The Garbage Collector

Seggan

00:22

you ever get a bug, but while fixing it you go off on a bunch of tangents so ur commits are really huge?

that just happened to me: a simple pop was missing, so to find out the issue i made a debugger, and while making the debugger i revamped my Span class :P

At least I don't keep the entire source code in memory for the length of the execution anymore :P

Redz

09:59

math was great today

they told us to start calculating standard deviation

i love python so much

imagine actually typing the whole formula into a CAS who even does that anymore (the rest of my class, apparently)

The_AH

Designed a lot of programming languages.

Never got to implementing them.

I just want to design a new programming language.

But I am afraid I won't implement it.

@RydwolfPrograms Wait you proposed this site?

I just looked on Area 51.

And I found that you proposed langdev.stackexchange.com.

Ginger

11:36

@Redz one of the best parts of being a programmer is getting to do this :p

someone: "hey Ginger gimme a random number"
me, starting Python: "one sec"

eyyyyy yearling!

Ginger

11:48

which is kinda odd since my account is only 4 months old

The_AH

13:38

The next programming language I am going to make is properly going to be a golflang.

Just saying that out there.

mousetail

Esolangs are much more fun, but do whatever you enjoy

The_AH

I meant the next programming language I make is going to be a golflang.

Going to edit the original message.

mousetail

Yea I got that

The_AH

@mousetail Esolangs are nice too. Anything that's easy to parse.

What should I name it though.

mousetail

Painrose is not at all easy to parse lol

The_AH

13:40

Huh that's not good.

For anyone trying to make a Painrose compiler or interpreter.

mousetail

Trying to build one now

The_AH

...

Anyways I am going to name it Golfs or Golfv or something like that.

It will be pretty compact like I want to make it super compact.

mousetail

Why not name it "aaaaaahhh" to fit your username

The_AH

My username did not orginate from "ah".

How about "AhGolf".

mousetail

That would make it a pun

The_AH

13:42

Yeah.

mousetail

AhGolf is nice

The_AH

Going to start working on it.

After that if I have time I am going to work on SvelteChat.

mousetail

Stack based?

The_AH

Stack based probably for simplicity.

mousetail

Any unique concepts or ideas you are adding?

The_AH

13:43

I am going to make an interpreter.

Unique concepts or ideas?

I am not sure.

mousetail

Like are you doing anything differently than existing golflangs?

The_AH

Well maybe everything will be stack-based.

Like even instructions.

mousetail

Ok interesting

The_AH

This is some boilerplate whenever your making an interpreter/compiler:

mousetail

Same stack for data and code or two different stacks?

The_AH

13:45

Two different stacks.

Boilerplate is this:

import sys

try:
	filename = sys.argv[1]
else:
	print("LangName: Filename needed")

mousetail

So instead of looping you'd need to duplicate N items from the instruction stack

The_AH

@mousetail That's a genius idea.

Rydwolf Programs

@The_AH Yup

The_AH

Going to take it.

@RydwolfPrograms |Good.

mousetail

Or instead of an "if" you'd conditinally pop N items from that stack

The_AH

13:46

That will be a pretty unique programming experience.

My next programming language will be AhGolf.

I am going with an interpreter.

mousetail

Yep, would be very cool, though unsure if it would really be golfy. Interesting from an esolang pespective though

The_AH

@mousetail It would be.

More compact then if and for loops.

mousetail

Are you sure? An if or loop can be just 2 bytes, while you need to push the length in your method almost certainly costing more

The_AH

Well if the syntax is this:

<

Pop a instruction, that's just one byte.

Well that's not complete if loop syntax.

mousetail

Popping one instruction is not that useful, you'd normally want to pop an entire block, so need some way to delimit it

The_AH

13:50

What about <5 to pop 5 instructions?

mousetail

That's good, but it would limit how big a block can be

The_AH

What do you mean it will limit it?

mousetail

What if a block is more than 9 instructions?

The_AH

Then <10 to pop 10 instructions.

Or <5 then another <5.

mousetail

Ok but then you'd need bytes to delimit numbers. How to distinguish pushing 10 or pusthing 1, and then 0

The_AH

13:52

Oh...

How about <(10).

Everything in the brackets is interpreted as a number.

mousetail

That works

The_AH

But then how do we pop from the normal stack...

mousetail

but it will cost 4 more bytes than a normal loop or branch

The_AH

@mousetail Ouch.

Okay so "@i" for instruction stack and "@n" for normal stack so do @i then < to pop an instruction or @n then < to pop a value from the normal stack.

Oh hello.

We were talking about my programming language, Ahgolf, that I am making.

Right now in real time.

mousetail

The concept is really cool, I'd really love to see it explored even if it isn't very golfy

The_AH

13:54

Good.

Okay so everything after . is going to be printed.

".Hello, world!" is going to print "Hello, world!".

jfami@Macs-MacBook-Pro ahgolf % echo '.Hello, world!' > code.ahg
jfami@Macs-MacBook-Pro ahgolf % python interpret.py code.ahg
Hello, world!

jfami@Macs-MacBook-Pro ahgolf %

The first Ahgolf interpreter has been invented.

Ginger

14:36

ginger@gingersrv:/home/ginger/lycsal$ ./hello
258169372

amazing

The_AH

Huh.

Good.

Not sure how.

Ginger

I'm pretty sure it output the raw pointer value instead of the actual string

oh wait, I forgot to add a case for processing strings in vystringify :/

The_AH

The Ahgolf interpreter for some reason.

Is saying invalid value.

When the value is a valid string.

Rydwolf Programs

← 77 messages moved from The Nineteenth Byte

The_AH

@RydwolfPrograms Wait this is a real message???

Ginger

14:50

oh hey I can access mod tools again

The_AH

@Ginger Oh no.

Ginger

will there ever be a situation in which it is impossible to statically analyze the metadata of a type?

The_AH

@Ginger Yes.

Ginger

you say that with extraordinary confidence considering you have no idea what I'm referring to

The_AH

I do.

Your saying if there will be a situation in which it is impossible to statically analyze the metadata (assuming this is the details, attributes, etc) of a type.

I am assuming builtin type here.

Monad is a difficult one.

Ginger

14:53

if the answer is no (which I think it is) then I don't need to preserve things like the length of an array into the LLIR

so I can just use TypeSpecs for everything

Rydwolf Programs

@Ginger Ohh so is this about whether all types will be sized?

Ginger

buddy you have no idea what I'm talking about

@RydwolfPrograms pretty much yeah

Rydwolf Programs

You can't know

But for anything ?Sized you can just put it in the heap and use a pointer to it

Ginger

actually nevermind

while I was typing a message to provide context I realized that I'm being silly and actually don't need to worry about it

can't believe I was trying to figure out how to stringify... a string

LLVMPositionBuilderAtEnd(builder, stringPath)
LLVMBuildRet(builder, valuePtr) // that was easy

hmmmm

it just occured to me that TypeSpecs having LLVMValueRefs with their metadata could pose a problem

since it'd be pretty easy to imagine one of those LLVMValueRefs being referred to in a context it doesn't exist in

but then again element implementations are effectively inlined into the main function and therefore share a context

ayyy, it worked!

Seggan

15:28

@The_AH yes. Any room owners can move messages from the room they own to any other room they can access

Ginger

15:44

ugh, I'm going to have to figure out how to store the length of an array alongside the array itself

I think I have two options here:

1. store the array and its length together in a struct

user

What's the second option? Slices like Rust?

Ginger

2. do what Clang does and have the array's length be a stack variable that gets kept around for as long as necessary

user

I like the first better

Ginger

unfortunately, it's both harder to implement and harder to optimize

user

oof

What about slices?

Ginger

15:54

what about them? :p

user

I'm typing

Ginger

I'm not really sure what you're referring to

user

That's because I haven't given you any pointers yet :P

Ginger

well hurry up and do that so I can run mem2reg

user

Let me type my damn message

The array is stored on the heap or wherever without any length, and instead of just having a pointer to the array, you have a pointer and the length

At least, I think that's how Rust stores slices, take this with a shaker of salt

Ginger

15:56

hmm

so currently arrays (and everything else) are stored on the stack

user

I have no idea what it was about option 1 that made it hard to optimize, so I don't know if that would apply here too

The_AH

Rust has brilliant design.

Along with Python and Javascript.

Every other language is too complicated.

For a person to understand.

user

@Ginger What happens if an array escapes a function?

Ginger

basically, a struct containing 1. an array, 2. a type tag, and 3. a length is unweildy and annoying to manipulate

@user they don't really, since "elements" are inlined into the main function

user

Wait, does LYCSAL not have user-defined functions?

Ginger

15:59

but LLVM's pointers can cross function boundaries without issue, so it wouldn't matter even then

@user not yet, maybe later

keep in mind I only have two elements right now and one of them is print :b

user

Well if you're going to add them later, you need to start thinking about them now

Ginger

I'd rather like to have non-broken types first though

user

If a function returns an array stored on the stack, the function's stack frame will be cleared when it returns, and anyone trying to use the array will probably get garbage

noodle man

@The_AH many would argue javascript is very poorly designed. after all, Brandon Eich made it in like 5 days

Ginger

hmm, good point

user

16:01

Just go with the heap

Ginger

right

but then how do I store the contents of the array?

user

Also as references, I guess

Ginger

my current solution involves type tags and pointers and is annoyingly complex

noodle man

@The_AH there are many other languages. Ruby, Elixir, Swift come to mind

Ginger

basically, since pointers and numbers both have a fixed size, I store those in the array

@The_AH wow, I didn't really process that message

user

16:02

Right, but arrays can contain more arrays, so you can't store everything inline

@Ginger I'm assuming it's satire?

Ginger

Rust I agree with, Python is kinda iffy, but JS???

@user I'm pretty sure it's not

user

Rust I'd definitely put under the "complicated" category, even if it's better designed than a lot of other languages

Ginger

fair

@user if I can use stack pointers and heap pointers interchangeably than this becomes much easier

since the compiler naïvely puts every value on the stack

user

Are numbers just 64-bit floats or proper rationals?

Ginger

yeah uh about that

user

16:05

I guess if they're just floats, putting them on the heap and only storing references to them wouldn't make much sense

Ginger

the compiler currently uses i64s for integer VNums and doubles for everything else

right, I think this solution will work: arrays are stored on the heap, and contain "structs" with a type tag and a pointer (stack or heap) to the element itself

I say "structs" because they're really just an i8 and pointer next to each other

user

So every "element" of your arrays contains a pointer/int/double and also a type tag?

@Ginger That's what structs are :P

Ginger

currently? yes

@user oh right lol

user

If the array only contains pointers, then you don't need it to also store a type tag, right? Because the object that the pointer points to will have a type tag?

Ginger

well uh no

user

16:08

@Ginger Oh did you mean to say it contains a pointer or a number here?

Ginger

normal values don't have type tags because the compiler can statically figure out types

@user no, it always contains a pointer

user

Ah

Ginger

if it's a number value being stored the pointer goes to the stack, else it goes to the heap

user

@Ginger Wait wow this is pretty amazing

Ginger

(strings are just arrays of bytes and can probably be treated in a similar fashion)

user

16:10

What if you return a number from a function? You can't store them on the stack either, right?

Ginger

you can

everything can be (and is) stored on the stack

user

Oh, do you make a copy of the number?

@Ginger No I mean in the future, when you have user-defined functions

It's better to plan for that now than to radically change how things work later

Ginger

ugh yeah you're right

wait, then why does my vystringify code work?

that returns a pointer...

user

sprintf doesn't return anything

Ginger

yeah, I'm creating the buffer

the buffer vystringify returns is allocad inside it, returned from it, and yet still usable outside it???

something terrible is going down (through the entire town)

user

16:14

Maybe that part of memory happens to not get clobbered by the time you use it?

Ginger

> allocad memory is automatically released when the function returns.

user

Oh wait what

Ginger

but I guess it's not actually cleared

so when vystringify returns, the pointer it returns points into unallocated memory

but since that pointer is printed directly after the return, nothing has time to overwrite the value there

that's extremely cursed

The_AH

16:41

@noodleman Well yes but after years it has improved.

@noodleman I have not programmed in any of these but I have seen their syntax and it isn't the most appealing. To me, C and C++ also come to mind. They are too complicated but essential in this world.

@Ginger Rust, Python and Javascript are greatly designed. Every other programming language is too hard to learn (including C and C++ but they are essential for this world)

noodle man

@The_AH this is a false statement, very broad

The_AH

@noodleman Well what I meant is almost every other programming language, there are still tons of languages out there that are not hard. I apologize.

noodle man

javascript has poor design, even if the language has grown to not be terrible. python is arguably not very well designed.

i just think you chose bad examples

Ginger

JS is a mess, and I say this as a person with considerable experience with it

noodle man

same here

Ginger

16:49

Rydwolf has even more experience and would (probably) agree

The_AH

Python and JS are the best languages ever.

I have been using JS for years.

Python is my first programming language.

I have a lot of experience with them.

noodle man

have you used, like, any other languages?

The_AH

I can tell that they are good.

Well I have of course.

Ginger

I love Python, it has a special place in my heart, but its package management mechanism is the worst of any language I have ever used

noodle man

they are good, until you put them in comparison with many other languages

The_AH

16:50

Maybe it's just because I don't use npm much but I like Python's package management.

Ginger

I conceptualized Rabbit specifically because of how much I despise Python's package mechanism

The_AH

Rust is also good.

I have used it a bit.

noodle man

python package management sucks. npm is..okay? but not great

Ginger

having experience with something doesn't make it better, as the MediaWiki developers can attest

The_AH

Dang, I haven't really used any other package management.

Maybe my standards are low because of that.

noodle man

16:51

you can admit JS has flaws right?

The_AH

yes.

JS has many flaws.

Ginger

and yet you still say it's the "best language ever"

The_AH

But it doesn't entirely suck, it still is good to an extent.

@Ginger I meant a good language.

noodle man

@The_AH i agree with this

Ginger

@The_AH then why didn't you say that?

noodle man

16:52

but it’s not fair to say it is a great language and every other language is too hard to understand

The_AH

I thought it would be clear.

@noodleman I really shouldn't have said that.

Ginger

I personally disagree with even that assessment, but you seem to not really know what you mean yourself

is it good or is it bad? what's your position?

The_AH

As someone who has been programming in JS for 4 years and in Python for 5, I do know what I mean myself.

@Ginger good but not great.

Ginger

your opinions are your own problem, but at least have them be consistent

The_AH

They are consistent.

Ginger

16:53

@The_AH unfortunately, we are not you and therefore do not

The_AH

I am just not good at explaining things.

Ginger

4 mins ago, by The_AH

Python and JS are the best languages ever.

The_AH

Well of course I will say that, they are my primary languages but that was an overreaction to some extent I admit.

Ginger

so it was hyperbole?

that's fine

The_AH

Yeah.

Ginger

16:54

but it sure wasn't clear

The_AH

I think a lot of programming languages suffer in one aspect.

noodle man

i would have read it as sarcasm if say Ginger said it

The_AH

So the aspect a lot of programming languages suffer in is error handling.

Go's error handling is horrible, Python's is not horrible but still not the best and JavaScript's error handling I haven't really used much so I cannot comment on that.

noodle man

JavaScript’s kinda sucks, you can’t specify which error will be thrown by a function, you can throw things that aren’t Errors, stuff like that

The_AH

@noodleman Hmm, I will need to use their error handling more to form an opinion.

noodle man

16:57

rust has pretty good error handling from what i’ve read though i haven’t used it myself

The_AH

Rust error handling sounds appealing but I haven't explored it much.

Rydwolf Programs

Rust's error handling is the worst part of it IMO

I hate it

The_AH

Hmm...

user

Yeah it sounded real cool when I was looking at the docs but it gets annoying real quick when you have multiple kinds of errors and then you just usinganyhow::Error everywhere

Rydwolf Programs

Not sure about anyhow but yeah in tons of Tokio stuff you just end up with Box<dyn Error>

Which is nasty since dynamic dispatch

And you can't match on it

noodle man

17:04

matching sounds like a nice luxury to someone like me who mainly has experience with JS

there’s actually a proposal to add matching to JS but it kinda sucks

Rydwolf Programs

Yeah pattern matching and tagged unions is suuuuch a nice feature

One of the few things I really miss when using JS

noodle man

github.com/tc39/proposal-pattern-matching

Rydwolf Programs

Without enums tho, pattern matching's pretty limited

Since a lot of the niceness about matching is the compiler can make sure you don't miss a case

Well unless you could somehow get it to work with ghetto enums. Like maybe if you used symbols in an object and passed the object to it

The_AH

PHP is probably in the middle.

It's not a bad language but not a great one either or a super good one.

Rydwolf Programs

ehhhh PHP's the butt of every joke for a reason

noodle man

17:11

PHP has gotten some nice quality-of-life improvements in recent years though

The_AH

Exactly noodle man, PHP is way better then people think it to be.

Rydwolf Programs

I remember trying to use it and having to have 50 manual pages open at all times since all the builtins take args in the weirdest order and they're all top-level functions instead of methods

noodle man

i watched some video called “PHP is not dead” or something

The_AH

@noodleman A video like that changed my perspective on PHP.

Before I thought it was absolutely trash and inferior to JS.

It still is inferior to JS but not that much.

Rydwolf Programs

I mean PHP and JS are kinda for entirely different purposes

PHP's an HTML preprocessor and the closest JS gets is a back-end (but using Node for prod is kinda bad)

The_AH

17:14

JS was initally designed as a HTML preprocessor.

Ginger

that seems dubious

got a citation?

The_AH

One min.

Wait it wasn't?

Rydwolf Programs

Even if it was planned to be it certainly never actually was

The_AH

Why didn't anyone tell me this?

I my entire time using JS thought that.

noodle man

@RydwolfPrograms citation-needed

(for the bit about Node in prod)

Ginger

17:25

alright, I've done enough LYCSALing for today

The_AH

What is LYCSAL?

Ginger

the Great Array Rewrite *thunder crash* begins tomorrow

@The_AH the Laconic vYxal Compiler System tArgeting Llvm

it's a compiler that compiles a subset of Vyxal to LLVM IR

The_AH

@Ginger Okay and you rmaking it and rewriting it.

Oh wow interesting.

Ginger

I am the ~~only~~ main developer, yes

The_AH

Want to hear all of my current projects?

Ginger

17:27

not really

The_AH

The list is long.

And there are some interesting ones in there.

Ginger

you do understand that we can't physically stop you, right? :b

The_AH

Well I shouldn't have asked in the first place really.

Ginger

dontasktoask.com

The_AH

Okay.

Here is the list.

SvelteChat
Noml
mytm
Ahgolf
Rbuilder
XTTP (and XBrow)
Youtubl

There are more but these are the most notable ones.

Some of the names can be very cryptic because I like to use short abbrevations.

noodle man

17:36

is sveltechat a chat made in svelte?

The_AH

18:59

@noodleman Sorry for the late reply, yes it is a chat made in Svelte and it will have a lot of features including the features of this chat.

Rydwolf Programs

Ooh will it include the bugs too

Ginger

I'll only use it if it has UTC midnight emulation d:

The_AH

@RydwolfPrograms What bugs?

@Ginger Sure. Not sure how to implement that.

The_AH

19:32

@RydwolfPrograms Oh there won't be any bugs.

Ginger

19:43

gotta admire that confidence

New Posts

0

Q: Unit Testing Built Into Language

I'm mostly looking to poke holes in this design, as well as suggestions on next steps. For the language I'm creating, I think I might have additional keywords and statements available just for test files. What are the issues with this? test suite // not sure what exactly to name this, but it mat...

Ginger

@NewPosts this is absolutely a bad idea

I can't immediately think of a single situation in which this would be useful

The_AH

I am making an account.

I do language design very often so I need it.

@Ginger Thanks.

I also have my first question pretty clear.

So my question will be about interesting loop syntax that will revolve around stacks.

I am spefically going to ask which syntaxes are better then others.

user

19:59

Be careful, that sounds like it'd be a very subjective question

Seggan

21:26

@Ginger erm... rust has it

@The_AH ^^

Redz

21:46

@Ginger wait huh yearling how does that work

Ginger

good question

David Young

@Ginger I've used something like this before and I like it for some small things. I have a directive %test "<name>" <type>: <expr> that means "When the testing framework looks at this file, in a test named <name> evaluate expression <expr> and print the result. Then compare to the golden test file"

Ginger

hm

Redz

is that for a badge on PLDI

Ginger

you get yearling after being active on a site for a year

Redz

21:48

i know that

but was yearling on PLDI

Ginger

yes, which is interesting, since PLDI's only existed for 4 months

Redz

huh

i don't see it on your badge list

Ginger

...oh wait

I got it on Area 51, of all things

misread the badge icon lol

Redz

oof

lyxal

23:07

6

Q: Rename the 'Blatantly off-topic' close reason

I noticed a community-specific close reason was created and called 'Blatantly off-topic.' However, for users without the close votes privilege, the 'Other' close reason is renamed to 'Blatantly off-topic.' I think it would look weird to <500 users to have 'Blatantly off-topic' listed twice. I thi...

feature-request status-completed close-reasons

status-complete

(cc @user16217248)

Ginger

@Seggan in what form?

Alexis King

@supercat Please stop arguing about the purpose or intent of the C standard in comments. Comments are just not a good place for that kind of discussion because they are limited, noisy (they ping the post author every time), and there is no way to downvote them. I know from our previous discussion in chat that you have strong opinions about this, but having a strong opinion is not, on its own, sufficient reason to leave a comment.

Ginger

let's not star that, folks

reprimands shouldn't be on the starboard

ATaco

23:47

@Ginger ★

3

Transcript for

The Garbage Collector