The Nineteenth Byte

Bbrk24

00:13

> TypeError: Cannot read properties of undefined (reading 'stackSave')

mfw the autogenerated glue code is broken

Unrelated String

lovely

Bbrk24

Also, my postMessage calls from within the worker seem to be going nowhere -- the window.onmessage handler isn't receiving them

lyxal

worker.onmessage

not window.onmessage

Bbrk24

oh

Also I know why the glue code broke: It tried to call the wasm before it was done setting up

It works!

lyxal

00:35

👏

Bbrk24

There's a weird visual glitch but I'm pretty sure that's a CSS mistake and not anything to do with the JS/wasm

OldSandboxPosts

01:00

Sandbox posts last active a week ago: Great Leap to the Right

Seggan

01:48

how did i just get an upvote literally 2 seconds after posting an answer

l4m2

What's wrong

Bbrk24

> Annoyingly, longer than Java.

Seggan

@l4m2 its too fast

bot maybe?

@Bbrk24 yes, i am very mad at that

kotlin is never longer than java in code golf

l4m2

I'd say reasonable

Seggan

2 seconds?

Bbrk24

01:54

Maybe the question asker was staring down their inbox

l4m2

Some users upvote e.g. first answer, and I sometimes upvote first and if it's bad I unupvote

emanresu A

1

A: What's missing (aka the vanilla multiset difference challenge)

Java (JDK), 21 bytes a->b->a.sum()-b.sum() Try it online!

I doubt kotlin can beat that

Bbrk24

One thing I don't like about SE is that you can't remove an upvote after five minutes. Sometimes I'll upvote an answer and then realize it doesn't work half an hour later

Seggan

@emanresuA ok, kotlin loses by 1 byte. and its only because kotlin needs curly braces around lambdas

usually it beats java

1

Q: Must a wavefunction be normalisable for a pdf to exist?

My course notes say, for normalised wave functions $\psi(x,t)$, the function $$\rho(x,t)=|\psi(x,t)|^2=\overline{\psi(x,t)}\psi(x,t)$$ gives the $\color{red}{\text{probability density}}$ for the position of the particle. I guessed from the bold that something about this is specific to normalise...

yes, because PDFs exist

they are the bane of document editing

if i had a time machine, id prevent them from being invented

idk how many times my teacher gives an assignment and its a pdf and i cant edit it so i have to print it write on it by hand and scan it back in

Unrelated String

~~surely you mean no, because otherwise the existence of pdfs would imply that all wavefunctions are normalizable~~

Bbrk24

02:08

Fun fact: before iOS 13, the only way you could embed vector graphics in a native iOS app was by making them PDFs. Otherwise you had to use a PNG or JPG. If you added an SVG, it would be converted to a PNG.

4

Jacob

@Seggan it was me :D

Seggan

how lol

Unrelated String

@Bbrk24 WHAT

Bbrk24

Yeah. iOS development is great

Today in "typos that work for some reason": git fetch --al

-al doesn't work. I guess it just recognizes --al as an alias for --all?

Bbrk24

04:29

I found this useless-looking (but presumably needed) piece of JS code: ~~+Math.ceil((R-+(~~R>>>0))/4294967296)>>>0

lyxal

> In summary, this code snippet converts a given numeric value R to a 32-bit unsigned integer.

from chatgpt

Bbrk24

Except that if R=200, it spits out 0

lyxal

Bbrk24

lyxal

@Bbrk24 interesting

no reason for it to produce 0

ATaco

04:34

~~R>>>0 DOES an unsigned integer make

Bbrk24

R-+(~~R>>>0) is going to be 0 if the input is already a 32-bit uint

ATaco

If you want it to be a 32 bit integer, I'd suggest ((~~R)&0xFFFFFFFF)>>>0

Bbrk24

Again, I didn't write this code. I found it (in a minified JS file)

ATaco

Ah

Bbrk24

Oh! I know what this is! It might be to find the high dword of a 64-bit number

ATaco

04:36

Ah, that seems to be it.

Bbrk24

@lyxal Note the parens, the R-+... happens before the .../4294967296

ATaco

In testing it definitely does taht

Unrelated String

07:43

come to think of it why are both ~~ and >>>0 necessary

since they both should do nothing on an integer, shouldn't they both just function as coercions

Bbrk24

10:57

@UnrelatedString Some operations convert something to a 32-bit signed int, others to a 32-bit unsigned int, and others to just any integral value (even if it’s stored as a double). I assume they’re intentionally mixing the three

mousetail

11:56

github.com/snoack/python-goto

Ginger

12:23

time to bash my head against the Rabbit combined-type-resolver-and-importer machinery

like seriously I hate this

because of how I've done it I have to handle local file imports differently even though their import declarations aren't different

so I have to do lots of jank to figure out which imports are local, and then even more jank to initialize them in the correct order

there's gotta be a better way to do this

Ginger

15:35

@Seggan exciting progress has been made

Seggan

noice

i feel like working on rol but i wanna work on cgr more

Ginger

I've ended up having to run the type checker and resolver on individual files in a module due to... reasons, but it should be fine

math scat

@Seggan ~~code golf romance?~~

Seggan

CGR

Mar 14 at 19:34, 52 minutes total – 80 messages, 6 users, 0 stars

Bookmarked Mar 14 at 23:06 by Seggan

me, user, and bbrk are building it

@mathscat stands for common golfing runtime

Ginger

16:01

@Seggan so, uh, things happened

and I now have 11 different classes for imports, including 4 abstract ones

Seggan

what

i have exactly 0

as in 11 classes the represent imports or 11 classes for handling import resolution?

Ginger

16:57

@Seggan 11 classes representing imports

Seggan

AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

why

Bbrk24

I might prefer node’s entire mess (two separate import systems) over that

Ginger

@Bbrk24 lemme clarify: this is an implementation detail

to the end user it just looks like one import system, as it should

Bbrk24

It sounds easier to make a require/module.exports system than that

Ginger

it is

Bbrk24

17:10

I’m morbidly curious, why/how is it so complicated?

Ginger

lemme type :p

@Seggan there are 3 types of imports: unresolved, resolved local, and resolved nonlocal, which each have 2 subtypes: normal and from

Seggan

AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

Ginger

the other 6 classes are abstract ones

Seggan

mine is just Map<String, Set<String>>

Ginger

AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

Seggan

17:12

the first string is the package, second is the stuff imported from that package

Ginger

that's worse somehow

also, you don't have to interpret your language

I do

also also your import system is quite different from mine

Seggan

@Ginger why

Ginger

why in the name of frick are you just passing raw strings around

oh, right, you don't have to interpret anything

Seggan

17:27

@Bbrk24 ive specified the binary format exactly

DLosc

@mousetail LOL, that's amazing. ^_^

Bbrk24

@Seggan Thanks. The code looked good, I just didn't want to approve it until I was certain on how it worked

Seggan

17:54

@Bbrk24 yeah one thing is missing from the pr tho... the magic number 0xDEADC0DE

Bbrk24

I remember one time I got an error because my wasm file began with '<!DO' rather than '\0asm'. I'm still not sure how that happened

lesobrod

Inline Latex not work!

Bbrk24

On codegolf.stackexchange you have to put backslashes before the dollar signs

\$this\$ rather than $this$

lesobrod

Yeah i do this but not works!

Seggan

it doesnt work in chat either unless youve got the chatjax userscript

lesobrod

18:00

No, in the main site, in the question body!

Bbrk24

Can you link to the broken question?

Ginger

@Seggan shouldn't ConcreteType also have a field for the members of the class it refers to?

lesobrod

only screenshot

Seggan

@Ginger why, if i can do just whatever.superclass.members

Ginger

whar ?

@Seggan superclass is a ConcreteType, no?

Seggan

18:07

yes

and i can access its members too

Ginger

with what property

ConcreteType has no members property

Seggan

it doesnt?

im quite sure i added a fields or smth similar

Ginger

Is your local copy different from the one on GH?

Seggan

it shouldnt be for that class

Ginger

there is quite distinctly no members field here

so like what do I do

Ginger

18:25

@Seggan is it in the classes branch?

Seggan

18:55

maybe

Ginger

19:09

@Seggan doesn't look like it :/

Seggan

huh

nvm then

Ginger

does your local copy have it?

Seggan

cant check

Ginger

mfw

Bbrk24

@Seggan since when are 0xDEAD and 0xC0DE one byte each? :P

I already left a comment on Github, but I'm curious to hear the logic (or lack thereof)

Seggan

19:14

since my head invented the 16 bit byte :P

Bbrk24

Oh here's a neat thing: The ... operator in Swift allows you to omit one or, for some reason, both, of the bounds. Instead of slicing over 4 ..< count, you can just slice over 4....

I'm not entirely sure what the purpose of the niladic ... operator is, but you can use it to take a slice that is the entire array

Seggan

kotlin 2 is coming

apparently its not going to be breaking

unlike when other langs use a new major version

Bbrk24

Swift 4 -> Swift 5 was barely breaking. Off the top of my head, the only breaking change was that they reworked how you implement custom string interpolations. Swift 5 -> Swift 6 will be source-breaking but not binary-breaking, interestingly enough

there's also some langs that can't do source breaks, like JS and C

Seggan

19:40

JavaScript 2: Even More Cursed™

DLosc

@Bbrk24 array.slice() in JavaScript is used to make a shallow copy of the array. Would that make sense in Swift?

Seggan

or python array[:]

Bbrk24

No, in Swift a slice of an array points to the same underlying buffer. It's CoW

DLosc

Copy on Write?

Bbrk24

yeah

All the standard collection types (String, Array, Dictionary, Set) are CoW

DLosc

19:43

Interesting. So if you assign a slice to a new variable, you can read from that variable without actually modifying anything, but as soon as you modify one of the elements, it makes a copy and modifies that?

Bbrk24

Yes

I'm not 100% sure about arrays, but at least for strings, retaining a slice retains the entire underlying buffer, even if the slice is only a couple characters in a several-kB string

Knowing how malloc works, the same is probably true for arrays

The only thing I can think of is that slices are different types from the underlying collection (ArraySlice and Substring from Array and String respectively), but you can just say Substring(str) to convert it

Seggan

is it copy on first write or copy on write in general?

Bbrk24

CoW implies RC (reference counted). When a CoW type is mutated, it first checks whether refcount == 1, and if not it makes a copy.

Seggan

huh

Bbrk24

Swift uses refcounting pretty universally -- it doesn't have Java-style GC.

Ginger

19:59

me when annotation parameter type restrictions

Ginger

20:28

new Rabbit thing: The standard package-naming convention shall be author.package

so for example the built-in system module would be rabbit.system

oh, also: I really like Kotlin's trailing-lambda syntax, so I will be yoinking it

I will also be yoinking the let and also functions, except I will be renaming them to apply and with respectively

(@Seggan)

user

eek

(also -> with makes sense, but why let -> apply?)

Ginger

because you're applying code to whatever you're calling it on

hmm actually that doesn't make much sense :b

but let makes even less sense

Bbrk24

Swift has some really cool trailing lambda syntax that's I believe a superset of Kotlin's

user

In the Java world, at least f.apply(args*) means f(args*), more or less

@Ginger Yeah let isn't the best name

Ginger

@user the java world is not what I come from

Bbrk24

20:32

Let's say that a function f has labelled arguments foo:, bar:, and baz:, in that order. You can say f { /* foo */ } bar: { /* ... */ } baz: { /* ... */ }

Ginger

brb, will think up better names

user

@Ginger Right, just pointing out how it could be confusing

@Bbrk24 That looks hard to parse, though (for a human)

Bbrk24

With the right whitespace it's fine

Button {
    print("clicked!")
} label: {
    Text("Click me!")
}

user

Ooh I do kinda like that

Perhaps I should become a Swiftie myself

Bbrk24

Especially when the alternative is

Button(action: { print("clicked!") }) {
    Text("Click me!")
}

user

20:34

eek

Bbrk24

@user It's painful to use on anything other than macOS, ime, so I don't use it anymore

user

A builder wouldn't be terrible

@Bbrk24 Apple >:(

Bbrk24

@user Oh, let me teach you about @resultBuilder

// Before @resultBuilder:
var builder = Builder()
builder.addInt(3)
builder.addString("hi")
let result = builder.build()

// After @resultBuilder:
@Builder var builder = {
  3
  "hi"
}
let result = builder()

Apple built an entire declarative UI framework out of this, and it is amazing (or at least better than UIKit)

user

Nice

Simd

21:20

Is 19:55 the time on a 24 hour clock with the largest sum of digits?

Seggan

23:55?

pxeger

19:59, surely?

Seggan

ah yes

wait no

23:59

ATaco

1+9 > 2+3

pxeger

1+9 > 2+3??

ATaco

21:24

Thus 19:59

Seggan

@pxeger sry, brain waint working

Seggan

22:08

here is a complete hello world program in CGR:

0000: de ad c0 de 03 a5 50 52 49 4e 54 a4 4c 4f 41 44   ......PRINT.LOAD
0010: a6 53 54 52 49 4e 47 01 00 ad 48 65 6c 6c 6f 2c   .STRING...Hello,
0020: 20 57 6f 72 6c 64 21 02 01 00 00                   World!....

emanresu A

Unicode strings when

Seggan

already supported

the hexdumper likes ascii

emanresu A

Ah are you using utf8

pxeger

Does the CGR need a byte serialisation / syntax? Can't it just have a structured API?

Bbrk24

@emanresuA Yeah, msgpack uses UTF-8

@Seggan I'm assuming you have local changes you haven't pushed, because the only opcode I see is NOP

Seggan

22:18

i havent added those opcodes either (yet) :P

this was a manually built file

@pxeger both

Bbrk24

I mean I figured the file was handwritten

Seggan

the actual program is the last 3 bytes

Bbrk24

Why are the opcodes indices in the string pool instead of a predefined list of bytes?

Seggan

it ends up being LOAD 0 (Hello, World!); PRINT

@Bbrk24 because of the extensibility

we cant have conflicting byte ids

ive had to suffer though byte id hell when writing certain minecraft plugins

until the parent plugin switched to string ids, much easier

Bbrk24

fair enough

Seggan

22:22

@pxeger the problem is making the api the same across the language barrier (say i write a lang in python, but the interpreter is jvm). that requires some cross language format

Bbrk24

Benefit to me making overflow unspecified in Trilangle: the compiler for CGR can use a 32-bit int instead of a 24-bit one

Seggan

~~well actually the number type is an arbitrary precision complex number~~

you can still treat it as reals ofc

Bbrk24

I thought there was three types (Int, Dec, Complex)?

Or is Int only for e.g. load instructions and not arithmetic

Really the only instruction where it matters is DIV

user

@Seggan Easy solution to that: tell the Python person to suck it

3

Ginger

suck it yourself you coward

Catija

22:37

Ouch.

ATaco

@Seggan A painful past...

Seggan

@Bbrk24 int is for that yes

@Bbrk24 Dec is a building block for Complex

Bbrk24

Then why can you put a Dec in the constant pool?

Seggan

for building complexii

well, technically, you can put anything on the stack

but the operators will only operate on Complex

Bbrk24

22:52

I'm just trying to think of what you can do with a Dec

Seggan

nothing you cant with a Complex

i intended Dec to be used internally, like for construction of complexii, printing, and such

Bbrk24

No I get that. I'm just confused as to what you can do with a Dec from the constant pool

Seggan

construct a Complex

see the BaseLabel code for COMPLEX

Bbrk24

Oh! Somehow I overlooked the fact that it unpacks Complex as two indices and not two strings

I just thought of a potential problem with that but give me a moment to construct the actual bytecode

@Seggan why are we using a number followed by loose objects rather than msgpack arrays? Does the unpacker not support unpacking arrays?

Seggan

23:13

ive tried, its very wonky working with unknown typed arrays

Bbrk24

Got it

As for the problem I mentioned earlier: here's two different bytecodes that attempt to construct the complex value 0+0i

Invalid:
dead c0de 02a7 434f 4d50 4c45 58a7 4445
4349 4d41 4c02 0001 0101 a130

i.e. (deadc0de, 2, ("COMPLEX", "DECIMAL"), 2, ((0, (1, 1)), (1, "0")))

Valid:
dead c0de 02a7 434f 4d50 4c45 58a7 4445
4349 4d41 4c02 01a1 3000 0000

i.e. (deadc0de, 2, ("COMPLEX", "DECIMAL"), 2, ((1, "0"), (0, (0, 0))))

Is that analysis correct? I think the first one is invalid because constantPool[0] references constantPool[1]

Seggan

yeah you can only reference earlier entries

Bbrk24

As long as that's intentional, I have no issue with it

Seggan

i think a ZERO opcode (like ICONST_0 in java) would be nice

Bbrk24

Kotlin's implementation of Long in JS has hardcoded const val ZERO, const val ONE, and const val MINUS_ONE or something like that

I feel like an int-to-complex opcode would be helpful for, if nothing else, shrinking the constant pool

Seggan

23:40

mhm

thing is, primitives are gonna be boxed on the jvm

Bbrk24

Also, you changed the line where you check the magic number, but not the line below it where you slice the buffer

Seggan

oop

my brain is not working today

CGR

Transcript for