The Nineteenth Byte

Aiden Chow

00:58

@Bubbler yoooooo nice!

yo wait theres a scratchpad too?

when u add that lol

ok bruh i just see osp join the room but not put any msg, weird

one last thing, can u have an export to png option for the explain thing

i cant seem to upload a svg into an answer

really annoying, im currently just screenshotting the explain thing

also an upgrade to the debugger would be great, but not really a must, just would be nice to have

like being able to go back one step or look at a history of all the steps since the start

and see stack history and stuff like that

oh and i have some ideas for the editor too lolol

like being able to delete a row/column in the middle of the grid instead of only on the sides

or like maybe i can highlight a certain section of the grid and move that entire section somewhere else? or like delete that section? (by delete i mean convert them all to black codels)

that would make editing so much easier lmao

but like the interpreter is already good enough as it is... im just throwing around some ideas :)

Bubbler

01:47

@AidenChow this is much harder, as what you see is actually an svg and I'd need to recreate dashed lines and stuff on a canvas

Aiden Chow

@Bubbler oh :(, ill go on screenshotting then

Bubbler

yeah that's the best for now

Aiden Chow

@Bubbler the scratchpad is a nice touch, im always opening up the notepad to plan out my code beforehand lol

naffetS

02:09

@Bubbler You can just show the svg on the canvas

174

A: Drawing an SVG file on a HTML5 canvas

EDIT: Dec 2019 The Path2D() constructor is supported by all major browsers now, "allowing path objects to be declared on 2D canvas surfaces". EDIT: Nov 2014 You can now use ctx.drawImage to draw HTMLImageElements that have a .svg source in some but not all browsers (75% coverage: Chrome, IE11, ...

Ginger

:O

Bbrk24

02:41

$ node <<<'23.46/21.61'
stdin is not a tty

Thank you node very cool

lyxal

I would expect that node js STDIN wouldn't be a telytypewriter :p

Bbrk24

I guess I'll use py -iq instead

$ py -iq <<<'23.46/21.61'
>>> 1.085608514576585
>>>

I don't like the >>> but at least it actually performs the calculation

(Yes, I know node -p exists, but this is in a long pipeline and I don't want to have to use $(...).)

emanresu A

TIL about prototype pollution attacks. Javascript is fun like that...

Bbrk24

Oh yeah that's certainly a thing

naffetS

03:24

@Bbrk24 Works fine for me, just needs a console.log

lyxal

oh hey my account turned 5 years old today

a minute ago actually :p

Aiden Chow

03:43

@lyxal nice!!

Bbrk24

$ node <<<'1/2'
stdin is not a tty
$ node.exe <<<'1/2'
$ node.exe -p <<<'1/2'
0.5

...huh?

(node -p, without the .exe, doesn't work either)

node and node.exe are literally the same file!

zoomlogo

@lyxal nice

lyxal

41

Q: The Stack Exchange reputation system: What's working? What's not?

Having incentive systems to participate in online communities (like Stack Overflow) is not a new concept. Basically every Q&A site or forum has had some type of points or karma system, and they all work to varying degrees. But as we’ve seen, our system and community in its current state has a hig...

there's suggestions there to cut the question rep gain back to +5

that'd be a big oof

Aiden Chow

zoomlogo, u new guy? i dont recognize ur username

@lyxal riiiipp

Bbrk24

@lyxal Would that be retroactive?

Aiden Chow

03:53

i like the +10 rep from upvotes lol

lyxal

@Bbrk24 some suggestions say yes

Bbrk24

oof

Aiden Chow

bruuh nahhh all our rep would be cut in half if thats the case....

Unrelated String

speak for yourself :P

lyxal

@AidenChow I'd only lose 2.9k rep :p

Unrelated String

03:56

in any case can't this be set site by site
i imagine they wouldn't do it on cgcc

considering the average question takes considerably more effort than the average answer

lyxal

@lyxal and that's a generous upper limit not accounting for repcap

Aiden Chow

i dont wanna lose my 10k privilege lmfao

i wanna see deleted stuff :P

@lyxal ok how is that even possible, where are u even getting ur rep from then

lyxal

@AidenChow the 770 answers I've written lol

Aiden Chow

but like if it go from +10 to +5 all the upvotes on ur answers would be half the value

lyxal

oh no this is question rep only

it used to be +5 for questions

but then it was bumped to +10 in like late 2019 I think

Aiden Chow

03:59

oh

really it used to be +5 for questions??

lyxal

I remember back then thinking "no wait let me ask a few more questions first before y'all bump it" :p

@AidenChow ye

Aiden Chow

damn i thought it was always like that

ok cool good to know that i wont be losing half my rep if this change is applied lol

lyxal

you used to be able to ask more questions in a day without repcapping :p

zoomlogo

@AidenChow well do you not recognise my pfp?

Aiden Chow

@zoomlogo oh i wasnt paying attention to pfp lmfao

how come u change ur username like that? whats it mean?

caird coinheringaahin g

04:39

@AidenChow yeah, was very annoying

also, the change was highly disliked on SO:

-516

Q: Upvotes on questions will now be worth the same as upvotes on answers

Those of you who suffer from banner blindness may not have noticed the announcement: Stack Overflow is changing the reputation scoring system to make the reputation earned from upvotes on questions equal to the reputation earned from upvotes on answers. Previously, upvotes on questions netted yo...

discussion reputation upvotes announcement

New Posts

05:19

0

Q: Language Word Checker

Develop a program to check if a given word is part of a language defined by the following rules: The language has a set of rules. The empty string is considered a valid word in the language. Two words from the language can be concatenated to create a new word in the language. If X is a word in t...

code-golf decision-problem

Adám

Since we made the bakery public, is there a non-public room to which ROs can move messages that should be hidden from the public?

caird coinheringaahin g

You can raise a mod flag, and we can move stuff to the mod-only trashcan room

Or just delete them

In the generally specific sense, no, there isn't a way for the TNB ROs to move messages to a private room

Adám

05:35

@cairdcoinheringaahing Can you nuke (no need to move) chat.stackexchange.com/transcript/message/63149363#63149363?

caird coinheringaahin g

status-complete

Adám

Thanks!

emanresu A

05:54

@lyxal Probably the worst change here (for us) is upvotes requiring 125 rep

Adám

@lyxal Can you see chat.stackexchange.com/messages/63149363/history?

emanresu A

@Adám Nope

Adám

Thanks!

Aiden Chow

well that confirms it

lyxal

07:38

@Adám a little late, but no, I can't

mousetail

08:53

@lyxal Reputation is even more meaningless on this site than on SO. So meaningless in fact that any changes don't even really matter for us

Aiden Chow

@mousetail well i mean its not completely useless. i still want rep privileges :P

mousetail

10K is really easy to get on this site though

Aiden Chow

@mousetail :\ it took me a while on this site to get 10k

mousetail

I still don't have 10K, but it's a matter of time, not of effort

Aiden Chow

@mousetail ig thats true, but like u still gotta spend some effort, cant just wait for the upvotes to come

mousetail

09:03

Just gotta be the first poorly golfed python answer to every trivial question

mousetail

10:51

Adám

11:13

Strange. Searching for "youtube podcasts" on Google gives youtube.com/podcasts as the first result, but that page seems to be missing. Only me?

Ginger

12:23

@Adám status-norepro

lyxal

@Adám status-repro

Adám

Maybe location restricted?

@Ginger Are you in the US, by any chance?

Ginger

@Adám I'm everywhere and nowhere

lyxal

@Adám timezone data suggests yes

Ginger

Nov 18, 2022 at 13:29, by Ginger

Apr 10 at 11:12, by Ginger

Mar 7 at 12:39, by Ginger Industries

yesterday, by Ginger Industries

Feb 3 at 23:32, by lyxal

@GingerIndustries Dang bro don't doxx me like that

lyxal

12:30

although timezone data also suggests vpn usage :p

Ginger

~~huh?~~ yeah sure whatever

@lyxal /srs wdym "suggests vpn usage"?

lyxal

well unless you've recently travelled from utc-5 to utc+0, there's no reason for your github reported time to jump from 7am to 12pm :p

Ginger

my github reported what now

lyxal

There's an extension called Refined Github

Ginger

w a t

lyxal

12:32

it uses commit metadata to determine what time it is for users

a moment ago, it said your time was 7am

Ginger

1. that's kinda creepy and 2. I probably know why that happened :b

lyxal

which a quick map search shows would be in utc-5

but after you created a branch on vyxalbot, it jumped to 12pm

which is in utc+0

Ginger

yeah I know why that is

Bbrk24

13:05

Speaking of timezones, daylight savings time starts soon

lyxal

rip

it ends soon here :p

Ginger

>:(

lyxal

you actually like DST?

Ginger

no, I do not

lyxal

oh

good

Ginger

13:08

I'm angry that I have to deal with it

lyxal

for a moment then you had me worried

Bbrk24

14:24

What's the most painless way to install make on Windows? inb4 "use a VM", but that's honestly a viable suggestion in my case, since I already have an Ubuntu VM set up

mousetail

WSL?

Rydwolf Programs

^

mousetail

It's a lot easier to work with then a VM

Rydwolf Programs

Trying to get a bunch of Linux utilities individually working on Windows is a really unsustainable and weird approach when WSL just works

Ginger

^

mousetail

14:27

make is a super simple tool, I failed to get it working on windows before but it seems to simple. It just runs some commands in bash or CMD yet it's pain

Bbrk24

For context, I'm trying to improve the performance of something that struggles in git bash but not WSL, and the automated tests require make to run. (Manual testing of the component I was working on didn't.)

mousetail

git bash is minGW based I think, might be a issue there

Bbrk24

I was able to improve the performance by 5-30% depending on the run (usually around 10%) but I can't verify that I didn't accidentally break something else

@mousetail About that: I tried copy-pasting some of the commands directly into bash and it didn't work because the tests use ps -o and my version of ps doesn't recognize -o. So even if I install make I still can't use the automated test runner

mousetail

So the issue isn't just make, but a POSTIX compatible shell behind it?

Bbrk24

I've said here before that git bash is missing some strange things, even man

in terms of syntax and shell builtins it's complete, it's just missing random parts of coreutils

mousetail

14:35

Yea MinGW is intentionally minimal

Cygwin is more complete but slower

Rydwolf Programs

LDQ: Should I have a distinction between single and double quoted strings?

For context the lang has backslash escapes, for double quoted strings at least

mousetail

No, unless you need them for a character type

Bbrk24

Well whatever you do, don't do what Swift did, where (like C) strings and characters are different types, but (unlike C) it's impossible to differentiate them by the literal alone. "a" could be a string literal or a character literal and without type info it's impossible to know which. ('a' is a syntax error.)

mousetail

Otherwise it's' confusing

Some kind of multiline string with reduced escaping is nice to have though

Bbrk24

Yeah, like Swift #""" ... """# or C++... I don't remember C++'s syntax but I know it has it

Ginger

14:40

@RydwolfPrograms NOOOOOOOOOO

DO NOT

I do not like languages where they're distinct

mousetail

Also a byte literal string type is very useful

Ginger

if you're going to have a char type have its literal be something like c"!", Python-style

Rydwolf Programs

...ew

mousetail

If you have chars they should defiantly be single quotes for consistency

Ginger

nooooooooooooooo

Rydwolf Programs

14:41

I'm not talking about char literals here tho

mousetail

If you don't have char literals they should be the same IMHO

It gets confusing if you have too many different confusing types of escaping rules

chars and strings being different is fine because they have the same escaping rules

Bbrk24

Having a raw multiline string literal would be a nice feature, but I don't know that I'd anticipate single quotes for that. Python/Swift use triple quotes and JS uses backticks for multiline string literals

mousetail

Maybe you could just accept line breaks in any strings? I don't really know what the downside would be

Rydwolf Programs

It's super ugly

Esp. with indentation in a non-significant-whitespace language

Bbrk24

Swift has a really clever trick for this: whatever indentation is on the line with the """ itself is stripped from the contents of the string

all three of these are the same string:

let foo = "a"

let bar = """
a
"""

let baz = """
          a
          """

mousetail

14:47

That's really nice, wish python had that

Ginger

@mousetail you know what will have it? Rabbit

mousetail

Awesome

Adám

15:15

@Bbrk24 Ooh, I have been looking for a solution for this problem.

Kamila Szewczyk

15:30

Realisation: Golfing languages are essentially manual Lempel-Ziv where the dictionary is built into the interpreter

hence, why not go a step further? One could train a context mixer on a large amount of perl/APL/Jelly code extracted using SE Data Explorer thing, then retrofit the oddball encodings into 8-bit ASCII

then use the resultant CM state, embed it in a "compiler" and "virtual machine", where the compiler simply transforms a readable piece of code by feeding it into the CM and intercepting the arithmetic coder results; the virtual machine would have the exact same CM state so it could with very little effort reconstruct the original program and execute it

there was some golflang ELO thing, does anyone remember where was it?

codegolf.meta.stackexchange.com/questions/8798/… - according to this, Jelly and osable would be the primary targets of this

existing golfing languages are a perfect target for this kind of thing: it is rather difficult to further analyse them to provide better scores by the traditional method of adding a new built-in that is a trivial composition of existing built-ins, because if this would be possible, i would conjecture that the language author would already look into this.

further, standard compression methods wouldn't work on golflangs, but CM models can find further relations in the input data stream (via sparse model, record model, dynamic markov coding, etc...)

golflang programs also tend to follow a rather rigid code structure, you rarely see e.g. literal text in the code hence a context model for it would be unnecessary yielding some savings in this department

Rydwolf Programs

15:47

error: could not compile `...` due to 428 previous errors

Well I'm having fun rn

DLosc

@Bbrk24 Okay, that's a big "wat" moment o_o

@RydwolfPrograms I think not: 1) it doesn't actually help that much with avoiding backslashes because what happens when a string has both " and ' in it, and 2) it's just another way to end up with inconsistent coding styles in the same project :P

Rydwolf Programs

Those sound like arguments for having a distinction

DLosc

Some secondary, more flexible string syntax in addition to your basic double-quoted strings would be useful, though--whether it's Python's """, JavaScript's `, or Perl's q{}, or something else like that

@RydwolfPrograms ... You like inconsistent coding styles?? :P

Rydwolf Programs

No as in

You can't inconsistently use "/' if they have distinct purposes

DLosc

Oh. I may have misunderstood your initial question

Rydwolf Programs

15:57

The options are "they're both identical string types, like in JS" or "single quote strings don't have backslash escapes, and are basically raw strings"

DLosc

I see. So, definitely not the first one. Not sure how I feel about the second one.

Bbrk24

16:09

@DLosc iirc it's because Swift 1 didn't have string interpolation or raw string literals, and they wanted to save single quotes for those. But they ended up going with other things (interpolation still uses "..." and raw string literals use #"..."#), and the community has yet to agree on how to add single-quoted literals. It's going to be even harder because, while Swift 6+ may be source breaking, they may not be ABI breaking -- let me find the forum thread

forums.swift.org/t/…

I have some strong opinions on this topic but you probably don't want to hear them

Sandbox Posts

0

A: Sandbox for Proposed Challenges

Paint you a Picture for Piet Perfection popularity-contest Make a Piet program that does whatever you want—but make it pretty. Your goal is to paint a beautiful picture with an interesting function. Your program may do anything, but your submission is much more likely to gain upvotes if it is int...

Kamila Szewczyk

16:55

a small "study" on how often each Jelly code-page byte is used in Jelly solutions, data gathered over ~5000 programs: pastebin.com/jURUhPsf

the average Jelly solution is 14.2 chars, the median is 10 chars

mousetail

17:46

@RydwolfPrograms Definitely not the second one

Rydwolf Programs

18:18

On the topic of using existing answers in golflangs to generate a more optimal golfing language encoding

I've been working on something that'll take all existing Jelly, Vyxal, Husk, etc. answers, with some filtering to remove ones that aren't actually optimally golfed, and analyze their structure and the operations they perform

And it'll try to create the optimal SBCS and parsing rules

A program in a tacit/stack-based/prefix golflang is just a tree of operators and nilads, with some branches being in multiple places at once, and implicit inputs as a special nilad type. So you could strip out the actual operators/nilads being used, and just find the structures of programs across different golflangs, and try to optimize your language's parsing rules to make those shorter than other structures

(or do something more catstruct-like and arithmetic/huffman code the different structures and store the structure and operator list separately)

And then I'll also write a thing that converts all of the answers into a sort of...golflang assembly, and write an automated thing that looks for common patterns in which built-ins are used where, to create the optimal list of operators and nilads

Kamila Szewczyk

18:37

i have fully implemented my idea, however it seems like it performs way worse than i had expected

the saving is sometimes a byte or two, usually nothing

Rydwolf Programs

A byte or two, if inconsistently, seems like it'd be a pretty big deal

Kamila Szewczyk

maybe, given that most jelly answers are below 10B

but it seems to be the case less often than not

maybe Stax would be a better target, but i doubt that it has as much training data as Jelly

Rydwolf Programs

Vyxal's got the second largest number of answers IIRC

You'd have to filter out flag ones tho which would take out a ton of them

Japt is suprisingly high too IIRC

But same flag issue

Kamila Szewczyk

japt transpiles to JS

the problem is that LZ action in context modelling compressors greatly decreases the ratio, because context is lost

e.g. if you had a sequence of primitives abc for which a separate primitive d was made for, the compressor will no longer recognise the neighbourhood of abc in the same way as the neighbourhood of d

the gains here are mitigated because 1) compression doesn't stack (try compressing a file first with lz4, then with lzma and see for yourself). 2) the context model will take care of eventual repetitions because the training data surely will contain them and ascribe short sequences to longer repetitive patterns optimally while also being aware of the order-N context

how things like Japt should be designed is to take an optimal matchfinder, e.g. github.com/IlyaGrebnov/esa-matchfinder that works via suffix sorting, then push a crapload of golfed JS code through it, then pick out the matches and ascribe single symbols to them

Rydwolf Programs

18:53

Oh I tried doing something sorta like that once I think

Problem was different choices of variable names contributed a ton of inefficiency

Was working on a thing using the Babel parser to parse and reconstruct the JS answers with deterministically chosen variable names but I got bored of it lol

The way I chose the strings that'd be replaced by a single dictionary item was by repeatedly finding the most common pair of adjacent dictionary items, adding it to the dictionary, removing any dictionary items that became redundant, and continuing until adding a pair of items as a single item would be overfitting

I actually made a really simple compression algo using that with huffman coding, which I've used in a few projects of mine to store (really) large string/JSON literals without needing a library

Kamila Szewczyk

some high level parts of my compression code i write in kamilalisp these days due to how convenient for almost all tasks it is

e.g. huffman coding is pretty short and nice: github.com/kspalaiologos/kamilalisp/blob/v0.2/examples/…

Rydwolf Programs

19:09

What if you put the test cases to your challenge in an image that changed to a new randomly generated test case every time you reloaded the page

Too bad SE's font isn't deterministic anymore, you could even completely hide that you're doing that until someone tries to hover/copy it

Altho I guess if you know the platform and browser, you can make a pretty good guess

Ooh, apparently when you copy an image with alt text and paste it, the alt text pastes

So you could even make the dynamic test case copiable...wait no the alt text doesn't get chosen server side nvm

I am determined to abuse this images-hosted-on-other-sites thing somehow :p

I suppose a particularly malicious way you could use it would be finding mods' IPs from chat, then showing one image to mods and another to everyone else (with a once-repeating GIF that changes the image to something harmless shortly after, preventing screenshotting), creating confusion and distrust

Kamila Szewczyk

20:08

regarding the discussion about languages: is there a place (on the meta?) to try introducing a my language? I've posted a few answers in it so far.

Rydwolf Programs

Language announcements aren't on-topic on Meta anymore I don't think (the one for Nibbles got closed)

@lyxal probably knows all the rules regarding language promotion, I'd ask him

New Posts

21:04

0

Q: Detect round trips on the order-4 pentagonal tiling

You're driving along a street in an infinite city whose blocks are pentagons arranged in the order-4 pentagonal tiling. At each step, you proceed to the next intersection and choose to go either left, right, or straight. A sequence of choices that returns you to your initial state (street and dir...

code-golf math geometry graph-theory abstract-algebra

ATaco

21:39

@RydwolfPrograms Have you tried running it with sudo compile -force

Sudo and -force always fixes everything

lyxal

22:10

@KamilaSzewczyk you might be interested in the jelly corpus which also has digraph and trigraph statistics

Bbrk24

23:09

So std::find_if uses linear search, and std::binary_search returns a boolean, not an iterator. Is there no builtin binary search that returns the location of the thing it found?

There's std::equal_range I guess but I feel like I shouldn't

Sandbox Posts

23:29

0

A: Sandbox for Proposed Challenges

Generate a Random Integer code-golfintegernumberrandom Your task is to create a program or function that outputs an integer from \$-\infty\$ to \$+\infty\$ non-deterministically (the output will not always be the same result). Rules Please explain what your code does to confirm that it can produ...

New Posts

23:53

0

Q: Ragged list pattern matching (2)

This is a sequel to Ragged list pattern matching. In this challenge, the wildcard may match a sequence of items of any length instead of just a single item. Given a pattern and a ragged list of integers, your task is to decide whether the pattern matches the ragged list. The pattern is also repre...

code-golf decision-problem ragged-list pattern-matching

Transcript for