of them. Same here I'd assume

Maybe if you defined a new operator, you could add its trait to an existing type tho

user

I guess you could make a wrapper around vectorizable lists or something

Rydwolf Programs

@RydwolfPrograms Which would make sense

user

Rust doesn't let you define new operators, but it'd be cool if Rabbit did

Ginger

19:59

okay, I've come up with some syntax that I like for structs and impls:

struct Something:
	int beans
	Display implementation:
		print(this.beans)

(this is a WIP btw)

Rydwolf Programs

@RydwolfPrograms Well, in a trait-based language, you'd make Button a trait instead

Done, simple

Ginger

@RydwolfPrograms ...what

Rydwolf Programs

@Ginger How do I add a trait to an existing struct?

Transcript for