The Nineteenth Byte

Ginger

15:00

should main be implicitly public?

Rydwolf Programs

Public to who

mousetail

No, it shouldn't normally be public

Ginger

who, exactly, will be accessing main other than the interpreter?

nobody, that's who

Rydwolf Programs

Would libraries ever need mains?

mousetail

Only reason to have it public is if it calls itself recursively via another function in a different namespace. This should be very rare

Ginger

15:01

@RydwolfPrograms nope, why would they?

so how about main gets to be (according to the parser) an access modifier, like public or private

Rydwolf Programs

I guess if it was a library that used some sorta thing in the background, like startin g a daemon or something, but that should definitely be done by calling a library.init() sorta function

Parcly Taxel

Back

Ginger

@RydwolfPrograms hmmmmm

Rydwolf Programs

@Ginger ...which is automatically applied to a function named main, right?

Why would the syntax look like main function name()

That's just boilerplate, main is the name

Ginger

it's 5 characters of boilerplate, only a Gentoo user would complain about that

mousetail

15:03

@Ginger In rare cases you should be able to make it public

Ginger

@mousetail huh, really?

Rydwolf Programs

The advantage of main as the name instead of a keyword is that libraries' main functions aren't special

What happens if I import something with a main function?

Can I call that function?

Does it get automatically called?

Ginger

a function marked main should be called automatically upon import, so running a module just involves importing it with no weird if __name__ == "__main__" stuff

Rydwolf Programs

No

That's gross

Ginger

if you don't want that don't mark it main

Rydwolf Programs

15:04

3 mins ago, by Rydwolf Programs

Would libraries ever need mains?

Ginger

if they have initialization to do, sure

Rydwolf Programs

No initialization should exist for the library as a whole

Global state is gross

Require the importer to do lib.init() to initiate an instance of that

Ginger

@RydwolfPrograms I agree wholeheartedly with that

Rydwolf Programs

And initialization of a whole library is global state

Ginger

I think it should be up to the developer

New Posts

15:06

0

Q: Simultaneously solve N-queens and no-3-in-line

An intriguing MathsSE question asked if there were large N-queens solutions where no three queens lie on a line. That question's body included the unique 4×4 solution up to symmetries . Q . . . . . Q Q . . . . . Q . and noted that there are no solutions for 5×5 to 7×7 because of knight lines. Ho...

Ginger

but I'm doing it this way and that's static final d:

Rydwolf Programs

And let's say two things I depend on import the same module

Ginger

it only gets called once, because of caching

Rydwolf Programs

Do they each run their mains separately?

If they don't that means a second library being run alongside mine can change the behavior of one of my dependencies

Parcly Taxel

Did we have a question on OEIS A326077?

Of course not. But would it make for a good challenge?

Ginger

15:08

@RydwolfPrograms then just... don't add a main function to a library! do init() the way you suggested!

Rydwolf Programs

Libraries should be "immutable". When I import it it should be the same every time. Anything that can change needs to be an instance of something in that library.

@Ginger But if you give people the option, they will take it

Alternatively...I guess you could just refuse to allow importing things with mains

Ginger

okay, how about this: main is only executed in the module being run

Rydwolf Programs

Which may actually be a better option than having main as a private function

Ginger

all other mains are ignored

actually, I like this idea

Rydwolf Programs

Yeah, I guess that'd be fine

Ginger

15:09

it means libraries can put little demo scripts in their mains that only run when you execute the library directly, like Python's __main__.py

Rydwolf Programs

OOOH trueee

Like python without an argument being a REPL

Ginger

exactly!

and if for some ungodly reason an importer wants to call the demo function directly they can still do that

ok cool

Rydwolf Programs

Would main be accessible from the outside? That seems unnecessary

They could always create a public wrapper around it

Ginger

yes, I can't think of any reason for it not to be

it hurts nothing (probably)

Rydwolf Programs

How about the library author not intending the main function to be run alongside other things

Ginger

15:11

hm, fair enough

Rydwolf Programs

Why are any functions private if they can just say "don't call this"

Ginger

next: should static be the keyword used for class methods? I don't really like it, but Python's @classmethod thing is too verbose (and janky anyway)

also, what should constructor syntax look like?

I like Vala's multi-constructor idea, but I feel like the syntax could be improved

first idea:

public class Main():
	constructor function():
		pass
	constructor function withFoobar():
		pass

doesn't feel right

public class Main():
	construct():
		pass
	construct withFoobar():
		pass

that feels a bit better

in Vala, extra constructors are called like class methods (Main.withFoobar()); I don't like this, so I think I'm going to do something like Main:withFoobar() for constructors and Main.classmethod() for class methods

although actually, do they have to be distinct?

it's not hard to make a class method that returns a new instance of the class

do I use new? because if I do I can re-use it for constructors

Seggan

@Ginger koltin's

Ginger

example?

Seggan

it has a primary constructor, and secondary constructors that call the primary

lemme make a quick example

Ginger

15:20

aight

Seggan

class Complex/* primary constructor */(val re: Double, val im: Double) {
	/* secondary constructor */ constructor(re: Double) : this(re, 0) {}
}

Rydwolf Programs

@Ginger No constructors

Seggan

the advantage is you see that val in the constructor? its boilerplate for creating a property and a parameter to set that property

Ginger

@RydwolfPrograms huh?

Rydwolf Programs

Just have static functions that return an instance

Like Rust does

Ginger

15:22

like Main.new()?

Rydwolf Programs

Yeah

Ginger

No

Seggan

@RydwolfPrograms doesnt really work well in a traditional oop lang

Ginger

^

Rydwolf Programs

Hmph, fine

As long as you don't have a new keyword

Anything but a new keyword

Seggan

15:23

how are you supposed to construct it in the static method?

Rydwolf Programs

Oh true

Ginger

@Seggan hm, what if I want to call a function in the primary constructor? where does that go?

Seggan

in a separate init block

Ginger

hmmmmmmm

Seggan

class Complex/* primary constructor */(val re: Double, val im: Double) {
	/* secondary constructor */ constructor(re: Double) : this(re, 0) {}

	// called by primary constructor
	init {
		whatever()
	}
}

im using this system for rol

Ginger

15:25

I think I'm going to do this:

public class Main():
	construct():
		pass
	construct withFoobar():
		pass
Main()
Main:withFoobar()

Rydwolf Programs

Single colon?

Why not .

Or ::

Ginger

:: is too chunky, . is for calling methods

Rydwolf Programs

Having a weird third property accessor for one specific case is just janky

@Ginger And...you're calling a method

A constructor is a method

Ginger

I don't like how Vala doesn't differentiate between calling a constructor and calling a static method

I could just have static methods that work like constructors tho

Seggan

@RydwolfPrograms lua go brrr

Ginger

15:27

public class Main():
	construct():
		pass
	<static keyword> withFoobar() -> self:
		pass
Main()
Main.withFoobar()

Rydwolf Programs

No, still have the constructor keyword and write them like constructors, but use .

Ginger

blugh no

Rydwolf Programs

'cause now your constructors are indistringuishable from methods syntactically, which is exactly what you didn't want

And you need some way to construct a self within your constructor functions

Ginger

either no extra constructors or extra constructors that use :

and I like option 2

Rydwolf Programs

Oh wait so with the static constructors option there'd still be a single primary one?

Seggan

15:29

@Ginger whats the parens after Main

Ginger

@RydwolfPrograms yes

@Seggan they aren't required, but they work the same way as they do in Python

Rydwolf Programs

oh frick I missed an NHS meeting

Ginger

F

Seggan

@Ginger ugh

Rydwolf Programs

Got buried in my emails lol

I'm prolly getting kicked out anyway so I might as well just not go I guess

I didn't do any volunteer hours or service projects 'cause I thought they were the same thing and there weren't enough service project opportunities for me to do one

Ginger

15:30

actually, should I do public class Foo(Bar) or public class Foo extends Bar?

Seggan

class Foo : Bar

kotlin again

and please make public the default

what access is without a modifier?

Ginger

private

Seggan

@Ginger use kotlin-style nullable typing

@Ginger aaaaaaaaaa

please make public default

Ginger

no

Seggan

i have nightmares of writing public everywhere in java

Ginger

15:35

@RydwolfPrograms I know you hate dunder methods, which is fair, so I'm going to avoid using them wherever possible

why do __init_subclass__() when you can do extends constructor()?

Seggan

@Ginger howd you do that

Ginger

@Seggan actually, I just realized that maybe making public the default is better, because private classes can only have private members

@Seggan magick™

Seggan

@Ginger YES please

Ginger

Rabbit will have infix methods, using the operator keyword, which will also be used for overriding mathematical operators

for example: override operator +()

the override keyword is required, to make sure you're actually doing what you intend

Rydwolf Programs

?

Oh wait nvm

But still override is unnecessary

It just wastes space and typing

You're making this so unnecessarily verbose

OOP is poison

It infects your brain

Ginger

15:45

yeah, but that's just, like, your opinion, maaaaaaaan

mousetail

@Ginger This would make sense if you need to inherit from some class to be able to do it

Ginger

@mousetail oh, yeah

if you're not inheriting it's just operator

mousetail

Then I'd say requireing override makes perfect sense

Ginger

ok, here's Python dunder methods in Rabbit syntax:

__new__ -> class construct()
__init__ -> construct()
__del__ -> delete function <name>()
comparison and math -> operator <name>()
pretty much everything else -> interfaces

that's right, we're doing interfaces

Seggan

@Ginger it will be the easy part. trust me, the hardest part will be the type checker (i speak from experience)

also dependency resolution will be very hard

mousetail

15:58

delete seems odd. Why do you need function there?

Seggan

opinion: no delete

mousetail

I like delete, it eliminates the need for with statements if it's sufficiently reliable

Seggan

@Ginger elegant, but i prefer rusts or javas way of doing it: implementing interfaces

@mousetail huh

how do delete and with go together

argh whats with my typing today

mousetail

For example, closing a file when it's deleted

So you don't need to do with open()

New Posts

0

Q: Shortest way to convert the result of a test to a string output (e.g. true or false) in bash

I'm trying to shorten a bash solution to a Code Golf challenge, which requires an output of true or false. I've managed to condense the key piece of the code down to a substring check against the sole argument to the program, of the form: [[ FIXEDSTRING = *$1* ]]&&echo true||echo false My questi...

tips bash

Seggan

16:02

just make a close() method like java

mousetail

Then it wouldn't run on error

Python has a close method too but it's recomended not to use it

it also closes on delete but since delete is not guarenteed to be called as soon as the reference counter reaches 0 it's considered too unrelaible

Seggan

@mousetail in java it does

if u use try with resources

any brb gonna take a walk

mousetail

try finally is even worse than with

even more verbose

Seggan

> try with resources

mousetail

Isn't that basically the same as a with statement?

Ginger

16:29

hi

should interfaces be able to supply default methods?

Seggan

@mousetail yeah

+ the same abilities as a normal try

@Ginger yes

Rydwolf Programs

@mousetail Isn't that like, purely a Python thing

Also who uses .open() directly in the 21st century

If Python makes you do that that's jank

Unrelated String

@mousetail i'd argue with statements are more desirable than convenient explicit closes :P

Rydwolf Programs

What does a Python with statement do?

In JS they put all of the properties of an object in scope as variables

Ginger

@RydwolfPrograms they work kinda like Rust lifetimes

mousetail

16:42

@UnrelatedString The issue is if you want to share a object between different functions. Like a function that returns a open file.

I've made multiple mistakes where I use a with statement around something that actually should outlive the current function

Ginger

@RydwolfPrograms if I do with open("data.txt") as file:, once execution of the with block finishes (or if an error occurs and execution ends early) the file object is closed

mousetail

If you close it on drop it's guarenteed to be closed and it's guarenteed nobody will use it after close

Rydwolf Programs

@Ginger Ah

Ginger

it's very useful and I'll probably put it in Rabbit

Rydwolf Programs

Yeah I'd avoid any operator named with personally, since JS and Python seem to use it completely differently

Ginger

16:44

speaking of: how should Rabbit do GC?

@RydwolfPrograms how about using? I'm doing import for imports

Rydwolf Programs

@Ginger What's a single case where it'd be useful other than for dealing with awful built-in filesystem stuff?

Unrelated String

fwiw i feel like js's with is slightly more intuitive but yeah can also just be using

Ginger

@RydwolfPrograms network connections, thread pools

Rydwolf Programs

1. An open file handle being dropped should close the file no matter what, this is 2023 people
2. You shouldn't need open() for anything but low level file stuff, and even then, rarely for that

Ginger

anything that should be cleaned up once it's done being used

mousetail

16:46

There are many things that need to be closed, like database connections, thread pools, transactions, etc.

Ginger

^

Rydwolf Programs

@Ginger Why would that need a with

Just have it do that when the object is deleted from GC

mousetail

@RydwolfPrograms What do you mean not use open? There is no alternative

Rydwolf Programs

If there's no alternative the language is bad

Unrelated String

@mousetail ideally the variable would go out of scope and have some special prohibition against being returned :P

Rydwolf Programs

16:46

You should be able to just do something like fs.read_file("xyz") to get it as a bytestring or something

As is the case in Node, Rust, etc.

Unrelated String

@RydwolfPrograms i think we have multiple ideas of what open is

Ginger

you can use os, but only masochists do that

mousetail

@RydwolfPrograms That could be a option but it's generally bad practice since it would load the entire file into memory

Unrelated String

oh yeah being able to slurp a file all at once isn't a bad idea

Ginger

2 mins ago, by Ginger

speaking of: how should Rabbit do GC?

mousetail

16:47

only usable for very small files

Rydwolf Programs

...which 90% of files you need are

Unrelated String

but quite useful for said small files :P

Ginger

I'm thinking Rust-style lifetimes

Seggan

@RydwolfPrograms which might not be in a while

Rydwolf Programs

And streams should exist too right?

Unrelated String

16:48

python does let you treat the handle from open() as an iterable

Rydwolf Programs

@Ginger What happens if I don't use a with with something that needs to be closed?

mousetail

@RydwolfPrograms Config files maybe

Rydwolf Programs

Does it silently create some sort of resource leak (very bad language design) or error (a sign you don't need with to be explicit at all)

mousetail

Most files shouldn't be loaded into memory

Ginger

@RydwolfPrograms usually Python closes it when the program ends or when the object gets GC'd, but that might be a while

Rydwolf Programs

16:49

@mousetail I can't think of a single situation where I've needed to read a file that wouldn't take up negligible amounts of RAM

Unrelated String

so a silent resource leak

Rydwolf Programs

that I wouldn't've needed otherwise

mousetail

@RydwolfPrograms Clearly you don't use enough XML

Rydwolf Programs

How little RAM do you have

mousetail

16 GB but the programs I write should be usable for people with much less

Unrelated String

16:50

i feel like incrementally reading is beneficial the moment you're even asking that question :P

mousetail

The data I work with comes in 15 GB XML files

Seggan

@RydwolfPrograms dealing with textures and gpu stuff needs closing too

Rydwolf Programs

Unless you're one of those 640 KB should be enough for anyone people, reading any file into RAM is probably a safe bet it's not gonna cause issues

I'm pretty sure the only >100 MB files I have are .isos and stuff

mousetail

@RydwolfPrograms lol

Seggan

@Ginger tracing GC

Ginger

16:51

@Seggan huh?

oh

Rydwolf Programs

@mousetail That seems like an edge case, and having convenience doesn't come at the cost of being able to do things the old way

Seggan

@RydwolfPrograms tars and zips :P

mousetail

@RydwolfPrograms I think any sane program should avoid loading files into memory if remotely possible

Rydwolf Programs

Guess I'm insane then

mousetail

You don't know how big they can get

Rydwolf Programs

16:52

A few megabytes here and there is nothing on any remotely modern machine

mousetail

If this was the only program maybe

But there is a ton going on on a typical computer

Ginger

I wonder what I should write the Rabbit interpreter in

Scala maybe d:

mousetail

At my work we use servers with over 100 GB of ram that are constantly nearly full

Rydwolf Programs

A ton of things that use a megabyte here and there

Even cheap lil' laptops come with 4 GB nowadays

Unrelated String

@RydwolfPrograms ever worked with video? :P

mousetail

16:54

At 4 GB you can not afford to load files into memory

Ginger

My RPi has 8GB of RAM

Rydwolf Programs

@mousetail I'm pretty sure I'll be fine reading a 500 byte config file

mousetail

Yea config files are fine

I'm talking about opening a file provided by a end user

Or some data

Rydwolf Programs

Or a couple kilobytes of input, or whatever. That's what 90% of file I/O is, after all, right? Config files and shit

mousetail

I'd support a convinience method for loading config files fast

But it shouldn't be the default

Seggan

16:55

@Ginger kotlin or rust

Rydwolf Programs

Well I guess if your program's a utility for doing stuff to files that just happens to be something you can do line-by-line or byte-by-byte, sure. But that's not the default any more than just reading a file all at once

And there wouldn't be any "default" anyway, just a set of options

mousetail

True

Maybe not default but "general recomendation"

Seggan

@Ginger all modern interpreted langs do that

ast eval is too slow

Rydwolf Programs

@Ginger Rust, 100%

It's low level enough that you'll have a lot of control, but will keep nasty bugs out of your language

You don't want to be the next PHP

But you also don't want to be the next Python

Ginger

@RydwolfPrograms sounds like it'll be hell to write, especially since I know very little Rust

and "learn rust by making a programming language in it" is maybe not a good idea

Rydwolf Programs

17:01

Y'know I think I'm going to start working on Cuprous

I think the world deserves a Rust that's easier to use

Ginger

~~and then I'll implement Rabbit in it~~

Rydwolf Programs

Altho soon Complement will replace all existing programming languages so there's little point

Ginger

fair enough

Seggan

@RydwolfPrograms do it

Rydwolf Programs

One of my biggest complains with Rust is Arc<Mutex<...>>

As soon as there's more than a single thread I want to jump into a pit of spikes

Seggan

17:08

type Easy<T> = Arc<Mutex<T>>

Rydwolf Programs

Barely helps

pxeger

(I feel like I should be able to flag locked posts)

Should codegolf.meta.stackexchange.com/q/25478 be no longer featured?

or, if it remains featured, could the title be changed?

user

17:23

@RydwolfPrograms And sometimes you need your own wrapper around an Arc<Mutex<...>> so that you can implement some trait for it

Rydwolf Programs

Example?

That seems kinda weird

user

I might've done it the wrong way, but I needed to pass a Read once and had to make a wrapper around an Arc<Mutex<Cursor<Vec<u8>>>> or something like that

Let me try to find it

Rydwolf Programs

That sounds like the wrong approach for sure

user

How do people usually do it?

Rydwolf Programs

The Cursor is Read right

user

17:25

Yeah

But the Arc<Mutex<...>> wasn't, hence the wrapper

Rydwolf Programs

You'd normally dereference the Arc<Mutex<...>> I think

But I don't know, maybe that is a situation where you'd need a wrapper

Wait do MutexGuards have the traits of the type they contain? Lemme check

user

github.com/ysthakur/gzp/blob/main/fuzz/fuzz_targets/…

Rydwolf Programs

Wait you can do &mut *thing I think

2

A: Calling method on Trait object behind Arc & Mutex

You are passing a reference to the MutexGuard to your runner instead of a reference to the wrapped runnable. You need to insert a deref: pub fn run<R>(&mut self, index: usize, mut runner: R) where R: FnMut(&mut Runnable), { let runnable_arc = self.runnables[index].clone(); let mut ru...

user

@RydwolfPrograms Looking at it again, the library's example uses a plain Vec instead of all that wrapped nonsense, so I guess I did overcomplicate it

Ginger

17:43

idk if I should use Rust or not

Rydwolf Programs

The answer to that question is always yes

Until Cuprous exists

Ginger

but, but, Kotlin!

Rydwolf Programs

LDQ: Is it worth distinguishing between fixed-size arrays and vectors syntax-wise?

The advantage of distinguishing between them is performance and low-level control, the advantage of not doing so is compactness since you can write a vector of strings as [String] instead of Vec<String>

Worth noting in many cases I could optimize it into a fixed size array

LDQ: print(pattern, ...stuff) vs. print("string" + stuff + "string" + stuff)

Basically should print take a string or a constant pattern and a list of things

Seggan

@RydwolfPrograms no

dont have a distinction between them, syntax wise or not

pxeger

How low-level is this language?

Seggan

17:52

thats one thing i like about python/js: no distinctions between arrays and lists

if i really want it i just write an immutable list

or a fixed list

@Ginger YES

@RydwolfPrograms just add string interpolation

print("string${stuff}string${stuff}")

13 mins ago, by Ginger

but, but, Kotlin!

ok time for a minor ad break

brought to you by Seggan

Kotlin is really close to being the ideal language imo

it has nice OOP which is not overcomplicated (like i sometimes feel about rust)

operator overloading is real nice

infix functions are sometimes nice, but i could do without them

the lambda syntax is one of the best ive seen

native function types (in a static language!), receiver types, implicit arguments

Ginger

@Seggan one thing: can Kotlin can be compiled to a native executable?

Seggan

yes

please lemme finish typing

Ginger

lol

Seggan

the lambda syntax allows for really nice DSLs (check gradle), as well as allowing you to embed code in config files n stuff

the multiplatforming ability is awesome

im currently working on a web game, and both the frontend and backend are in kotlin

it also allows you to share code across platforms, so i dont have to do a ton of duplication

back to lambdas: you dont need to put parentheses around lambda args if its the last one

so as well as having syntax like list.map { it + 1 }, you can also make functions that look like inbuilt structures

i.e. with(obj) { code } is actually a stdlib function taking an obj and a lambda

receiver objects allow you to have an implicit this in lambdas

kotlin can target JS, JVM, iOS, Android, and can compile natively

one thing ii really love as well is extension functions

oh the nightmares of making tons of utility classes in java

now youve got extension methods

primary constructors are a really nice feature, as well as specifying fields in constructors allows you to remove a bunch of boilerplate

when statements are a souped up switch, with limited pattern matching

the absence of full blown rust style pattern matching is one of my main grievances with kotlin

nullable types are mwah

no more NPEs

the interop with java is the best in any jvm lang besides probably groovy

i commonly use java libs if i cant find a kotlin lib

its integration with gradle is the best

(gradle is another topic entirely)

the large amount of functional extension functions to collection types in the stdlib i love

no more Streams, no more for loops just for a map

</ad-break>

user

18:20

@Seggan it will happen someday, kotlin is still young

I can’t wait for that and contracts to get good

@Seggan well rust isn’t meant to do oop

That snippet ginger posted was cursed af

@Seggan idk about that baut o love the intellij support for it

I once converted a small codebase from java to kotlin during a school day

Intellij is that good

Seggan

^

Ginger

18:38

ok I'm going to do it in Kotlin

how tf does ANTLR work

Seggan

18:58

ok this might take a bit of explaining

do you have intellij + gradle

New Posts

0

Q: Compact my variables (and code)!

(Notice the differences from this post. Specifically, this references JavaScript, a programming langauge that happens to be one you can run on StackExchange. However, you may use any programming language to do this challege. An example of JS looks like this (try running all the snippets in here):...

code-golf javascript

Transcript for