The Nineteenth Byte

12:16 AM

@mousetail that image is missing the best bit...

also it can be simulated perfectly with union types

type X = Y | Z
class Y...
class Z...

val x = Y()
match x { is Y(...) -> ... is Z(...) -> ... }

Bubbler

12:34 AM

Now the problem: how would you match on a union type where each branch is not a concrete type?

type Foo = (X & Y) | (Y & Z) | (Z & X)

AviFS

We're almost at 50% commitment for PLDI!

Seggan

@Bubbler i dont see why you cant just enforce destructuring just on concrete types

you can still do x is (X & Y)

and x is X(...)

Seggan

12:56 AM

alright, finished rewriting the package and function system

cracks knuckles time to start working on classes

this has been vastly simplified by the function rewrite

AviFS

1:12 AM

@Seggan What are you guys working on?

ATaco

I've implemented a Macro system in BrainChild, with the intent of it being usable to hack in Syntax.

macro ( (assignable) (/(\*\*|[*\/+-])/) "=" (expression) ) { $1 = $1 $2 $3 } is used to define x += 3

Seggan

@AviFS Rol, a language ive been mentioning several times over the past few months

its intended to be a replacement for lua for those who hate lua's syntax (me)

it compiles to lua

AviFS

@Seggan Ooh la la, link to info?

Seggan

imm currently writing up some docs

AviFS

Awesome

Seggan

1:18 AM

remind me tomorrow if i forget

lyxal

@AviFS you wanna see something cool?

that's what I've been working on :p

Ginger

1:58 AM

@lyxal what DLosc said :p

@lyxal that syntax looks startlingly like Dysfunctional's syntax :b

lyxal

I could say the same thing but the other way around :p

Ginger

eh?

what language is that, exactly?

verbose vyxal?

lyxal

Vyxal 3 literate mode

Ginger

thought so

lyxal

Implemented just yesterday too

vyxal.github.io/Vyxal

Ginger

2:02 AM

neat

lyxal

It's available to play around with online

It's limited at the moment because a lot of stuff hasn't been implemented

But enough should be there that it's Turing complete.

Ginger

at least vyxal doesn't have chain filters or stack commas or the other crazy syntax Dysfunctional uses

lyxal

chain filters? stack commas?

the hell are you on about

Ginger

I'll tell you... tomorrow q:

lyxal

I take it stack commas treat things as stack items

Ginger

2:08 AM

eh, not quite

you'll see

~~you'll all see~~

lyxal

@Ginger this is actually quite similar to how v3 literate mode looks

Ginger

here's a marginally more spicy Dysfunctional program that also happens to showcase those two features I mentioned:

yesterday, by Ginger

fib(n) -> ,1,1 swaplast ,{(##$)sum} replacelast ,{,1 (#$)sub} replacefirst ,{fib} ,{(#$)} (^#)if call

(it's a Fibonacci program, in case you couldn't tell)

lyxal

it'd be 1 10 range { ctx-n 2 mod 0 ==} filter { 10 mul } map { sum } reduce

which is only slightly different to dysfunctional

Ginger

huh, weird

~~great minds think alike (but fools rarely differ)~~

lyxal

the end could also be fold- add though

@Ginger some lag behind several months :p

Ginger

2:21 AM

@lyxal yeah the end is basically identical

Seggan

2:32 AM

@AviFS pushed syntax examples, no actual docs yet tho. a bit more info in the readme

adding example problems

Seggan

2:45 AM

LDQ: should [x, y, z, ...] be a mutable or immutable list?

Bubbler

3:03 AM

@Seggan Is it array destructuring syntax?

user

@Seggan Should be the same as what Lua uses if it's for Rol, immutable otherwise

Bubbler

Or just a list literal?

Seggan

literal

@user well erm lua has mutable lists

user

Best go with that then

Unless you don't care too much about compat, in which case immut is better

Bubbler

Also mutability is usually attached to a variable, not a value (like const x in JS)

unless you're considering something analogous to Object.freeze

Seggan

3:45 AM

@Bubbler huh

i only heard of variable conferring immutability in rust

what i mean is i cant add to an immutable list but can to a mutable

user

3:59 AM

@Bubbler If you do const x = [], the list that x refers to is still mutable

PyGamer0

@lyxal cool

Bubbler

@user Yes but x cannot be reassigned

user

But the object that x refers to can be changed

It's pretty common for "mutable" to refer to both variables and values

Bubbler

4:14 AM

Ok.

Then there's Python's tuple vs. list

user

What about it?

Bubbler

another example of object mutability

I believe Python has both because list is index-assignable while tuple is hashable

@Seggan So one possible answer to this question is "have two kinds of list literals, one for mutable and one for immutable"

user

+1 to ^

Jo King

i think mutable is more referring to whether the container is mutable or not, where both variables and lists are containers

ATaco

4:41 AM

The limitations of a 16bit Instruction Pointer are catching up to me

emanresu A

You can emulate a 32-bit cell with two 16-bit cells. Gets slow quickly though.

ATaco

5:00 AM

I could change the underlying structure to 32-bit cells instead of 16-bit cells, but that would destroy brainfuck compilablility

(That or re-write the step that I use to emulate 16-bit cells)

(Then again, a "True" BF implementation only supports 30k cells anyway, and a 32bit IP would be unuable anyway)

emanresu A

5:25 AM

Any substantial program is probably going to use more than 30k cells

emanresu A

8:39 AM

@DLosc You somehow left your filesystem structure in a source map for the HBL interpreter.

Unrelated String

9:32 AM

@Seggan but why would you

lyxal

10:58 AM

4

Copilot knows too much

Fatalize

Pretty worrying if it starts learning to suggest code based on this site’s examples

mousetail

Does it always use patterns from the project you are currently working on, or was the project just in the training data

lyxal

@mousetail I don't have any project currently open

+ even the code I have in my vyxal directory never mentions what Vyxal is nor where its website is

+ vyxal.pythonanywhere.com was way too late to be in the training data

unless of course it's always being retrained

New Posts

11:23 AM

Best of CGCC 2023 - Call for categories

6

lyxal

@mousetail the latter

I asked someone who I'm 99% sure doesn't have vyxal on their computer, and it gives the same completion

pxeger

it must have been retrained

lyxal

yeah, because copilot originally released in mid-2021 and v.pa only became a thing in late-2021

pxeger

although the fact it "knows" the website suggests it's been overtrained

lyxal

I suppose it's learned about it because it's autocompleted so much for it

if you try What is 05AB1E? it just gives [^1]

but if you try What is Vyxal, it gives

Vyxal is a stack-based esoteric programming language. It is designed to be easy to learn, and to be easy to write programs in. It is also designed to be easy to implement, and to be easy to write interpreters for

05AB1E hasn't had much of a chance for copilot to do anything

Vyxal on the other hand is at least 50% copilot generated

pxeger

11:33 AM

I conjecture that the majority of bugs in Vyxal are due to copilot

7

mousetail

Seconded

pxeger

actually if the majority of the code is copilot generated then you might expect the majority of bugs to be because of copilot

I conjecture that disproportionately many bugs in Vyxal are due to copilot

lyxal

you're not wrong

Neil

12:34 PM

UNBELIEVABLE!! 80% of people get this simple maths problem WRONG!!!

mathcat

12:44 PM

> simple

mousetail

20% of people used wolfram alpha

Seggan

1:02 PM

@UnrelatedString because it doesnt work that way?

New Posts

0

Q: Roll a painted cube

There is a 1x1x1 cube placed on a infinite grid of 1x1 squares. The cube is painted on every side, so it leaves a mark on the grid when it moves. The sides of the cube are colored 6 distinct colors, re-presentable with any 6 distinct values. A 7th value represents a blank space. The cube can rol...

PyGamer0

1:16 PM

@lyxal what!

Ginger

@lyxal so, I heard you wanted to know more about Dysfunctional

lyxal

well yes but also maybe not now

12:19am isn't good time for new language after getting frustrated at JS

Ginger

sadge

I'm pretty busy for most of your morning

lyxal

type up some stuff and I'll read as I get ready for bed

Ginger

aight

so, Dysfunctional programs make use of something that kinda resembles a stack

mousetail

1:21 PM

Daring

Ginger

for example, take the program input print, which functions as a cat program

to the untrained eye it may look like what we are doing is calling print on the result of calling input

but what really happens is that input gets called, its return value is appended to the stack, and print pops that value off and prints it

now, take the program input input print

you'd think it would take input twice and print both inputs, but that is wrong and bad ~~and if you thought it you should be ashamed of yourself how could you~~

mousetail

@Ginger That's how print(input()) would be run internally in a practlang

Ginger

it does take input twice, but the second input call overwrites the return value of the first, so only that gets printed

this is where stack commas™ come in

a functional version of that program (assuming print pops all values off the stack) would be input ,input print

note the extra comma; it tells the Dysfunctional interpreter that instead of replacing the bottom stack value, that next input call's return value should be appended to the stack

lyxal

That just sounds like a "pair" built in with extra steps

Ginger

(you can also put a comma at the end, i.e. input input, print, which prepends the return value to the stack instead and would result in the two inputs being printed in reverse)

now we get to the spicy bit: stack filters (or "chain filters", but "stack filters" is more accurate)

(also I just realized that what input input print would actually do is take input, take input again with the first input as the prompt string, and print that :p)

lyxal

1:29 PM

@Ginger joined by newlines or as a list of values?

Ginger

@lyxal that's an implementation detail :b

I'm just figuring out the syntax

so suppose we have the program 1,2,3 + print

this program will sum 1, 2, and 3, and print the sum

but what if I only want to sum 2 and 3? well that's where stack filters come in handy

they are a way to filter the stack (shocker) so that a function only operates on certain items

lyxal

@Ginger 1 2,3 + swap pop print

No need for your precious stack filters

Ginger

@lyxal it's a demonstration, lyxal

:|

as I was saying: in this case, if I only wanted to sum 2 and 3, I could change this program to 1,2,3 (##$)+ print

the bit in the parentheses is a stack filter, and it's basically saying "pass the last two items on the stack to this function and discard the rest"

@lyxal also wrong syntax lol, you probably meant 1 ,2 ,3 swap pop print

lyxal

No

I explicitly removed the 1 from the chain

Because it isn't being summed

Ginger

that program currently tries to call 2 on 1, which is an error

waiit

hmm, what that program does depends on the order in which commas are interpreted

it could either be 1 2, 3 or 1 2 ,3

anyway, stack filters use a syntax similar to regex

the ^ and $ characters are anchors for the beginning and end of the stack, respectively

# means "keep" and : means discard

and that's it, basically

for example the filter (^#:#) would keep the first and third stack items from the top and discard everything else

lyxal

1:36 PM

Can I dynamically generate these filters?

Ginger

@lyxal I don't have any syntax for that... yet

but it's probably possible

oh one other thing: you can declare lambdas with the syntax {code}, which can be moved around on the stack and called using the call function

they're used by if, for example

lyxal

So if I have ` 1,2,3 (##$)+ print` why don't I just write 2, 3 + print?

Ginger

it was an example lyxal

lyxal

I know but where would I have excess data in my stack as an example? Wouldn't I always just have as much data as I need?

Ginger

recursive functions

lyxal

1:39 PM

How do I make lists?

Ginger

and other cases I can't be bothered to think about right now

@lyxal [1, 2, 3, etc]

lyxal

Can I turn 1,2,3 into a list?

Ginger

you can do ranges and stuff with [0...] or [1...10] or [1.2.10] or such

@lyxal yes, there will probably be a function for that

lyxal

Interesting

So I can index into that list and completely bypass stack filters

Good to know :p

Ginger

...yes, I guess

mfw my exciting new features are actually new ways of doing the same thing

lyxal

1:42 PM

Where we're going we don't need stack filters! :p

Chains and list indexing all the way!

Ginger

~~maybe I should remove lists entirely~~

lyxal

It's dysfunctional not non-functional

Ginger

the stack is the only list you'll ever need :b

lyxal

Is it a real stack though?

mousetail

Maybe have only pairs? But no dynamically sized lists

lyxal

1:44 PM

@Ginger Or is it some sort of pseudo-stack

Ginger

it's not quite a stack because things can be added from both sides

lyxal

So it's a double ended queue kind of a thing

Ginger

sorta

lyxal

Where all operations try and consume everything on the thing

Ginger

the math operators are greedy, yes

functions like if will only consume the amount they're supposed to, tho

lyxal

1:46 PM

That just sounds inconvenient

Ginger

it's Dysfunctional, I don't care about convenience

lyxal

forks your dysfunctional to make it convenient

Ginger

~~call it Therapy~~

lyxal

No that's javascript

Ginger

Therapy is what you need after writing javascript

lyxal

1:48 PM

Jan 2 at 16:29, by Ginger

JS is a therapy group :p

Ginger

Jan 2 at 16:29 Ginger was wrong

lyxal

No, Jan 2 at 16:29 Ginger was Jan 2 at 16:29 Ginger

o/

Ginger

⦂|

mousetail

Language idea:

Some languages let you effectively define your own builtin. I propose a language where every function, including buit in ones and basic ones like assignment, can only be used once.

If you want to use one multiple times you must first copy the builtin to several new different names so you can use each of them once. This includes the operator you would use to redefine and copy builtins

Ginger

yknow I could probably make Dysfunctional do that

Transcript for