The Nineteenth Byte

I built a compiler for a grammar that had maybe five rules for a class project four years ago. I remember decades of English class better than I remember that. So that's all I can help you with.

Ismael Miguel

5:49 PM

If I made an interpreter, would you use this language?

Zgarb

@IsmaelMiguel You have two distinct ways of explicitly producing a quine. Wouldn't one be enough?

Martin Büttner

(the right answer is, zero would be enough :P)

Rainbolt

@IsmaelMiguel I enjoy golfing in Python and Java, just to beat other Python and Java submissions. Golfing in a "look it up" language doesn't sound appealing.

Ismael Miguel

@Zgarb Not really. I had 3, but deleted one. One simply outputs the code and does nothing, the other fills the stack with the source code. Those are 2 very distinct actions

@Rainbolt Thanks dude. How could I made this language in a way that would at least wake some curiosity to try it?

TheBestOne

I just realized that drinking/eating all potions/rations before attacking at all does not give any additional health/energy :(

Geobits

5:52 PM

I don't see how the quine bits are/would be good for golfing (at least here). Any decent challenge related to quines would have a "no builtin quine functions" line, right?

Martin Büttner

@Geobits it usually doesn't, I think, because so far only HQ9+ can do that, using which is a standard loophole anyway

Rainbolt

@IsmaelMiguel I don't know. The last language that really sparked some interest here was Marbelous. Maybe you can take some pointers from its creators.

Martin Büttner

I just think it's generally a bit lame

takes the whole fun out of quining

Geobits

I think introducing one would be a good way to push the builtin loophole into "valid"

Rainbolt

github.com/marbelous-lang

Martin Büttner

5:53 PM

@Rainbolt that one wasn't at all made for golfing though

that was more like a fun puzzle language with design goals similar to Manufactoria

Zgarb

What purpose does "|Q" serve that can't be done with "K|"?

Ismael Miguel

@Rainbolt I've read about it, but briefly.

Rainbolt

@MartinBüttner I know. I'm just pointing out that those guys knew how to get people interested.

Ismael Miguel

@Zgarb K| runs the code, |Q simply outputs the quine without running it.

feersum

the way some people might get interested in a language is if you implement it, and then get a shorter score than other langs in challenges

3

Geobits

5:55 PM

Worked for pyth, right?

feersum

not me though..I don't like golfing languages.

Ismael Miguel

@Rainbolt I know I don't know how to interest people, therefore asking opinions about it might at least spark a tiny bit of curiosity

Martin Büttner

I think the "get a shorter score than other langs" part isn't even important

just implement it and answer challenges with it to show off some cool features

Rainbolt

I think winning is helpful.

Ismael Miguel

I've seen @Doorknob 's language he showed a while ago

Rainbolt

5:56 PM

Nobody cares about my cool MTG decks until I win with them at FNM.

Ismael Miguel

He didn't won, but his language looked promising

Lembik

Hi

Martin Büttner

@Rainbolt true, but if you win because you have a generalised-quine built-in, that's probably less helpful than not winning and just showing off a nice unconventional paradigm

not this again...

Rainbolt

Disclaimer: I starred Greetings because it reminded me of something that got starred last week.

Martin Büttner

that was different

feersum

5:58 PM

if instead you simply added a variable that tells the index of your position in the file, then you could win every CH challenge

Martin Büttner

@feersum you also need to be able to ignore all further or all previous code

Ismael Miguel

@feersum That would be a bit hard to implement

feersum

I assume there would be a commenting method in most languages

Lembik

how can you get the first million characters of a text file in windows? I only know linux where you would use head -c

feersum

hard to implement ???

lol

Martin Büttner

5:59 PM

@feersum there's almost never a left-facing comment, especially not covering multiple lines

TheBestOne

@Lembik Copy paste :D

Lembik

And any thoughts onhttp://meta.codegolf.stackexchange.com/questions/2140/sandbox-for-proposed-cha‌llenges?cb=1 ?

Martin Büttner

(and GolfScript is the only language I know that has a right-facing comment covering more than one line)

feersum

loads of languages have /*

Martin Büttner

@Lembik that doesn't seem to have the answer ID

feersum

6:00 PM

many unfortunately barf on unterminated comment

Martin Büttner

@feersum without matching */?

feersum

AutoHotkey is the only one I found that doesn't barf

Martin Büttner

see :P

and I don't know any language that has the opposite

Lembik

1

A: Sandbox for Proposed Challenges

Finding all-but-one matches This challenge is about writing code to solve the following problem. Given two strings A and B, your code should output the start and end indices of a substring of A with the following properties. The substring of A should also match some substring of B with up to ...

feersum

anyway, a comment is by no means required in all cases

Martin Büttner

6:01 PM

no but for CH's challenges it would often be very useful

Lembik

How do you know exactly how many characters you have copied?

Optimizer

@MartinBüttner C*

Martin Büttner

no.

@Lembik position the cursor at the beginning of the file. hold shift. hit the right arrow exactly one million times. press ctrl+c.

feersum

@Lembik the indices would make more sense if you decreased the starting index by 1

then same index twice = empty string

do you have a text editor?

Lembik

Thanks.. android seems to have no support for chat except through chrome so apologies for not mentioning properly

Yes I have every free text editor :)

How about the coding/algorithms side of the question? Too hard/easy?

oh did someone downvote it??

TheBestOne

6:09 PM

RoranStronghammer started dancing and throwing a rock at the Fatalis to get its attention.

I like Thrax's humor :D

Geobits

@Lembik I use ChatSEy. Not perfect, but better than the alternatives I've seen.

Ismael Miguel

Well, I've updated it again and added a syntax for comments.

I promise this will be the last time I bother you again

If you want to check it again, you can go to github.com/ismael-miguel/mariogolf

Most likely I will skip my gaming session to make a tremendously basic implementation

@MartinBüttner I don't mind the star there, I just hate bothering people

Whoever put it there, removed it.

Lembik

6:25 PM

@Geobits Thanks

Optimizer

6:42 PM

Hi

No

Hi G

O, hi O

OhiO

O, geez

6:47 PM

Joke successfully explained.

Doorknob

7:04 PM

H OGF

Optimizer

HD

Wow, I am currently leading in source-layout

TheBestOne

@Doorknob @Optimizer Translation ?

Optimizer

No

H1

Rainbolt

Haha. Someone made a snowman outside my workplace. imgur.com/cPcEr2u

Doorknob

7 hours ago, by Calvin's Hobbies

Am I the only one referred to by initials around here? We are all about golfing things down. We can have MB and D-nob and ... O.

Read from there down

Rainbolt

7:06 PM

And then our Website guy turned it into this imgur.com/0SWepCZ

Optimizer

So you work at Genesis ?

Rainbolt

Yes. It's on my profile too codegolf.stackexchange.com/users/18487/rainbolt

HO HO HO

H1!

No that's HO3

7:09 PM

H3K

FUZxxl

3#'HO '

Rainbolt

Three pound hoe?

That's rather light

Geobits

You missed the space. It's three pound hospice. Still pretty light...

Rainbolt

7:29 PM

I just got invited to walk from one end of my building to the other end, three times, as a walk of shame for not turning in my steps every day. Along with two others.

Geobits

wow

Rainbolt

productivity at its best

TheBestOne

@Rainbolt What's more shameful, declining the invitation or accepting it?

Rainbolt

@TheBestOne Depends on your weight

PhiNotPi

CH has a sandbox submission with a negative score!?

I would have never expected that.

Geobits

7:40 PM

Wasn't me :P

randomra

Started a BF Sierpinsky solution but it would be over 200 characters...

and not too interesting either

PhiNotPi

I have an idea for a kolmogorov-complexity-hybrid question.

TheBestOne

@randomra That challenge is impossible in Java :(

PhiNotPi

The goal is to reproduce a list of numbers (say atomic electronegativity), where you are allowed to round numbers.

The fewer chars, the lower the score. The more rounding, the higher the score.

So, if you cut a byte off your program but lose a byte of accuracy, your score stays the same.

Martin Büttner

sounds interesting but the balancing could be tricky.

randomra

7:54 PM

@PhiNotPi apart from scoring/rounding is it the same as compressing a random digit-sequence

Martin Büttner

but a) you need numbers with a finite maximum accuracy or you'll get -∞ score and b) there's a chance that outputting 1 for any input will be the best tactic

PhiNotPi

I'm going to have to define "a byte of accuracy" precisely.

Rainbolt

You mean accurately?

PhiNotPi

Both, probably.

"Precisely" meaning that the definition is unambiguous, "accurately" meaning that it is balanced well.

Rainbolt

I always thought of accurate as "close to correct"

randomra

8:06 PM

@TheBestOne why impossible? I don't know java that much but don't see why impossible

TheBestOne

@randomra All possible ways of taking a Java program and stacking it on top of itself produce an invalid program.

Horizontal concatenation is possible; but not vertical concatenation.

PhiNotPi

The first step is defining a "digit" of accuracy. If I take a random integer and round it to the ten's place, then I would round it by an average of 2.5.

Rainbolt

@TheBestOne I managed to get this to stack once. Twice and it breaks.

class A {} /*/
class A {} /*/

PhiNotPi

If I round to the hundred's place, I round by 25 on average.

Rainbolt

Clarification: That's the code already stacked once.

feersum

8:21 PM

you didn't even get it to stack once because it's an error if there is only 1 line of that

Rainbolt

Yea, you're right. My compiler was just slow to let me know lol

randomra

maybe there is a unicode char to cancel out linebreaks or something similar

TheBestOne

Concatenate program in Linux, compile in Windows :D

randomra

@TheBestOne :D A special cross-platform program.

TheBestOne

8:36 PM

@Doorknob Where did you learn to code ?

Doorknob

Couldn't you use something like `// \` at the end of the line in Java to comment out the next line?

TheBestOne

No

Geobits

Even if you could, what would comment out the next one?

Doorknob

@TheBestOne I'll get back to you on that one later; have to go in a few minutes

Rainbolt

I'm stumped. I don't think it can be done without some overlapping blocks

TheBestOne

8:45 PM

@Rainbolt It's simpler to prove impossible.

Rainbolt

Really? Because a single example would prove it possible.

And you wouldn't know it was simpler to prove impossible until after you actually did so.

New Meta Posts

0

Q: Can I use Meta to vote for the type of my puzzle?

I am about to write a puzzle, but I don't know what type it should be. I think of Code Golf, but perhaps it needs to allow libraries. Then, I don't know whether and how much I should penalize the use of a library (e.g. triple your score if you used a library) Can I initially omit the puzzle type...

support

feersum

how about this:

/*
xx <CODE> /* *\u00
// */

where xx is hex for a /

then <CODE> is only uncommented the second time

Rainbolt

Is that Java?

feersum

yes

Rainbolt

8:48 PM

Illegal unicode escape

according to NetBeans anyway

feersum

oh, does it error if you have \u with less than 4 numbers?

Geobits

Right

Rainbolt

/*
xx <CODE> /* *\u0000
// */

Works fine

feersum

maybe if we split between \ and u

Geobits

But then you don't get the / from it.

1) You can't comment all blocks except one
2) A valid program consists of a class
3) You can't have multiples of that class
4) Java sucks for this

feersum

8:53 PM

it hardly helps for this challenge, but just as a curiosity, this code defines a class only when repeated twice horizontally:

/*
u002f class X {} /* *\
// */

Rainbolt

Wow you took proofs in elementary school?

Geobits

3rd grade to be specific.

feersum

maybe I can find an old CH challenge which only requires horizontal repetition

TheBestOne

I skipped 3rd grade :)

Geobits

See, that's why I had to prove it :P

We need anti-stars in chat.

5

Rainbolt

8:57 PM

We'll get anti-stars when Facebook gets unlike.

Geobits

I heard they're getting that on May 14th.

feersum

aww, I can't find any source-layout challenges with horizontal-only block duplication

Geobits

But you could write one ;)

feersum

lol

Geobits

The problem there is that it still only works for exactly two blocks, right?

feersum

9:02 PM

it could work for more

err

Geobits

Without defining X more than once?

feersum

actually, this would not be useful

since a horizontal block challenge would also have the trivial answer with // comments

Optimizer

Hi

Geobits

@feersum :D

feersum

int x = 1 // int x = //
+ 1 + 1
; //

Geobits

9:04 PM

I.... can't.... type....

Rainbolt

Would have worked thanks to a bug in 2004. bugs.eclipse.org/bugs/show_bug.cgi?id=60848

Martin Büttner

we're slowly reaching the point where I wish stars weren't anonymous so Doorknob could just ban the people from chat who keep starring those "Hi" messages

Optimizer

I agree

Martin Büttner

then again, he could just ban the people who post those messages to provoke the stars.

Rainbolt

I swear it's not me, but I imagine your anger only encourages whomever it is.

Optimizer

9:05 PM

No

feersum

isn't that a bug in the text editor's syntax checker allowing that, and not the compiler?

Optimizer

@MartinBüttner that thinking only led Infian Govt. to ban github, pastebin and co

Geobits

Exactly. They're being starred to provoke a response. Quit feeding ;)

Optimizer

ban the sites which encourages people to share code

instead of banning/punishing such people (who share bad code and all)

Martin Büttner

@Geobits wait, I just remembered... I'm a room owner :)

Rainbolt

9:07 PM

I knew the day would come when he abuses.

Geobits

Oh crap. Ban-spamming time.

Optimizer

@M you can unstar and stuff ?

Rainbolt

Either I'm really good (and fast) with photoshop or I am not the culprit: i.imgur.com/G297dJ6.png

Martin Büttner

no, but I can kick you out :D

Optimizer

abusing

Rainbolt

9:09 PM

Everyone else post a pic of your sidebar

Narrow down the suspects.

Geobits

No thanks. My past statements on the topic speak for themselves, I hope.

Martin Büttner

that's what they all say

Optimizer

^ - prime suspect

@MartinBüttner ughh

Martin Büttner

I shouldn't have gotten in the way there

Optimizer

Geobits

9:10 PM

Then again, with a newly-realized banner in the room, it might be a good idea to present evidence. Don't ban me Martin!

@Optimizer I was pretty sure it wasn't you, since you can't star your own messages. And you own both starred 'hi' posts.

Martin Büttner

(which is even worse)

Geobits

At least not today...

Optimizer

just helping

Sp3000

Obviously O did the starring on Optimizer2, right?

Optimizer

we have an O2 ?

Geobits

9:13 PM

I breathe it, along with nitrogen and trace gasses.

(given the mood, I almost want to edit that last one to say hi)

43

Q: I'm seeing stars! (I can see who starred a message and so can you)

So I was making a chatbot in Ruby for SE chat, and I discovered that I could find out the starrer of a message. I'm pretty sure stars, like votes, are supposed to be anonymous. Although this knowledge would help for cases of star trolls like this. Here's the specific slice of code that do...

support bug chat chat-stars

Rainbolt

I photoshopped i.imgur.com/FR3vOGO.png

Optimizer

i CSS'ed

Lembik

I'll leave my question one more day..I am hoping Peter Taylor will take a look

Rainbolt

@Optimizer That's... way smarter.

Optimizer

obv.

Rainbolt

9:18 PM

Well that blows this tactic :(

New Sandboxed Posts

0

A: Sandbox for Proposed Challenges

Story This is based on a real world puzzle. A colleague had a very old program. It was announced to be discontinued by the company and then it was really discontinued in 2009. Now I should use that program as well, but it is full of bugs and I refused to use it. Instead I wanted to help my coll...

Thomas W.

Hi. A high rep user may want to add my new sandbox answer to the list in the question, please. Thanks a lot!

Rainbolt

Done

I added a note "Story [Challenge type up for discussion]"

Thomas W.

9:33 PM

@Rainbolt: Thank you. "Story" reminds me of thinking for a good title :-)

Optimizer

source-layouts are the reasons why we need stop execution in cjam

Sp3000

And also why CJam should have infinite zeroes at the bottom of the stack

This challenge makes me think Golfish was a good idea after all, for the log

Optimizer

i dont see a benefit of infinite 0 wrt this question

Sp3000

So you can do something like 8\^)8\mL

Oh that's off by 1 isn't it

Point is, so you don't need to initialise, and make sure the initialisation happens once

Optimizer

U

the combo can save me a byte

)1+(
3mLn

Sp3000

9:49 PM

Yeah, something like that :)

PhiNotPi

I'm going on a function hunt!

TheBestOne

@PhiNotPi Why not a Monster Hunt ?

PhiNotPi

I need to find a function f(x) such that...
f(-x) = f(x)
f(0) = 0
The average value from 0 to 5 is 1.
The average value from 0 to 50 is 2.
The average value from 0 to 500 is 3.
etc.

@TheBestOne I haven't really taken a look at that challenge yet.

Monster Hunter looks pretty complex.

Also, I found this image, which would have been relevant to yesterday's discussion:

Optimizer

10:10 PM

4 more upvotes to cap, but i dont have any hope

later folks.

PhiNotPi

@Optimizer Cap of what?

daily rep cap?

randomra

10:21 PM

@PhiNotPi what kind of function? Z->R?

PhiNotPi

@randomra I'm thinking non-negative reals to non-negative reals.

Perhaps I should explain the purpose more.

randomra

than f(-x)=f(x) is quite a challenge

PhiNotPi

lol

Then just ignore the whole negative half then.

randomra

tell the purpose than

TheBestOne

f(x) = g(abs(x))

Now to find g(x)

randomra

10:25 PM

is it for sample generation?

TheBestOne

Does it have to be continuous?

PhiNotPi

I am trying to find a way to measure how much "rounding" has been done to a number, in terms of "digits". If you have a random number and round it to the tens place (1 digit lost), you lose from 0 to 5 from the value.

I am trying to create a challenge in which the program tries to output a certain set of numbers, but is allowed to round the numbers some.

If you round a number by 2 digits, then your program length incurs a penalty of 2.

I want a way to assign a certain amount of error to a certain amount of rounding.
I anticipate, also, that programs could output non-integers.

randomra

i see, a continuous, differentiable function would probably include li(x)

but can't you just check the first wrong digit?

or similarly like ceil(log(dif))

PhiNotPi

If I round a large set of random integers by two places (to the nearest hundred), the errors would be random numbers from 0 to 50, so I want the function to return an average value of 2 given random inputs from 0 to 50.

The first wrong digit wouldn't be useful in a case like 2000 -> 1999 or 120 -> 119. The percentage error is small.

randomra

@PhiNotPi you don't want to return 2 for 504 -> 500

PhiNotPi

10:38 PM

I know.

Similarly, the average value from 0 to 5 would be 1, I think.

randomra

but you want average 2 for only 6..50 right?

PhiNotPi

I think so...

It could be a step-like function, if we wanted to make it easy.

0 -> 0; 1..5 -> 1; 6..50 -> 2; etc.

randomra

that's 1+ceil(log(dif/5)) if I'm correct

yep, that

PhiNotPi

The idea for the challenge is to attempt to reproduce a set of data, like the list of atomic radii, as accurately as possible.

So that there is a balance between shorter programs and more accuracy.

Simply rounding/truncating the values should have no benefit, which is why I need a penalty function like this.

The key to scoring well is to try to use patterns (like the atomic trends) to compress the program.

TheBestOne

10:53 PM

(x * ln(10) - 10 * x * (ln(x) - 1)) / 9

I think that should work ?

PhiNotPi

@TheBestOne I graphed it and the curve goes downwards.

TheBestOne

Nevermind...

feersum

11:09 PM

what's wrong with log |err| ?

with a minimum error

PhiNotPi

11:22 PM

If I'm measuring in digits, I've got (1 + ln(2 * |err|))/ln(10).

feersum

why are you multiplying by 2?

PhiNotPi

Because an error of 50 signifies rounding to the nearest 100.

TheBestOne

x != 0: log10(|x|) + .7353015051
x == 0: 0

PhiNotPi

There would be a cutoff value of 0.183949, below which all errors give zero rounding.

TheBestOne

@PhiNotPi Then the averages won't work :(

PhiNotPi

11:36 PM

Well, log10 gives a negative value for err < 1

feersum

This answer is nice.. codegolf.stackexchange.com/a/47075/30688

the question sounded impossible without cheating by examining the program's own code, but it's apparently possible in c++

TheBestOne

|x| >= .1910955809: log10(|x|) + .7187493558
|x| <  .1910955809: 0

The cutoff doesn't miss up the averages anymore :D

Transcript for