The Nineteenth Byte

Neil

00:50

@lyxal what am I looking at and why would it upset anyone?

lyxal

> Rust free forever

user

01:03

@Bbrk24 I mean, such sites usually are wikis (e.g. Wookieepedia)

mousetail

07:22

Pain

flawr

08:14

... au chocolat

lyxal

That stuff is tasty af

Unrelated String

yeah

mousetail

12:46

LDQ Should iterables and iterators be different things?

Bbrk24

13:07

I think so. It'd be weird for a set/array/dict/string to have a .next() method

Swift calls the former "Sequence" instead of "Iterable," but it does have the distinction. The only real requirement of Sequence is the makeIterator method, which means it's very easy to make an iterator iterable: just add func makeIterator() -> Self { self }

JavaScript Generators are what TypeScript refers to as an IterableIterator -- that is, both Iterable and Iterator, in much the same way I just described.

I don't think it would be unreasonable for a language to have interface Iterator extends Iterable, but they should still be distinct.

Like if we just consider an array to be

struct Array<class T> {
  T* const buffer;
  const int count;

  T& operator[](int idx) {
    return buffer[idx];
  }
};

How could that implement next()? You'd need to add a third stored property, nextIdx or something. Which I don't think an array should have

Rydwolf Programs

@hyper-neutrino Yeah but the wiki- part has given its meaning to the -pedia part too through osmosis. Now that vyxalpedia exists, vyxal can be edited by any internet user as well.

Bbrk24

Yes, that's how github works

Rydwolf Programs

ssshhh

Well I mean, not unless you want to make a lot of people mad

Bbrk24

You'd have to fork it probably

Rydwolf Programs

Well I mean you can edit your own copy of like, anything lol

Bbrk24

13:21

but with github you can open a PR to get your changes back in the original

(granted, they're still free to say no)

lyxal

~~laughs in required approvals~~

hyper-neutrino

@RydwolfPrograms fair enough

mousetail

I know some madlads who have a bot that automatically approves and merges PRs if they pass certain automated checks. That would create more the "wiki" idea

lyxal

that just sounds like auto-merge with extra steps

mousetail

It's literally auto-merge

Rydwolf Programs

13:26

We should do that for OpenSSL and the Linux kernel

Sure some of the changes would be bad

lyxal

@mousetail that's what I said

mousetail

You can always revert them

Bbrk24

@RydwolfPrograms I had a classmate who was convinced the entire Linux kernel should be thrown out and rewritten in Rust

Rydwolf Programs

...as it should?

Bbrk24

unironically, I could not convince him otherwise

Rydwolf Programs

13:31

Why would you want to

Every giant piece of software written in C is a ticking time bomb

Bbrk24

At some point it's easier to use C than to keep writing unsafe everywhere

Rydwolf Programs

Or you could just avoid unsafe code

And put the necessary stuff in isolated places

Which is what having to write unsafe is meant to encourage you to do

Bbrk24

@RydwolfPrograms ...you're being facetious aren't you

Rydwolf Programs

?

Vulns are discovered all the time from things Rust trivially prevents

In very very important software

Bbrk24

('tism moment, I have no idea how to read your messages)

Sure but throwing the whole thing out and starting over from nothing?

Rydwolf Programs

13:35

Sure

It'd take a while, but the kernel already supports Rust code, you could do it gradually

Bbrk24

Also, aren't there some places where a pointer can be 0x00000000 and that's not nil, it's just, a valid pointer? Does Rust have any support for that? I know Swift doesn't

mousetail

Kernel doesn't "support rust code", just modules

Rydwolf Programs

@Bbrk24 Rust is about as low level as C

(if you leave out the standard library)

mousetail

@Bbrk24 That might break, since Option<&x> is guaranteed to represent None as just a null pointer

Bbrk24

Yeah that's my point

Rydwolf Programs

13:38

But you could also just not put it in an Option

mousetail

Though you may be able to set some compiler functions to disable that feature

Rydwolf Programs

Or reimplement Option without that optimization

Which is all of four lines of code

mousetail

@RydwolfPrograms That would mean you couldn't use any libraries at all, a Option<&X> is super common

Rydwolf Programs

Libraries? In a kernel?

I don't think that's a concern

Bbrk24

Zig has different T.*.allowZero and T.*.? types iirc, they both have the same memory layout as T.*, but then if you have both allowZero and ? then it doesn't

mousetail

13:39

True you'd be using #[no_std] probably

Bbrk24

I don't know about Rust, but if you reimplemented Optional in Swift, you'd lose massive amounts of syntactic sugar (e.g. if let)

mousetail

You can use if let with custom types in rust

Rydwolf Programs

^

Bbrk24

Swift lets you say if let foo as shorthand for if case Optional.some(let foo) = foo as a special case in the compiler

Rydwolf Programs

All you'd lose is ? I think

And some convenience things like Result::ok

mousetail

13:41

There is a experimental feature that allows you to implement ? for custom types, when that gets stabilized you'd be able to implement that too

Rydwolf Programs

But, you only need the special Option in the (likely pretty rare) cases where you're both storing a pointer, and one that could be all zeroes

@Bbrk24 Yeah looking at Redox (a kernel written in Rust), I don't see much unsafe

Bbrk24

huh

Rydwolf Programs

Only ones I can find are in places where it's safely isolated and probably required, like Deref traits

Bbrk24

Still, I feel like there are better things to do than throw the entire Linux kernel out and start over. It would be better to replace it piece by piece, as bugs are found and new features are added

mousetail

For sure

Rydwolf Programs

13:48

Well I mean that's probably what they mean

Bbrk24

no it was not.

mousetail

There's a wide range of opinions on this matter, there are for sure people who would prefer to shift+delete everything

Rydwolf Programs

Wait is shift+delete a thing

What's it do

mousetail

No recycle bin, actually delete

Rydwolf Programs

Oh right

Bbrk24

13:56

conversation somewhere else reminded me of this diagram i made once

   C      ObjC     Java     Swift     Rust
 <----------------------------------------->
  struct           class             struct
composition      inheritance       composition

Rydwolf Programs

I mean ideally we'd burn every piece of software and hardware in existence and start from scratch

mousetail

Rust is close to C in many ways actually

Rydwolf Programs

^

mousetail

I don't think it's on the far side of Java in any meaningfull aspect

Bbrk24

trait

Rydwolf Programs

13:57

It's like C++ and JavaScript had a baby who against all odds went to Harvard

4

mousetail

Java doesn't really have traits though right? Interfaces are very different

Bbrk24

neither does C

Rydwolf Programs

Yeah traits are kinda a weird third thing, but most of its syntax and ideas come from C/C++

mousetail

How does that make it more Java than Java

Rydwolf Programs

Make the diagram a circle then we'll talk

mousetail

13:59

Triangle might be more appropriate

Bbrk24

@RydwolfPrograms I made this when the IQ bell curve meme was popular. I had more examples towards the middle (e.g. C++ at the same point as ObjC, JS at the same point as Java) in one version

Rydwolf Programs

ah

I mean IMO (early) C's a better language than Java. At least C was simple, and upfront about its flaws

Seggan

@mousetail traits are interfaces you can implement on anything

Rydwolf Programs

Yeah but they'

Bbrk24

idk about Rust traits, but Swift protocols let you require inits and static things

Seggan

14:03

@mousetail ofc

Bbrk24

Java/C#/TS interfaces don't

Rydwolf Programs

Rust traits let you add static methods

Since a static method is just one that doesn't take a self

Traits don't really "require" things tho, they add them

Seggan

@mousetail in kotlin, they are different, but you can iterate an iterator

Bbrk24

I already left a long paragraph about iterableiterators

Seggan

gotta add my own thoughts :P

mousetail

14:05

Yes thank you for that

Java-style interfaces are odd because the same function can implement different interfaces

Bbrk24

Swift does that too with its protocols

Seggan

@mousetail same... function?

mousetail

Method

Bbrk24

There's an experimental feature to prevent this though, @_implements

mousetail

If 2 interfaces happen to have a method with the same signature the same method will implement both of them

Bbrk24

14:06

github.com/apple/swift/blob/main/docs/ReferenceGuides/…

mousetail

Even if they have a entirely different purpose

Seggan

that shouldnt happen. if 2 methods have the same signature then... they have same/similar purposes?

Rydwolf Programs

No mousetail's talking about a single method

Seggan

yes ik, i meant the 2 methods in the 2 interfaces

Rydwolf Programs

Oh right

Yeah but what if they don't do similar things

Seggan

14:08

thats bad design

mousetail

One concrete example: 2 libraries create a interface for a event handler. One method has a boolean parameter inverted though. You couldn't implement a method to implement both interfaces

Rydwolf Programs

@Seggan On whose part

Seggan

the interface writer/implementer

Rydwolf Programs

Which one?

mousetail

Even if both variations are entirely reasonable

Rydwolf Programs

14:09

Both interfaces could make perfectly reasonable decisions

If they don't know the other exists, nothing they can do about it

Seggan

then dont implement both interfaces

Rydwolf Programs

What if you need both?

Traits eliminate this problem entirely

mousetail

What if you want to support both libraries?

A very common thing to do

Seggan

have inner classes then

@RydwolfPrograms how

Bbrk24

This is why Swift has @_implements

Seggan

14:10

how does the compiler know which function to call

Rydwolf Programs

You need to import traits to use their methods

Seggan

interesting

Bbrk24

Ah that's quite different from Swift

mousetail

Of course you can work around it but it would be much more convenient if you needed to specify the interface you are implementing to avoid bugs and confusion

Bbrk24

in Swift, you don't import individual types like in Java, you typically import whole libraries like you can in Python

mousetail

14:12

You can specify what you impelment in C#

Though in C# it's done in a weird confusing syntax

Seggan

C# has weird and confusing syntax everywhere

mousetail

True

also why name callbacks delegates? No other language uses that word

Bbrk24

Yeah that confused me so much at first

Rydwolf Programs

Does anyone remember the name for types that are a subset of the values of another type

E.g., a char that can only be a lowercase letter

pxeger

en.wikipedia.org/wiki/Refinement_type?

Rydwolf Programs

14:18

oh thanks

Things like this really annoy me:

(..., char @ ('b' | 'n' | 'r' | 't')) => { string.push(match char { 'b' => '\x07', 'n' => '\n', 'r' => '\r', 't' => '\t', _ => unreachable!() }); ... }

Bbrk24

I said float: right. Why has it gone to the top left corner

Seggan

@RydwolfPrograms @?

Rydwolf Programs

@RydwolfPrograms I have to add that _ => unreachable!() to the match

Despite the fact that char can only be one of b, n, r, or t

Which hides possible bugs

@Seggan That assigns which variant of that list of possibilities is chosen to char

@Bbrk24 is there something forcing the parent to be wide enough?

Could be it collapses to 0 width once you make its child float

Bbrk24

It was inside a flexbox and so it was following parent.alignItems instead

Rydwolf Programs

ah

Seggan

14:21

@RydwolfPrograms like python case ['b' | 'n' | 't' | 'r'] as char?

Bbrk24

I may have other layout difficulties as I'm trying to get this guy to occupy something else's margin

Rydwolf Programs

@Seggan Probably, idk Python's pattern matching syntax

I love how, since both Haskell and Curry are names for things, Haskell Curry seems like an obviously fake name

Bbrk24

why are my buttons invisible

This div contains four buttons and the word "text" four times. I see the right fifth of a button, and the word "text" once.

Ginger

they're shy

ask them nicely to come out

Rydwolf Programs

Code pls?

Bbrk24

14:37

<div id="menu">
  text text<br />
  <button type="button" onclick="setDarkMode(true)">Dark</button>
  <button type="button" onclick="setDarkMode(false)">Light</button><br />
  <button type="button" onclick="setHighContrast(true)">High-contrast</button>
  <button type="button" onclick="setHighContrast(false)">Low-contrast</button><br />
  text text
</div>

Plus a lot of sass

I tried again using grid layout and I actually see the right column now but not the left one

Rydwolf Programs

Well I mean the CSS is gonna be the issue, not the HTML

Bbrk24

Oh I know what happened

I'm trying to make this wider than it has room to be

so the buttons are hidden behind <main>

I have the inspector panel on the right, and whether there's enough room depends on whether I have the inspector panel open

Seggan

CSS is cursed

Bbrk24

Apparently, 0 auto is valid for margin but not padding

zoomlogo

14:54

@Seggan what exactly is C# tho?

Microsoft's rival to Java?

Seggan

pretty much

but its syntax is so bad i prefer java

im still trying to find a way to run kotlin in unity

or at least rust

Bbrk24

I honestly prefer C# over Java

Jacob

15:08

i think unity used to support JS but they called it unityscript or something because it was a weird variant

Seggan

i think it was a basic (the lang) variant, no?

Jacob

idk sounds cursed tho

almost as bad as shudders actionscript

Bbrk24

Aw, min(100%, max-content) isn't valid CSS

Oh, setting width: max-content; max-width: /* parent width */ does work though

wait, max-width: 100% works here? That's never worked for me before

user

15:49

@Bbrk24 I think it's generally agreed that C# is better than Java in many/most ways. It was made later and they had a chance to improve on some of its mistakes

Although reification is really weird

Unrelated String

@RydwolfPrograms it is pretty obvious that haskell is named after somebody, but it feels like way more of a last name

Seggan

16:09

@user tbh i dont see "many" differences

properties are nice ofc

Bbrk24

The way you import namespaces instead of classes means fewer using/import statements, which is nice when autocomplete dies

The fact that generics actually exist means you can have classes with the same name but different number of generic arguments, which is useful in niche cases

Seggan

@Bbrk24 ~~when you use a good ide so autocomplete never dies~~

user

@Seggan Lots of QoL features

Seggan

like?

user

I'm typing

Seggan

16:14

hrm chat got a new notification dot for pinned tabs?

its a much lighter blue

user

Getters and setters, ref parameters, that SQL-like syntax (idk why people use it but apparently they like it), reification (just looked it up and it's weird but also very useful)

Null safety

Seggan

null safety? never seen it

user

I think you can make your own value types

Seggan

refs shouldnt exist imo

@user yeah java getting those soon

reification too i think

user

@Seggan Older versions don't have it

Seggan

16:17

in any case, the syntax + MS casing/naming conventions just put me off

@user ig unity uses an older version then lol

user

I think it has covariance and contravariance too, unlike Java, where everything's assumed to be invariant

Seggan

kotlin go brr

user

@Seggan Or perhaps they simply don't bother with it (it's an opt-in feature)

Seggan

ah

Bbrk24

@Seggan learn.microsoft.com/en-us/dotnet/csharp/nullable-references

user

16:18

Yeah Kotlin's slightly better than C# but C# is slightly better than Java

Extension methods! I don't know how I forgot about those

Bbrk24

The star board just disappeared for me??

Seggan

@user Kotlin is much better than C# what are you talking about

user

How's that?

Seggan

kotlins lambdas already make it better

plus java interop, not so jank naming/casing conventions

Bbrk24

C# has lambdas too?

Seggan

16:19

but kotlin has better lambdas

user

I think Seggan's talking about syntax?

Which is subjective and a minor improvement imo

Seggan

^^ + receiver types

user

Not a huge fan of receivers tbh

I get that they make DSLs much easier to make, but still

Seggan

i love them

user

Some of these features I fluctuate between loving and hating

Bbrk24

16:21

I accidentally made a very strange-looking transition

* {
    background-color: inherit;
    transition: 0.2s;
}

Each element starts the transition after its parent finishes

user

Interesting, what's it for?

Bbrk24

I'm trying to make it so that switching between dark and light mode isn't so sudden

but that's not it

user

@Seggan Wikipedia article if you want to see more about C#

Seggan

ty

Bbrk24

Even when I only put the transition on the body, the background-color: inherit makes it look broken

no transition then ig

Seggan

16:23

@user hrm apparently javas enums seem better than c# (no parameters accepted)

user

Yeah I like Java enums

They're not ADTs but they're nice

Bbrk24

yeah Java's enums are its one good thing

Swift has Rust-like enums but sometimes I'd prefer Java-like enums

user

No, checked exceptions are also one amazing thing about it

I am prepared to die on this hill

Bbrk24

See forums.swift.org/t/new-old-idea-enum-case-blocks/60824 for context about enums

Seggan

@user then you will die

i dislike checked exceptions so much

mostly because of IOException being checked

Bbrk24

16:34

ObjC and by extension Swift use "Error" and "Exception" in the exact opposite way from Java

Generally, NSError is recoverable and NSException is not

user

lmao

If I make a language I'll be sure to have more descriptive names. Maybe Oopsie for a recoverable error and AAAHHHHHHHHH for an unrecoverable one

Rydwolf Programs

Or just don't do exceptions

Bbrk24

Honestly if a language only had Result and not throw I probably wouldn't even be mad about it, it's a logical decision to make

Rydwolf Programs

Rust

Altho it still has a lot of flaws

Bbrk24

16:49

As long as you still have fatalError or panic! or whatever you want to call it

Transcript for