@Seggan does this make any sense to you

sorta

if I'm doing something wrong I'd like to know how to fix it now before I cause any more problems for myself :p

Six-byte trilangle program to print the number of days since last Thursday: D'%7!@

ooh

@Ginger End users wouldn't have carrot right?

depends

if you want to do a carrotless install you have to use an application bundle, which is a whole other thing

I thought the whole point of bundling was so that people could install Rabbit programs without needing the development package manager

normal bundles are... kinda useless; application bundles are what make ^ possible

I think you're overcomplicating things. I write code. That code, across a few files, is a module, and that module has a config file, which specifies certain metadata, and optionally some dependencies. I get those dependencies as other modules, which are in some sort of tree, or maybe in some user-wide package store. I can bundle this together into bundles, which are how the code gets distributed to end users, who only need to have Rabbit installed, which then runs the bundle.

If I'm a developer, I have carrot, which is how I actually install those dependencies to put in the bundle

It gets those from various repository types, including bundles, which are just big directories of unorganized modules

It defaults to a central repository you, Ginger, run, but you can also install from other sources

what you've described is almost correct... except for one thing

the Rabbit interpreter cannot exist independently from carrot, except for within application bundles (okay technically it can but it's going to be a real pain in the ass to use)

...why

That'd be like if NPM was part of Node, or cargo was part of rustc

The language should not depend on the package manager, and you should need to waste disk space and stuff on a package manager you'll never use

because of a detail of carrot I may have not elaborated on until now: carrot manages, in addition to modules, installing copies of the Rabbit interpreter and launching Rabbit modules

Ew

That should definitely be a separate program

well, it's not

And you should be able to install a version of Rabbit without that program

E.g., I can still install a specific Node or rustc version without nvm or rustup

Those are tools for developers, not end users

you can, but it'll be a real b*tch to use, because one of the things carrot does is pass arguments like library paths and such to the interpreter

the interpreter does not know where to find libraries without being explicitly told

If I install a bundle file from apt or dnf, it should be able to list Rabbit as a dependency, instead of carrot, which then is maintaining Rabbit in parallel to the system package manager

@Ginger ...isn't that how interpreters work?

wrong: carrot will be designed so that it can be invoked by the system package manager to do the actual install

Like I'd assume you give the interpreter a path to either an individual script, or to a module config which would specify what to run and what dependencies it needs

@RydwolfPrograms you tell the interpreter two things: the module name and version to run and paths to the libraries to look for it in

Wait so it has to traverse through directories and stuff just to find the file to run?

@RydwolfPrograms yes

Nasty

too bad

It should follow a predictable pattern

You're making something worse than Python

what the frick are you talking about

it is predictable

for any combination of arguments the same module will be run every single time

Then why does it have to search through directories to find the right file to run?

It should always be a specific file in a specific place, not "some file in one of these folders, idk, figure it out"

what do you mean "search"? the first library containing the correct module is the one that will be used

What is the point of doing it that way

and if a library doesn't have it, on to the next, and so on until all are exhausted

Why

it'd probably make more sense for carrot to do that, but it'd still be done

That's way unnecessary abstraction

this is the same thing Python does

what do you think sys.path is for?

I thought your goal was to do better than Python

sys.path is a fine way of doing it; Python's issue was that you could add whatever the hell you wanted to it

Oh god

So with Rabbit you can't

Why are you even doing it that way then

Wait, is it trying to find these files at runtime?

Yes

Or when you install the modules?

Rabbit solves that by making sure that only valid and known libraries are being used; in fact, generally only two libraries are searched: the user one and the global one

@RydwolfPrograms :|

this is why you should use carrot to run Rabbit code: it ensures that all the libraries are valid, among many other checks

Ginger, this is a solved problem. There are two valid approaches. Either all the dependencies get stored in one central user-wide store (just a directory full of modules), or you use a tree structure and store all of the dependencies of a module alongside it

#1 is what I'm doing

except it also checks a system-wide store

lemme clarify: there should only ever be two libraries used by the interpreter and they are the global one and the user one

Calling those "libraries" is really confusing

^

if you're adding extras you are doing it wrong (unless it's for developing code, in which case it's fine to have your workspace as an additional library)

okay sure fine whatever

Store is a pretty common term. Or simply "folder"

sure, stores

Why are you giving people the option to do it wrong

because setting the paths in stone only causes problems for people with an actual need to modify them

Fair enoiugh I guess

I think calling them libraries is what confused me

okay cool

You earn one "not as cursed as it could be" point

I'll call them stores from now on

@RydwolfPrograms :D

I'm the edge case person; I hate it when people make things like module paths set in stone because it means I often can't use their tools in the conditions I need to use them in

Example: Poetry doesn't handle extra Python module repos well, which is annoying because it means I can't use Piwheels with it

also, side note re. OS package managers: carrot does support installing modules via an OS package manager

also also: because carrot is itself a Rabbit module, it can also be installed through an OS package manager

Woah woah woah...carrot is run by Rabbit, which is installed and run by carrot?

yes, I'm typing an explanation right now :p

to fix the issue of carrot needing a Rabbit interpreter to run itself, its OS package would probably include a bootstrap script that installs a bundled interpreter for carrot to run on; after that point carrot can install a newer version as needed

the same thing would be done in the "official" install shell script, just with a dash more interactivity

(and I say "official" here because the OS packages will be as official as any other; the script's just in case you don't have a package manager to use)

There's a reason nvm is written in shell and not JS

@RydwolfPrograms Isn't Pip kinda done like that?

But yeah, slightly cursed

Python is cursed according to Ginger

And if Ginger calls something cursed

@user it is (assuming you mean the Python package manager)

I believe it

@RydwolfPrograms glad to know you put so much faith in me :b

I think it's more that something would have to be really cursed for even you to call it cursed

@Bbrk24 I don't take advice from a company whose name could stand for Nude Pasta Manufacturers q:

@user I think that's why he trusts me

Do you put sweaters and shoes on your pasta?

yeah, who doesn't?

All pasta is nude

@user Yes, but not pants

fun fact: Rabbit is implemented in Kotlin, so it can hypothetically run on your Blu-Ray player

I forgot, are you transpiling Rabbit to something or interpreting it?

interpreting; that's why the source is so cursed

also, what exactly did you mean by chat.stackexchange.com/transcript/message/63220557#63220557?

Oh, I thought it was compiled

I was going to say that the type objects that your compiler uses don't need to be at all connected to the type objects that you use for runtime reflection

But if you're interpreting, then ignore all that

@user perhaps you're possibly a prudish pasta producer panicked by pants-less Pennine?

@user cool!

I still think you're doing it wrong, but I know even less about interpreters than I do about compilers so I can't really give you unsolicited advice there

@user I'm almost certainly doing it wrong, don't worry d:

I technically have more experience writing interpreters than compilers but Trilangle isn't really comparable to Rabbit

Bah, I missed another language design discussion! D:

s/discussion/argument/ :p

Arguments are just discussions with sides.

@Ginger theres a reason nvm and npm are separate

@DLosc madlad actually did it

It's beautiful

Fun compiler idea: when compiling a program, it tells you if any of the lines are brilliant moves (!!) best moves (!) mistakes (?) and blunders (?!)

int main() {
    println("Hello, World") //!!
    int* p //?
    del p //?!
}

delete (new LinkedList()); // ?

And if you somehow checkmate the compiler, it halts compiling and just prints a # after the line that checkmated it

You say that but I have caused the Swift compiler to segfault before

But is that checkmate or was it just stalemate

Like checkmate is when the compiler just resigns because it knows you're going to do something irrevocably stupid

Like it tries to get out of it but it can't

It finishes without throwing an error

Doesn't do any clean up because you've won but at what cost

I got my nullability qualifiers wrong and it broke. idk

@lyxal We've got ? and ?! already, they're called warnings and errors

@ATaco Discussions can have sides too. People can push a particular solution to a problem they're discussing. Arguments are just less open, more heated

Arguments are often more heated, but not strictly.

@user I'm talking about things that aren't compiler errors

Anything that isn't an error or warning is a best or brilliant move :P

Brilliant move if all tests pass

Memory leaks are Blunders

PHP is a Blunder

...what if there was an esolang where the program is a series of legal chess moves but the execution only depends on an engine's evaluation of them

I like that idea. The specification could just have one sides moves be "undefined behavior".