The Nineteenth Byte

lyxal

10:11

@Simd how about the python or pari answers?

Simd

@lyxal they are not as good imho

lyxal

Why's that?

Simd

Look at the brain power that went into the C answer :)

the python and pari ones are much easier

also, I guess I should wait until the bounty time is up in one hour

so it's a little early now

lyxal

@Simd what's to say brain power didn't go into the other answers?

UndoneStudios

I guess brain power is subjective

Simd

10:16

@lyxal they use the standard libraries

brains are definitely subjective

mathcat

~~except mine~~

Simd

:)

UndoneStudios

@Simd That's true

The python answers used outside libraries

give a bounty to that which does not

so not the python answers

Simd

it's really C or Julia for me

I think it will be C

I am rewarding hard work and brains

also the fact that they made 128 bit look trivial in C

might have actually taught people something too!

UndoneStudios

But C isn't the most golfed

that's my concern :(

Simd

10:21

it is a per language competition don't forget

I feel everyone forgets that all the time

UndoneStudios

Oh

But in that case then C is the only competitor in its own lang

Simd

yes

that is why typically there is no accepted answer

UndoneStudios

yeah

but anyways, who are you giving the bounty to?

lyxal

@Simd there's a third answer you haven't considered

Simd

@lyxal vyxal?

UndoneStudios

10:31

vyxal?

lyxal

That's right

It had hella brain power put into it

Simd

it doesn't satisfy the criteria of the question

UndoneStudios

ninja'd

lyxal

@Simd it does

UndoneStudios

@Simd What was the criteria to get the bounty?

Simd

10:33

@lyxal "Your code doesn't need to be fast, but ideally it should complete in under a minute."

@UndoneStudios that is entirely up to the bounty giver

lyxal

@Simd ideally

UndoneStudios

@Simd "Ideally"

lyxal

That's just a suggestion

UndoneStudios

ninja'd again

Simd

if you want the bounty :)

lyxal

10:34

So basically only compiled languages ever stand a chance

That's kinda cringe

UndoneStudios

Huh get a better language lyxal :)

Simd

@lyxal that's wrong. python would work fine

and so would basic

in fact almost all interpreted languages

lyxal

@Simd if using standard libraries

Simd

but you just have to implemennt it right

lyxal

And besides, it's only slow because it's going to a high precision

If it was a lower precision it'd be faster

Simd

10:36

@lyxal no that is not right. If you translate the julia code directly into python it would be plenty fast enough

you would use decimal of course

@lyxal there is no problem with speed if you just use the loops set out in the C and Julia answers

in fact I can't think of an interpreted language where that would be too slow

UndoneStudios

Hold on a minute. Is the criteria fastest-code?

Simd

@UndoneStudios no

UndoneStudios

For the bounty

Simd

you can look a few thousands times in pretty much any language

look at the loop.

for(X a=*y=k=1,q;k<200;){
/* estimate the Riemann zeta function */
for(i=99,q=0;--i;)
q+=(i%2?k:-k)*pow(i,~k);
*y-=(ldexp(a*=logq(x)/k++,-k)-a)/q
}

in which language would that take more than a minute?

UndoneStudios

"Estimate"? Does that mean the zeta function is not the zeta function?

@Simd depends

Simd

10:39

it is doing less than 2000 iterations in total

lyxal

Then that's not as accurate as my answer

Mine takes as many iterations as it needs to get to a value that doesn't change

UndoneStudios

in that case I think the bounty should go to vyxal

accuracy is more important in this case

lyxal

and to 256 deicmal places too, twice the required accuracy

That's why it takes so long

UndoneStudios

@lyxal The bounty has to go here

Simd

@lyxal the needed accuracy was specified in the question

if you could implement those loops in vyxal you would have won

UndoneStudios

10:43

Is there a penalty if you go beyond that?

lyxal

@Simd I've still got an hour

Simd

@lyxal true!!

lyxal

And if 2k loops is all that's needed, it'll be shorter

Simd

@lyxal just check you get the right answers

mathcat

lyxal speedrunning vyxal

pro gemer

UndoneStudios

10:46

hah

Simd

@UndoneStudios sorry, beyond what?

UndoneStudios

@Simd the precision needed

Simd

@UndoneStudios no penalty from me. It just has to finish in time

UndoneStudios

@Simd "ideally"

Simd

@UndoneStudios if you want the bounty :)

UndoneStudios

10:48

@Simd ah ok

Simd

everyone is free to do what they want if they don't want my approval :)

it's a free world

lyxal

11:02

well turns out it won't be the Julia answer

it gives 4.564496573408498235184002179924633035696951973113798875845920480384681651501851 for 10^1

when 4.56458314100509023986577469558 is required

UndoneStudios

Hooray for C!

lyxal

and it's not C either

4.56631846465729586315692453417579 for 10^1

UndoneStudios

oh

lyxal

so @Simd who gets the bounty now

when the two options aren't valid to the required accuracy

mathcat

lol

lyxal

11:05

sure they finish fast, but at the cost of being valid for only 2 decimal places

UndoneStudios

I pray it is anything but Vyxal

mathcat

why though?

Simd

@lyxal oh the julia one looks wrong too now!

UndoneStudios

Mainly because chat.stackexchange.com/transcript/240?m=62465060#62465060

Nov 30, 2022 at 4:04, by lyxal

and given that Jelly is less practical than Vyxal, I also question your practicality rating

Which we've disproved

Ever since then I never trusted Vyxal

Simd

just checking if any of them are right!

lyxal

11:08

@Simd it's not just 10^1

it's 10^all the powers

Simd

the python ones are right

lyxal

but it uses standard libraries though

I thought that wasn't "brain power"

UndoneStudios

Yeah

Guess nobody gets the bounty?

Simd

@lyxal right it would have been beaten by a correct cleverer answer

UndoneStudios

Mathematica uses a builtin

lyxal

11:09

then award it to PARI/GP

Simd

not the new vyxal answer?

can you see if you can get the right answer by increasing the number of iterations?

lyxal

we've shown that all estimate based answers that are fast aren't valid

@Simd not without timing out

Simd

the number of iterations determines the accuracy

how long does it take for 99 iterations in the inner loop in vyxal?

lyxal

@Simd I think the problem is that the Gram sequence just isn't accurate enough

Simd

@lyxal try increasing the number of loops

lyxal

11:12

no I mean fundamentally

Simd

all the methods are using that

including the ones giving the right answer

e.g. codegolf.stackexchange.com/a/256130/116117

UndoneStudios

Yeah kind of explains why I do not like making trade-offs

Simd

@UndoneStudios we know you don't like real numbers on computers :)

UndoneStudios

@Simd Give the bounty to that

It's a good answer

Simd

@UndoneStudios I may well do

but I am holding out for a vyxal answer :)

lyxal

11:14

@Simd those are using sum methods that aren't estimations

they aren't iteration based

they give exact results from doing things like clever algebra under the hood

UndoneStudios

In 47 minutes !?

@lyxal ~~then use it~~

Simd

@lyxal no they don't do any clever algebra

lyxal

internally they do

Simd

the sum in pari just adds until the term it is adding goes below a threshold which you can set

lyxal

suminf looks to not be iteration based

Simd

11:15

it's just a for loop

suminf is just a for loop

lyxal

while loop

it's a while loop

> The evaluation stops when the relative error of the expression is less than the default precision for 3 consecutive evaluations.

source

Simd

or a for loop with a break :)

but yes

lyxal

point is, there's no magic number of iterations

Simd

so it's easy to implement in any language

nothing clever about it

UndoneStudios

I think Simd isn't getting lyxal

Simd

11:18

sure but you just run it once, see the number of iterations and then use that number

lyxal

it might change for each input

Simd

you just do a preliminary experiment to see what it should be

mathcat

@UndoneStudios how tf is jelly/golfscript more practical than apl lol

APL isn't even meant to be estoric

Simd

@lyxal if you test it for 10^1, 10^31 then largest will be the right answer

UndoneStudios

@mathcat That's such an old message

lyxal

11:19

with less than 40 minutes to get an answer, I don't think I have the mental effort left to do so

UndoneStudios

read the conversation

Simd

@lyxal :(

lyxal

it'll just have to go to one of the python library answers

Simd

or pari...

lyxal

@Simd I mean, technically that's what my current answer is doing, it just takes a while

UndoneStudios

11:20

PARI/GP

Simd

@lyxal is it actually running the same loops?

lyxal

it's going until there's no change in value

mathcat

@UndoneStudios apl wasn't mentioned at all

Simd

@lyxal just change to set a threshold

that is a one minute fix

UndoneStudios

@mathcat What I meant was that the conclusion was the graph itself was too subjective

The graph is my way of seeing the world

lyxal

11:26

@Simd still times out

I'm gonna give up on it. I've spent way too much time on this now

for 50 rep, not worth it :p

Simd

k

thanks though

Simd

12:03

pari won the day

I can sort a list of pairs by second element with sorted(
[('abc', 121), ('abc', 231), ('abc', 148), ('abc', 221)],
key=lambda x: x[1]
)

but what if the first elements differ too and I want to sort by them second?

pxeger

12:21

@Bubbler If you need autocompletion to write productively, then you're not writing a terse enough language

Simd

14:37

hello all!

New Posts

0

Q: Write shortest .class file in java which is equivalent to this program

This is a program: import java.util.*; public class testtekovanje{ public static void main(String args[]){ Scanner sc = new Scanner(System.in); int b = sc.nextInt(); // 2 int c = sc.nextInt(); // 3 int d = 2; // 4 int e = 0; // 5 while(b>1){ ...

code-golf

Ginger

VTC

pxeger

no?

Simd

14:51

I have lists of grades like [('MATHS', 'A'), ('PHYS', 'A*'), ('ECON', 'B'), ('PHYS', 'B')] in python. I want to only have the highest grade per subject. Is there a golfy way to do that?

A* is higher than A

CMQ ^^

mousetail

A*? Not A+

Simd

yes, it's an odd system

it just means the letter before A :)

caird coinheringaahing

15:21

@Simd it's the UK a-level grade system :P

Simd

@cairdcoinheringaahing :)

Radvylf Programs

@pxeger or you've got stupidly long variable/function names

There's things in between w and wordsInInputWhichContainOneOrMoreSpaces

PyGamer0

15:39

@Bubbler i am trying to program in aarch64 again, can i improve this code? thanks

DLosc

16:38

Imagine a "language" that consists of an accumulator and two commands: + increases the accumulator value by 1, and * multiplies the accumulator value by 2. The accumulator starts at 0. CMC: Given a nonnegative integer, output the shortest program in this language to set the accumulator to that value.

Radvylf Programs

x=>x.toString(2).replace(/0/g,"*").replace(/1/g,"*+").replace(/^\*/g, "")

Actually at this point "lol its just binary" is longer than doing it properly...

DLosc

@RadvylfPrograms Can you just do .slice(1), since I think the intermediate result will always start with a single asterisk?

Radvylf Programs

Oh true

x=>x.toString(2).replace(/./g,x=>+x?"*+":"*").slice(1)

f=x=>x%2?f(x-1)+"+":x?f(x/2)+"*":"" is way shorter :/

DLosc

@RadvylfPrograms Ah, good thinking with the ternary. I was trying to combine the two cases using string slicing, which is probably the shortest way to do it in Python

Radvylf Programs

Yeah, I tried that too

DLosc

16:58

@RadvylfPrograms Ungolfed port of this solution into Elm:

f x =
  if modBy 2 x == 1 then
    f (x-1) ++ "+"
  else
    if x == 0 then
      ""
    else
      f (x//2) ++ "*"

Now let's see how much I can golf it.

70 bytes: f x=if modBy 2 x>0 then f(x-1)++"+"else if x<1 then""else f(x//2)++"*"

Unfortunately, it looks like case requires newlines and indentation, which makes it a bit longer.

f x=case(x,modBy 2 x)of
 (0,0)->""
 (_,0)->f(x//2)++"*"
 _->f(x-1)++"+"

user41805

@RadvylfPrograms 27b

@DLosc sad code golf will be done better by computers than humans at some point. what will we do then

DLosc

17:13

@user41805 I mean, people still play chess even though computers are better at it

user41805

fair

DLosc

Also reminds me of Ken Jennings's TEDx talk about whether it's still useful to know stuff when Google is a thing.

user41805

17:46

@DLosc thanks, gave it a watch

mousetail

18:09

@RadvylfPrograms when are you planning to post your KOTH?

New Posts

0

Q: Division between two strings

Inverse function of this challenge To (properly) multiply an string by an integer, you split the string into characters, repeat each character a number of times equal to the integer, and then stick the characters back together. If the integer is negative, we use its absolute value in the first s...

code-golf string number arithmetic

Sandbox Posts

18:26

0

A: Sandbox for Proposed Challenges

Write a proper quine. Community decided that you're not allowed to claim you're not reading source code but data, and read a stored string.

Ginger

lmao at this line from who's help:

> If ARG1 ARG2 given, -m presumed: 'am i' or 'mom likes' are usual.

(-m is a flag that makes who only list the user running it)

pxeger

$ who are you
pxeger

impostor!

Ginger

süs

Radvylf Programs

18:54

Considering changing my name again, the a in Radvylf makes it kind of an ugly color still

And it seems like it should be pronounced "rad-vilf"

"Rydwolf"

It looks so much nicer

@mousetail Some time in the next few weeks

mathcat

Hey, almost completely destroyed my Raspberry Pi today with sudo apt-get purge python3, nothing to be concerned about

It looks like purge deletes python and its dependencies, almost everything on my Pi was dependent on python

user

19:12

...why would you remove Python?

Seggan

me

Radvylf Programs

> why

Python's a dependency for lots of stuff

user

Were you trying to upgrade it?

Seggan

@RadvylfPrograms i read who

Radvylf Programs

I mean it makes sense that way too, you're why we should remove Python :p

If we don't, you'll make us regret it :p

user

19:15

Or a less dark interpretation: We don't need Python anymore because Seggan will do the job themselves

AI will no longer be taking our jobs, Seggan will

mathcat

@user yup lol

Radvylf Programs

19:28

@user If Seggan took over all of those customer service AI chat popups on websites we'd be in a much better world

The hivemind is already two-strong, we just need exponential growth to kick in

Seggan can just be Larry from AWoG

Ginger

there's only room for one distributed artificial intelligence on this planet, and I've already got that spot filled

Radvylf Programs

Idea: A programming language where import trains a transformer AI on your code and the library's code, and has it write out a new program inspired by both

Ginger

19:44

me: import tensorflow
language: oh god oh frick oh god oh frick

@RadvylfPrograms corollary: importing multiple modules on the same line mixes them together

Seggan

@RadvylfPrograms oh i will

Ginger

@Seggan you're a hivemind right? q:

Seggan

ignore that then :P

caird coinheringaahing

22:26

I'm considering accepting my answer to this, given the votes, but would like any more feedback/opinions

lyxal

22:39

It seems fine, but I guess in future, there needs to be emphasis on avoiding situations where a problem would possibly arise

like creating a url shortened link to a site is obviously going to lead to some issues down the line, especially if it's registered just by a code golfer

feel like that should be a standard loophole

caird coinheringaahing

What's weird is that using URL shorteners is a loophole

lyxal

23:01

Url shortners are, registering custom domains for challenges aren't

It's basically metaurlscript

caird coinheringaahing

Feel free to suggest it as a loophole

lyxal

turns out it kind of already is

42

A: Loopholes that are forbidden by default

URL shorteners / shortened URLs Every once in a while, we get a challenge that requires fetching some data from the internet. While some of them manage to ban URL shorteners in time, other don't. I propose to forbid them by default because of the following reasons. URL shorteners that were cre...

> I propose counting all URLs on domains as shortened (and therefore forbidden) if the domain name isn't owned by the same individual/company as the canonical one. Thus, e.g., if the task at hand involves querying api.stackexchange.com, only URLs that belong to Stack Exchange, Inc. may be used in the challenge.

and the ppcg.ga link was used after that loophole was proposed

so now not only are the answers invalid due to external resources, they're invalid due to standard loopholes

which means that even if it's decided to mark everything as invalid because of the external resource issue, they still need to be handled as invalid due to loopholes

~~new meta question time?~~

caird coinheringaahing

23:31

I'm not entirely convinced that loophole covers the linked question

And, even if it does, I'm also not convinced that it should be a standard loophole to use "non-officially owned" URLs

lyxal

@cairdcoinheringaahing why not? The task of the challenge involves opening codegolf.stackexchange.com which belongs to Stack Exchange, Inc. Therefore, only links owned by Stack Exchange, Inc are allowable

New Posts

0

Q: n-digit primes given the first m digits

I just discovered this site and this is my first question, I hope I'm doing it correctly. The challenge is the following, you must write a code that prints all the prime numbers of \$n\$ digits. Since this problem is computationally complicated, I think that one way to simplify it is to give a nu...

code-golf primes

lyxal

@cairdcoinheringaahing you may not be, but the score is more than 5+ with a distribution of +53/-11

meaning it's recognised as a loophole

caird coinheringaahing

I will also just bluntly say: I'm not happy deleting 46/50 answers from a post in the first place, especially given that, at the time, all the answers were accepted - and flags on those posts were handled - as valid

If the numbers were lower, I'd be more open to doing so. But deleting literally 92% of 50 answers is a massive amount

lyxal

so if a lot of answers to a challenge are edited to violate a loophole and no one notices for a long time that's allowable?

caird coinheringaahing

23:41

I'm not saying that

lyxal

well that's what happened

caird coinheringaahing

I'm saying that I want to avoid having to go through literally two pages of answers deleting every single one of them, and would like to find a resolution that allows me to avoid that

If the site wants them all deleted? Fine, I'll spend an afternoon leaving comments and mass-deleting them. But, I'd rather avoid that if possible

Additionally, that loophole was explicitly allowed:

@issacg Yes, it is allowed. — noɥʇʎԀʎzɐɹƆ Mar 30, 2016 at 15:26

lyxal

hm

fair enough

CMQ: Are infinite lists/generators allowed to be returned for challenges where the output would otherwise be a finite list but you know that logically, the infinite list will end up being equal to the finite list?

More specifically, does anyone know of any existing meta posts explicitly allowing/disallowing this

I know this exists, but that doesn't really cover the case of what I'm asking. It's just saying lazily evaluated lists are allowed. It doesn't state anything about infinite being guaranteed the same as finite

Transcript for