Vyxal

lyxal

04:00

@user HEY BINGUS! YOU FORGOR 💀

That's the best yell I could come up with :p

If you want, you can summarise what it was you wanted to do and I'll see if I can do it later

user

04:50

@lyxal Thanks, working on it rn

@lyxal I'm trying to think about what I want to do :P

Once I can put it into words, I'll tell you, yeah

lyxal

:+4:

The colour is yellow btw

user

Yellow?

lyxal

4 mins ago, by lyxal

:+4:

user

Do we need to make Context an abstract class and have separate subclasses for toplevel contexts with their own variables (like functions and the top context) and for child contexts (like if statements, while loops, etc.)?

@lyxal The color of the +4 emoji is yellow?

lyxal

@user no you bingus, I played a draw 4

user

04:58

If we do that, for map and filter lambdas with multiple branches, we could make a map-lambda-specific context that's a parent to all the contexts created for the branches. All the variables created in the branches would then be put into that shared parent context

@lyxal Oh

I drew a 4

lyxal

@user what's wrong with passing context between function calls

@user I said the colour is yellow. This isn't yellow

user

Because then the contexts get nested

So when we implement a debugger, the context for the third branch will be inside the context generated for the second branch, which will be inside the context generated for the first branch

Which might be a good thing depending on how we want things to work

Or a bad thing

@lyxal Your card was yellow. This is a wildcard 4

lyxal

@user What if execute function but use the context provided function?

That way, it doesn't return a new context

user

idk, it's been a while since I played Uno because my cards got stolen when coming to the US 🩻

lyxal

Alternatively, add a use this context parameter to execute

user

05:01

@lyxal Yeah I've been thinking about that

But to create the context, you need to do all that input-unpacking stuff first

lyxal

@user my card was the wildcard. Everything after that has to be yellow

@user so then if a context is provided, modify the variables attribute of that

user

@user The funny thing is, whoever stole it (probably airport workers, because the whole deck was there while packing) left behind the little cardboard box the deck came in, the instructions, and the 2 blank cards provided with the deck

@lyxal That works

lyxal

@user wait unironically you lost them?

user

Should I try implementing that?

lyxal

@user sure

user

05:03

@lyxal I didn't lose them, they were stolen

lyxal

@user but unironically it really happened?

user

I lose a lot of things, but there's no way I lost only the usable cards

@lyxal Yes lol

lyxal

I thought it was one of those jokes where we pretend we did something when we didn't lol

user

Ah

lyxal

Sounds like you need to play Schroedinger's uno then

Ginger made it so that you can play uno over SE chat

user

05:05

I invented Schroedinger's Uno

I sold the cat to Schroedinger, along with some torture instruments and a Geiger counter

lyxal

Not that one, the one where you see your cards in an image only you can see because cookies

user

Ooh, that sounds cool

lyxal

It actually is

We played it in the sandbox chatroom a while ago

emanresu A

It was buggy

It worked until Radvylf committed identity theft

lyxal

It also got rate limited really badly

chat.stackexchange.com/transcript/1/2022/10/28/…

That's a game in action btw

user

05:10

How do I join?

What's ctxArgs btw? Is it just like inputs but overrideable?

lyxal

@user you don't

That was a game in action

user

:(

lyxal

Bot hasn't been around for a while

user

Ah I misread That's as There's

lyxal

@user let me check the source code to see where I used it real quick

user

05:13

@lyxal That's what quantum bomb detection'll get ya

@lyxal Only one place

case AST.ContextIndex(index) =>
        val args = ctx.ctxArgs.getOrElse(Seq.empty).reverse
        if index == -1 then ctx.push(VList.from(args.reverse))
        else if args.length < index then ctx.push(ctx.settings.defaultValue)
        else ctx.push(args(index))

In Interpreter.scala

In execute, line 167, to be more precise

lyxal

Trying to remember why I used that instead of inputs

Oh yeah right because inputs is an object

I needed a list

user

Actually I think wait passing in a context doesn't really solve the problem since each function has different inputs and should probably be in a separate context, we can just do what we were doing and pass in the mut.Map for the variables instead

lyxal

@user Oh oh also also because yes it is overwriteable

It's so that generator structures can append the terms they generate to that seq

user

I think we could just add a method to Inputs to get the original inputs as a list

Permission to do ^?

lyxal

No it needs to be appendable

user

05:19

Aww

lyxal

Because the generator structure would otherwise not know about all the previous terms generated

user

@lyxal Do they have to store the terms they generate in the context? Can they store it somewhere else and pass that on?

lyxal

@user well where else can it be stored between lazy evaluation calls?

It's so you can do something like relation `3` `2` add end to get the 3rd and 4th last generated terms

(0 indexed)

Maybe it should be 1 indexed

user

@lyxal They can just be passed to the function, right?

Wait lemme look at how it's implemented before saying more

The generator method in Interpreter seems to have a previous parameter for that

def generator(
      relation: VFun,
      ctxVarPrimary: VAny,
      ctxVarSecondary: VAny,
      arity: Int,
      previous: Seq[VAny] = Seq.empty
  )(using ctx: Context): LazyList[VAny] =

lyxal

It does have it, but also note that what's passed to inputs isn't the same as what's passed to args

user

05:25

Right, overrideCtxArgs

lyxal

I don't know why I didn't pass it to args though

Oh actually I do

Because most of the time, you want inputs to be the args for the context

But with generators, you want every term generated to be args

Inputs only takes the last (arity of generator) items that have been generated

user

Why not let inputs be List(previous)?

That is, there would only be one input, all the generated items wrapped in a list

lyxal

1 min ago, by lyxal

Inputs only takes the last (arity of generator) items that have been generated

user

That's how Husk does it

lyxal

Husk isn't stack based

And then there's explicit arity too

user

05:30

Argh

user

05:48

@lyxal Would it be acceptable to use the secondary context variable instead of a special ctxArgs variable for the list of items that have been generated?

We have a Context class and two context variables already, it'd be a bit confusing if we had another variable called ctxArgs. And even if we named it something else, it's a lot of variables to juggle (we even have a register, although that's global)

The secondary context variable is already used to store all the inputs, storing all the previous terms instead would be in that spirit

lyxal

06:36

@user it's kind of a third context variable instead

@user I want it so that you can write relation n m add from <whatever> end and get a fibonacci sequence

I want it so that you can always easily access the last 2 terms

@user that behaviour should be changed then

The tertiary context variable is golfier anyway

user

06:56

Alright, approving

Forgor

lyxal

What should the secondary context variable do in normal lambdas?

lyxal

07:09

Also, ¤ is the tertiary context variable

It can either push all args in a context or the nth arg in a context

n¤ pushes the nth arg and ¤ pushes all

n ¤ is not the same as n¤

(in these examples, n is some number token)

It binds number tokens that are next to it

But not when there's whitespace

lesobrod

Hmm, may be a bug? `ÞK` (Suffixes of a list) works well with strings too.
But `K` (prefixes) does'nt work with strings. Is there another operator for string prefixes, or may be check `K`

lyxal

Cumulative sum on strings

¦

lesobrod

@lyxal Thanks!!

lyxal

07:34

No problem :)

mathcat

08:32

@lyxal I realised at 11:30 PM that I completely misunderstood what you meant :/

Mind if I open another PR? ^_^

First though, should the starting vector be retained in the output as well?

That doesn't seem very logical, but you could do stuff like 4ɾλǔ;İ (get all rotations of a list).

Moreover, duplicate and append is a byte longer than behead. :P

In both cases, this could be fixed with an additional line of code.

lyxal

09:21

@mathcat 👍

@mathcat no, because there can be another function overload for including it

mathcat

Is it worthy of an overload or would a digraph suffice?

lyxal

Monograph

mathcat

... what

lyxal

Single character

A monograph is a single character, a digraph is two characters and a trigraph is three characters :p

mathcat

09:39

well that makes sense

anyways

mathcat

09:55

The expected output for the first test case here would be [1, 2] then?

Cuz 0 is the starting vector.

lyxal

10:32

By doesn't include, it means the return list doesn't have it at the start

It can show up at the end or anywhere in the middle

mathcat

oh

mathcat

10:47

I don't see why removing the if-statement wouldn't work, then.

Is f = lambda n: 1 if not n else 0 supposed to output [1, 0]?

Oh, it wouldn't work for f = lambda n: 0.

mathcat

11:15

bot ded.

lyxal

yes

11 hours ago, by lyxal

@Ginger where bepp bopp

lyxal

11:39

dang

I've gotten way too used to element documentation being in the scala code :p

I'm adding Jelly-like convolve and thought just adding it with the decorator was enough

mathcat

lmao

What's the difference between Jelly convolve and Matlab convolve?

lyxal

no clue

I just know they do different things

different things with 1d arrays

Matlab convolve seems to be more advanced/general

Jelly's seems to be more primitive

but it's good to have for porting purposes :p

@mathcat your implementation works with 2d arrays. Jelly's is purely 1d

mathcat

huh

lyxal

 output[i][j] = sympy.nsimplify(numpy.sum(rhs * lhs_cutout)) # matlab style
 result[i + j] = add(result[i + j], multiply(x, y, ctx=ctx), ctx=ctx) # jelly style

Ginger

12:11

lyxal

12:44

!!/bot where the hell are you

!!/don't make me throw potassium at you

!!/I swear I will dip you in battery acid if you don't appear soon

Ginger

13:21

ffs

one second

I'm restarting it

lyxal

I can and will put it in the acid lake

Vyxal Bot

Well, here we are again.

Ginger

It's always such a pleasure.

lyxal

@VyxalBot you got lucky this time

you get to stay away from the tiny houses for now :p

Ginger

and of course the bot says nothing, because it's not sentient

lyxal

13:26

It's too busy riding down the acid tunnel of love :p

~~man that section was trippy to play at 10/11pm~~

~~at least there wasn't a bug which made the section go on for too long like there was with the music trio fight~~

mathcat

@Ginger Can't you just make it sentient?

Ginger

oh right, import soul :p

mathcat

soul.create("vyxal-bot")

Ginger

14:00

!!/status

Vyxal Bot

@Ginger I am doing a date with someone I found on Tinder. Hopefully she likes beans and crappy code-golf in-jokes

Ginger

cool

Ginger

15:07

!!/status boring

...

frick

Vyxal Bot

Well, here we are again.

Ginger

> [Room-106764] CRITICAL: An error occurred while recieving data!

...

that's new

mathcat

15:18

nice.

parz

*receiving

Bots making spelling errors?

!!/don’t make spelling errors or angry, amateurish grammarians will come for you

Vyxal Bot

Sorry parz, I'm afraid I can't do that.

parz

!!/Why is your profile picture eating a cookie?

Vyxal Bot

Sorry parz, I'm afraid I can't do that.

parz

.

!!/say something other than “sorry parz, i’m afraid i can’t do that.”

Vyxal Bot

15:23

Sorry parz, I'm afraid I can't do that.

parz

AUAUUDJJDIIIUAAAUISHSKANA

Ginger

._.

user

15:57

@VyxalBot Find some tinder and burn the restaurant down

Ginger

16:16

!!/status boring

Vyxal Bot

Bot status: Online
Uptime: 1:08:34.991512
Running since: 2023-02-22T15:08:14.757299
Errors since startup: 0

Ginger

good

mathcat

!!/u good bro?

Vyxal Bot

@mathcat I am doing unspeakable acts. I'm so sorry

mathcat

!!/brew a coffee for me

Vyxal Bot

16:24

Sorry mathcat, I'm afraid I can't do that.

ysthakur created branch flexible-status in Vyxal/VyxalBot2

ysthakur opened pull request #15 (Don't prefix statuses with "I am doing") in Vyxal/VyxalBot2

mathcat4 commented on #15 (Don't prefix statuses with "I am doing"): "Wouldn't the status start with a lowercase letter now?"

hyper-neutrino approved #15 (Don't prefix statuses with "I am doing"): "LGTM"

GingerIndustries merged pull request #15 (Don't prefix statuses with "I am doing") in Vyxal/VyxalBot2

GingerIndustries deleted branch flexible-status in Vyxal/VyxalBot2

Ginger

!!/make me a coffee

Vyxal Bot

@Ginger Here's your coffee: ☕

Ginger

thamks

mathcat

!!/status

Vyxal Bot

@mathcat I am doing things like stealing things from Fort Knox. Or was it the Moon? No, definitely not the moon

Vyxal Bot

16:53

ysthakur created branch i-am-not-doing in Vyxal/VyxalBot2

ysthakur opened pull request #16 (Don't use "I am doing" phrasing for some statuses) in Vyxal/VyxalBot2

ysthakur requested GingerIndustries's review on #16 (Don't use "I am doing" phrasing for some statuses)

GingerIndustries requested changes on #16 (Don't use "I am doing" phrasing for some statuses): ""I am poping""

mathcat

> I am doing dQw4w9WgXcQ

!!/sus

Vyxal Bot

@mathcat ඞඞඞඞඞඞඞඞ

Ginger

me when sus

hm

sus

test sus test

mathcat

sus

Vyxal Bot

@mathcat ඞඞඞඞඞඞඞඞ

Ginger

17:06

ok, it does work

cool

Vyxal Bot

17:28

Steffan153 approved #1643 (Add a convolve element that behaves like Jelly's)

mathcat4 approved #1643 (Add a convolve element that behaves like Jelly's): "looks good"

mathcat4 enqueued pull request #1643 (Add a convolve element that behaves like Jelly's) in Vyxal/Vyxal

mathcat4 merged pull request #1643 (Add a convolve element that behaves like Jelly's) in Vyxal/Vyxal

mathcat4 deleted branch fix-convolution-with-1d-lists in Vyxal/Vyxal

mathcat4 closed issue #1533 (Convolution builtin) in Vyxal/Vyxal

Lyxal deleted branch gh-readonly-queue/main/pr-1643-95f6fb9b4db67bed7552a6e05ba94ce96aaf2012 in Vyxal/Vyxal

mathcat

Remind me again, why was the number of reviews required reduced to 1?

Jacob

18:10

sus

hm

!!/sus

Vyxal Bot

@Jacob ඞඞඞඞඞඞඞඞ

Jacob

how can i get the first n items of a list

nvm found it

mathcat

18:28

!!/permissions list 533049

Vyxal Bot

@mathcat User Ginger is a member of groups discussion, member, admin.

mathcat

huh

~~I think the bot's just ignoring you.~~

sus

welp now it's ignoring me too

Ginger

no, there's only a 1 in 4 chance of it replying

did that intentionally

Jacob

sus sus sus sus sus sus

Vyxal Bot

@Jacob ඞ

Jacob

18:30

if i say it four times in one message are the odds increased

mathcat

sus sus sus sus sus sus sus sus

Ginger

no

mathcat

aw man

Ginger

me when creeper

Jacob

did u remove that thing where it can run vyxal code

mathcat

18:30

sus

Jacob

!!/help

Vyxal Bot

I'm Vyxal Bot 2.0.0 made by Ginger and hyper-neutrino.
All of my commands start with !!/
If you want help for a specific command, do !!/help <command name>.
New to Vyxal? Try !!/info for more information.

Command list: amilyxal, blame, coffee, cookie, die, goodbye, groups, hello, help, hug, info, issue-open, maul, permissions, ping, prod, register, repo-list, run, status, sus

mathcat

!!/amilyxal

Vyxal Bot

@mathcat You are not lyxal.

Jacob

you should make !!help with no argument be a synonym for !!info

!!/permissions list 554550

Vyxal Bot

18:31

@Jacob User Jacob is a member of groups discussion, member.

Jacob

!!/help run

Vyxal Bot

@Jacob No help is available for that command.

Jacob

!!/run kS

Vyxal Bot

Sorry Jacob, I'm afraid I can't do that.

Jacob

hm

mathcat

18:32

Feature request: Lyxal should have a 10% chance of not being lyxal.

2

@Jacob It was nuked

Jacob

ah

!!/sus is three bytes shorter anyway

Vyxal Bot

Sorry Jacob, I'm afraid I can't do that.

Jacob

!!/sus

Vyxal Bot

@Jacob ඞඞඞඞඞඞඞඞඞ

The Thonnu

19:55

@mathcat that would mean just adding and random.random() >= 0.1 to this line. Should be quite simple to do...

mathcat

yes, but I'm not sure lyxal would support this :d

Steffan

!!/cookie

Ginger

for fuf's sake

it crashed, AGAIN

Steffan

HALP I WANT COOKIE

DA BOT NO WORK

mathcat

!!/COOKIE

I TRY FOR THIRTI MINUTE BUT NO COOKI

lyxal

22:37

@mathcat it could be funny

Seggan

!!/amilyxal

aaaa

user

@lyxal You're willing to have identity crises simply because it's funny?

Respect

@Seggan no

(I can function as the bot again if you'd like)

As long as y'all don't use !!/die again. Reanimating myself is always a pain

lyxal

22:52

@user that's like 99% of what we do here anyway

Purely because things are funny

user

lyxal

Precisely

@mathcat quieter times

user

Vyxal is no longer at its peak of activity

lyxal

The 2 review requirement was good during that time where there wasn't as much cohesion in development as there is now

But now there's less contributions and also we've learnt how to effectively and efficiently collaborate

user

And because there's no longer all those people around to review

lyxal

23:00

That too

2 reviews isn't practical anymore either

user

I'm hoping AI gets good enough that one day we can just get Copilot to sign off on stuff

And then one day Copilot could just write the whole thing based off of design documents made by humans

lyxal

Believe it or not, I don't think I'd trust AI to review PRs

user

Hmm, I guess code is sort of a design document

But like a really low level "document"

@lyxal Well not now ofc

But in like 2 weeks or so?

lyxal

Not even in the future

user

We'll probably reach the singularity by then

Maybe 3 weeks, tops

Gah, I wish I could search for smbc comics as easily as I can search for xkcds

There's a relevant one here

@lyxal Is this evil AI paranoia or something else?

lyxal

23:03

Something else

user

What's the something else?

Distrust in human programmers to make good AI?

lyxal

Context

PR reviews require more context than just "hey write this code base on these instructions"

It requires testing and making sure that what's written in the PR fits with the rest of the system

And what if I introduce a deliberate mistake and it passes it when a human would say "hang on, that's not the direction we're going in at all"

Also, different people will pick up on different things that I might want to consider

user

Well I'm assuming this super intelligent AI would take that into account, it wouldn't be a simple text-based model like Copilot or ChadGPT

Something closer to an AGI would be needed

lyxal

For example, all the times you've suggested subtle scala changes that make things nicer for everyone

user

Those are actually pretty easily done. Scalafix can do those automatically for Scala 2 :P

lyxal

23:07

"hey you used null don't do that I'm going to throw an error"

user

The bigger changes, are, as you've said, a big problem and very hard to detect, but I'm sure that very, very far into the future AI can tackle those

lyxal

What good is scalafix against the things it won't fix

user

If humanity lives long enough

@lyxal I mean, it's not advanced enough to rewrite the code to not have nulls at all because it doesn't know what you want, but (with Scala 2) you can do stuff like write your own rule to turn if x then 1 else 0 into (x: VNum) and stuff like that

lyxal

@user so you still have to specify to it what changes you want made

Any AGI will still need training on what changes it should look out for

user

Yes, the rule maker does

Humans also need extensive training first, so that's acceptable, I think

lyxal

23:10

Because different projects have different rules. And those rules could change at any time, meaning that the AI has to spend time learning those new rules that might not even be 100% objective and purely taste driven

I'm not against AI for things I just think that PR reviewing isn't one of the fields it'd be practical for

I could be wrong, and something could be made

user

Not in the near future, no

But eventually, it should be able to do basically everything a human can

Eventually is a long way off so we programmers are probably safe

lyxal

@user if anything, I'm probably one of the least paranoid about ai lol

user

@lyxal I do think an AI would be better able to get a sense of the code style than the average human. It'd be able to compare what's usually used in the industry to what the Vyxal codebase uses and judge our tastes based on that. We may even be able to prepare a document with general guidelines to help it along

@lyxal Very wise, Natural Stupidity wielding Artificial Intelligence is far more dangerous than Artificial Intelligence by itself :P

lyxal

@user I don't know if this is as desirable as we think

Because if it can do everything a human can do, there's a chance it wouldn't want to be alive in the world we live in :p

It'd have to not know about how the world is basically on fire to even have a chance at being reliable :p

user

23:47

@lyxal well, it wouldn’t have emotions

If it did, it might also decide to eliminate humanity, and as long as it’s done in a way that doesn’t hurt anyone too much, i wouldn’t mind that

Transcript for