The Nineteenth Byte

Ginger

5:00 PM

wat

Radvylf Programs

The key isn't in using fewer operators, it's in using the same number of operators more effectively

Ginger

I am not an information theorist; I don't know how I would do that

Radvylf Programs

You don't need to understand information theory, you need to understand golflangs

mathcat

@RadvylfPrograms excellent metaphor

Ginger

except I do not understand golflangs

mousetail

5:01 PM

The key is making common operators shorter and uncommon operators longer

Ginger

hm

mousetail

that way you can support a potentially infinite number of operators

without loss of density

Ginger

~~time to write a program that checks every single answer to find the most common operators~~

Radvylf Programs

Look at four bit golfing languages. Despite having operators half the size of SBCS langs, they're often multiple times longer. That's because you get 1/16th as many operators for just 1/2 the size, and 256 operators you know are useful is a lot better than 256 pairs of two operators, many of which will perform niche or redundant operations

thejonymyster

have we explored "if your program is syntactically invalid, it is reinterpreted as a structurally different program" as a golflang mechanism

Radvylf Programs

5:02 PM

@mousetail a key

mousetail

A very OP key

Radvylf Programs

Neither catstruct nor pxeger's idea involve variable width encoding, that's only one approach

mousetail

There is no way you can create 256 operators with exactly equal number of use cases

it's always going to be inefficient

Radvylf Programs

@Ginger We've already done that for years :p

@mousetail Sure, but making one operator shorter makes the others longer

mousetail

Yes, the trick is to make common operators shorter and uncommon ones longer

Radvylf Programs

5:04 PM

Having addition be shorter is nice, but if it comes at the cost of 20 other less common operators being longer, it maybe isn't a good idea

mousetail

It is

Radvylf Programs

it depends on the numbers

mousetail

even if a program uses uncommon operators it's likely to use it just a few times, while every program will use common operators a lot

Radvylf Programs

If each of those 20 are used half as often as addition, suddenly it's definitely not a good choice

@mousetail That doesn't really occur in practice though, the slope is rather shallow

And most of the more commonly used operators can actually be eliminated with more clever tricks than shrinking them with encodings

mousetail

I literally proved a average -5 byte reducion on lyxal answers

@RadvylfPrograms True, it's better to remove operators all together and make them implicit

Radvylf Programs

5:06 PM

@mousetail That seems slightly suspect to me, given that the average Vyxal answer is probably not much longer than 5 bytes

mathcat

@RadvylfPrograms yeah, encoding is just a side-effect of golflang-evolution

Radvylf Programs

Although it definitely would not surprise me if huffman/arithmetic coding would be a significant byte save for most 3rd-gen golflangs

mousetail

Ok it was skewed by a few very long answers that got reduces a ton

average percentage is 2%

Radvylf Programs

I think it's mostly just been avoided as a low-hanging and uninteresting golf

mousetail

I mean SBCS is a low hanging uninteresting golf

yet everyone does it

Radvylf Programs

5:08 PM

SBCS is sort of the baseline though

mousetail

Exactly

We're setting a new baseline

Radvylf Programs

And I'd consider it a lot more interesting to have 160 more possible operators

@mousetail It's not a sensible baseline though. Unlike SBCSs, which were already in use before CGCC, and which make sense when bytes are the scoring method.

mousetail

Yes, that's the point. If you don't limit yourself to having every operator the same length you can have 160 or 200 or 1000 operators without being too long

Radvylf Programs

Well that's slightly different from just huffman coding an existing language

(Personally I think 256 is a pretty good sweet spot for interestingness of the number of operators though; less than 128 is too few to allow for creativity, and more than 512 is too many to make designing languages fun and picking which operators make the cut difficult)

Ginger

ok here's another sample: p{?=a"cat""cat goes meow"a|i

this language uses the functional paradigm

Seggan

5:13 PM

@RadvylfPrograms IMO fig somehow stays optimal below sbcs

Radvylf Programs

I was talking about interestingness, not golfiness there

Seggan

ah

i was thinking of huffman coding fig, but i got tired of manually building a tree

Radvylf Programs

But 96 operators would indeed be less efficient than 256 operators in theory. I expect that being able to put more focus on each operator allowed you to optimize it better though.

Seggan

yep

i think more datatypes helps too (more overloads)

CMP: is a dedicated boolean datatype useful?

Ginger

here's a different one for codegolf.stackexchange.com/questions/241080/is-it-a-pure-word: p!%smi{fÁa}2

thejonymyster

5:22 PM

@Seggan i think maybe definitely not, truthy and falsy work fine :-)

however that has me thinking of the "not" operator, usually that implicitly returns whatever is being used as the standin for bools or something

wait, nvm the thing i was starting to say

Adám

@Seggan No.

Radvylf Programs

@Seggan Yes

Seggan

why

clarification: i meant boolean on top of truthy/falsey

Radvylf Programs

Only minimally so, but it does not conflict with other data types (unlike, e.g., different number types), meaning that it can only possibly save you bytes

Adám

K/Q/kbd+ makes millions or billions without a Boolean datatype.

Ginger

5:24 PM

@Ginger Here's a pseudocode version of that: print(not(mod(sum(map(input() a => find(alphabet a))) 2)))

Adám

@Seggan I don't follow.

Seggan

truthy/falsey idea + a boolean datatype

kinda like c(?)

Radvylf Programs

E.g., let's say your not operator returns a boolean. You now have ~48 operator overloads to do things with that boolean, like really obscure control flow, at exactly zero cost.

Seggan

@RadvylfPrograms 20067

Ginger

remember, you can't spell "functional programming" without "malfunctioning program"

or, for that matter, "planigram uncomforting"

mousetail

5:37 PM

@RadvylfPrograms Coding a existing language was just a demonstration, to do it properly would require a new language built that way from the start

Radvylf Programs

CMQ: If signed zero is a thing, should the sign of +0 be 0 or 1? Is it positive?

hyper-neutrino

the sign of +0 should be +0 and for -0 it should be -0 (partially joking)

Radvylf Programs

That's how JS does it

Rust considers 0.0 to be positive, but not 0

mousetail

I think that makes the most sense

Radvylf Programs

To me this is the worst way to do it

mousetail

5:47 PM

Maybe not have a negative 0 but have a omega value instead that is distinct

thejonymyster

mr incredible slamming finger into book 0 is 0

mousetail

exactly, that's why having a seperate value to represent a value infinitesimally close to 0 makes sense

thejonymyster

epsilon <3_<3

Radvylf Programs

Something that I think is really cool: When I type, my fingers are basically really well trained markov chains

I keep typing "Minux" instead of "Minux"

thejonymyster

needs more training

Radvylf Programs

5:51 PM

Because my fingers are trained to expect "(L)inux"

Any time I do Shift+? i n u

mousetail

I know a guy who has modded his keyboard so he can type a lot of common words with 2 keypresses

Seggan

to me shift is {

thejonymyster

@mousetail i can do that too: if or an im we no oh

Ginger

@mousetail idea: a keyboard where every key is a common set of three letters and when you press a sequence of them it types the shortest word that includes all of them in sequence

mousetail

@Ginger That's literally the keyboards court transcribers use

Ginger

6:04 PM

example: int + nat + ion + Compose = internationalization

thejonymyster

that owns. we should have a court transcription challenge

mousetail

Go for it

Ginger

lemme go write that up rq

~~dibs~~

mousetail

They also do odd stuff like merging similar sounding letters and ommiting vowels when not nececairy

so it's a "lossy" compression

thejonymyster

i dont wanna write it :P doh

ive got too many irons in the sandbox (wait thats a mixed metaphor) ive got too many castles in the sand

mousetail

6:06 PM

I want to see the NERPS challenge

Ginger

hmm, this might actually already exist

because the challenge is just "Given a list composed of sequences of three letters, as well as another list of dictionary words, return the shortest word containing all of the sequences in order"

can someone check if this exists or not? I don wanna

thejonymyster

@mousetail god that took me way too long to remember what the acronym was

Seggan

@Ginger i think thats called a stenograph?

Stenotype

‹See TfM›A steno machine, stenotype machine, shorthand machine, stenograph or steno writer is a specialized chorded keyboard or typewriter used by stenographers for shorthand use. In order to pass the United States Registered Professional Reporter test, a trained court reporter or closed captioner must write speeds of approximately 180, 200, and 225 words per minute (wpm) at very high accuracy in the categories of literary, jury charge, and testimony, respectively. Some stenographers can reach 300 words per minute. The website of the California Official Court Reporters Association (COCRA) gives...

mousetail

It's basically code golf but for spoken language

Ginger

that's SMS-speak lol

mousetail

6:10 PM

It looks a lot like SMS speak

Though SMS is different, since a lot of acronyms represent emotions. You can't objectively judge emotions in a court

mathcat

well just use AI

mousetail

It's not really useful information either

though I'd think it would be hilarious if there was a AI announcer occasionally going like

"The plaintiff is bored"

"The defendant is angry"

"The judge is tired and wants to go home"

thejonymyster

the stenographer is tired and wants to stop typ

mathcat

Imagine a next-level AI could recognize "guilty-ness"

thejonymyster

reminds me of minority report lol

Ginger

6:17 PM

@RadvylfPrograms would you happen to know if this exists?

mousetail

@mathcat Yes and probably be incredibly biassed

Ginger

or anyone else for that matter?

it's simple enough that I think it does but if it doesn't, I'm going to post it

@thejonymyster a stegenogenotegnoeggenegogglogeographer is a person that transcribes encoded musical charts being etched into the surface of the earth while drinking eggnog and wearing swim goggles

5

thejonymyster

holy McMoly. thats my dream job sincei was 9

Ginger

yea me too :b

Codidact Posts

New Posts - Challenges - Code Golf: Balanced quinary quasiquine

posted on October 06, 2022 by trichoplax‭

Give an integer $N$ in balanced quinary, output the first $N$ characters of your source code if $N$ is positive, or the last $-N$ characters of your source code if $N$ is negative. Term...

Radvylf Programs

6:34 PM

@mathcat My guess is it would almost immedaitely become racist

emanresu A

^

Radvylf Programs

It's hard to pick up on subtle emotional cues, and those are unreliably anyway. Guessing someone's socioeconomic status correctly, however, could be a pretty good predictor of whether they're "guilty" (keep in mind you'd have to train it with already-biased real world data), and unfortunately due to a variety of factors race and socioeconomic status are closely related

emanresu A

AI magnifies human bias

mousetail

There are already AIs used to "judge risk" but they have been shown to have a huge racial bias

Ginger

0

A: Sandbox for Proposed Challenges

Make a Court Transcriber Discussion in courtrooms often occurs at a high speed and needs to be accurately transcribed in case the text of the decision is needed din future cases. For this purpose, the stenographers who transcribe court discussions uses specially designed keyboards with a small se...

Seggan

6:43 PM

@Ginger needed din

Ginger

whoops d:

fixed it

mousetail

Will all words have 11 letters?

Radvylf Programs

My career goal: become a steganostenographer: someone who keeps track of what's said in court rooms, but disguises it as something else

3

thejonymyster

@Ginger can we assume there wont be multiple possible matches?

mousetail

My career goal: Become a stenogangster who illegally breaks into courts to record what is said

6

thejonymyster

6:46 PM

@RadvylfPrograms i knew i was getting those two confused lol

Ginger

@thejonymyster Nope, you must return the shortest match

@mousetail Nope

thejonymyster

@Ginger sorry, multiple *shortest matches

as in [abc def ghi], abcdefghia abcdefghib

Ginger

@thejonymyster If there are multiple shortest matches you may return any of them

thejonymyster

ok cool (^_^)b

@Ginger will the dictionary contain words shorter than 3 letters long?

idk if that matters but it crossed my mind :P

Ginger

does it matter?

thejonymyster

6:52 PM

LOL yeah i guess not

Ginger

I don't think it does

@thejonymyster dancing-emoji.gingerindustries.repl.co/…

3

I saw the emoji and remembered I made that

thejonymyster

so happy

Ginger

~~if this message gets 10 stars I'll reprogram it and add more features and an actual stylesheet~~

thejonymyster

tempting :P

Ginger

@thejonymyster you or the emoji? d:

thejonymyster

6:55 PM

the former as a result of the latter :-)

Ginger

yay :D

thejonymyster

@Ginger CMC: this

Ginger

@thejonymyster VTC as dupe of

Jan 6 at 19:43, by pxeger

CMC: animate alternating between

Jan 6 at 19:43, by pxeger

⌐(■_■)⌐
¬(■_■)¬
⌐(■_■)⌐
¬(■_■)¬

thejonymyster

lmao tragic

Sandbox Posts

0

A: Sandbox for Proposed Challenges

A decimal-based unit of time code-golf math Background In 1960, the 11th General Conference on Weights and Measures defined the Système International d'Unités (SI) Units which scientists still use today. The metre and the kilogram became standard units in that conference. These were based on powe...

2

A: Sandbox for Proposed Challenges

Make a Court Transcriber Discussion in courtrooms often occurs at a high speed and needs to be accurately transcribed in case the text of the decision is needed in future cases. For this purpose, the stenographers who transcribe court discussions uses specially designed keyboards with a small set...

Seggan

7:05 PM

yay my transpiler is actually taking shape now

Unknown reference error at row 12, column 0, statement 'print(fibonacci(10))': print was not defined

Ginger

transpiler for what?

Seggan

my language, Rol

a clike for Lua

Ginger

another one? :b

Seggan

no

this is the only one i had

Ginger

github.com/Seggan/Fig

Seggan

7:10 PM

ah yes, another one :P

Sep 19 at 15:38, by Seggan

A short showoff of my language: https://pastebin.com/HEBj4XvR

Ginger

@Seggan can you Gistify that? I can't see it

Seggan

@Ginger gist.github.com/Seggan/48620562010ca82f37cfa38904125d68

Ginger

thanks

Seggan

its a touch different from the pastebin bc of design changes i made

this is a praclang, if you didnt get it

Ginger

I did

thejonymyster

7:59 PM

Any feedback?

Might post it later if nobody finds a glarin errorin :-)

Seggan

@thejonymyster maybe a worked out example will be nice

thejonymyster

Oh right lol

ill write something up when i get home thanks

Ginger

okay, replacement NPSP is now running in the Sandbox on my server instead of replit

nvm I lied

Sandbox Posts

8:18 PM

0

A: Sandbox for Proposed Challenges

Partition square into squares Given integers \$n\$, \$m\$ and \$k\$, randomly output a list of \$k\$ numbers between \$1\$ and \$m\$ (inclusive) such that the sum of the squares of the numbers in the list is equal to \$n\$ squared. In other words, find a way to randomly split \$n^{2}\$ into \$k\$...

Seggan

hrm i may need to work a bit on my error messages :P

Type mismatch error at row 1, column 4, statement '"a"': Expected String, got Num

turns out i accidentally switched the expected and actual type arguments in the error generating function

much better: Type mismatch error at row 1, column 4, statement '"a"': Expected Num, got String

why is writing a type checker so much easier than i thought it would be

Ginger

8:34 PM

okay NOW it works d:

(I hope, please don't fail me)

okay so here are the profiles for Unit Tester and Twonit Tester respectively:

I am very funny [citation needed]

@RadvylfPrograms ok, I have confirmed that GBNPSP works

now we just need it to remain stable for a week

DLosc

9:30 PM

@Ginger That one does use non-ASCII characters, tho, which poses extra difficulties for some languages.

DLosc

10:17 PM

QBasic, 49 bytes:

CLS
?"d (^_^)b
SLEEP 1
CLS
?"d(^_^) b
SLEEP 1
RUN

Sadly, a more clever approach comes in at 53 bytes:

emanresu A

lol

DLosc

DO
CLS
?"d"MID$(" (^_^) ",i+1,6)"b
SLEEP 1
i=1-i
LOOP

lyxal

Well that's what you get for having mid code

DLosc

It'd be 51 if QBasic used 0-indexing

DLosc

10:41 PM

feature-request for languages with /* ... */ comments: ?8 ... 8? should be an alternate comment syntax :P

3

DialFrost

10:54 PM

@cairdcoinheringaahing luv the new pfp! :3

caird coinheringaahing

11:23 PM

@DialFrost thank Redwolf for the edit :P

Transcript for