The Nineteenth Byte

TheNumberOne

12:23 AM

How to search deleted answers in the sandbox?

Doorknob

2:00 AM

@TheNumberOne You can't—only mods have access to the deleted:{1,0,all} search operators. What are you looking for?

TheNumberOne

2:15 AM

@Doorknob Already found it.

Doorknob

Alright

Sparr

4:11 AM

@NathanMerrill ping

@NathanMerrill the A thread's first copy B+1 B in this bot seems to put a random value in B. have I misread the spec? gist.github.com/sparr/e186d287a527f6c9d372

Sparr

5:34 AM

@NathanMerrill yeah, I think it's a bug. I've got Copy B+1 B in one thread and Lock #B+5 in another thread. I expect the lock to apply to a random line (plus five, so still aa random line) and the copy to work. Instead the copy is producing a random result, too.

The copy has a random result even when Copy B+1 B happens all by itself.

I don't actually see any provision in the code for NOT randoming a read when a thread copies from a variable to the same variable.

Lembik

5:55 AM

@Agawa001 no unicode

Sparr

6:11 AM

@NathanMerrill updated gist to reflect another bug. D thread should be locking lines 4-12ish before A thread gets around to overwriting lines 7+ with flags. What's failing there?

Lembik

7:41 AM

hi @Sparr

Mitch Schwartz

8:16 AM

my uncorrected formula, multiplied by 2, gives the number of (A,B) if we also allow A[n] = -A[1] in addition to the A[n] = A[1] of the problem

and it seems that for even n, there is no way to have both A[n] = -A[1], and B without at least one 0

sorry, those A[n] should have been A[n+1], or the A[1] changed to A[0]

Mitch Schwartz

8:47 AM

well, also for odd n, because there needs to be an even number of non-zeros

Mitch Schwartz

9:44 AM

the 1-to-1 correspondence can be established as: for an allowable (A,B) where B has a 0, transform it into one satisfying the A[n+1] = -A[1] property by negating A[m+1:] and B[m+1:] where B[m] is the last 0 in B (this is reversible). so it just remains to prove that if B has no 0 then for A[n+1] = -A[1] there is no solution, probably a simple parity argument or something

Agawa001

9:59 AM

@Lembik u make my task harder

New Sandboxed Posts

10:22 AM

0

A: Sandbox for Proposed Challenges

How Many Isomers of Undecane Are There? Background Hydrogen and carbon form various series of compounds called hydrocarbons. Carbon (nearly) always forms four bonds and hydrogen forms 1 bond. The alkanes are the series of hydrocarbons without any double bonds between carbon atoms. The first fo...

Mitch Schwartz

@Lembik where should i post my alternate formula? as part of the old answer, as a new answer ?

Agawa001

@Lembik whats the max length of string input

(not here :/) anyone else knows ?

Nathan Merrill

10:42 AM

@Sparr pingback

@Sparr What should be happening is that both the lock and the write works

the lock shouldn't be locking a random line

@Sparr actually, I know why Copy B+1 B fails.

from the spec:

If two threads both copy to the same variable, the variable will end up as a random value. If they both copy to the same line, neither will work. If a thread copies to the same variable another thread is copying from, then the latter thread will copy a random value. If two threads are both copying from the same variable, they will both work fine.

oh, I talk about different threads

I never check to see if the write and read are on different threads

simply if I am writing and reading on the same turn

Mitch Schwartz

@Lembik i'll plan on posting it as a new answer later today, unless people think it's a bad idea

Agawa001

@MitchSchwartz this user isnt actually here

Mitch Schwartz

@Agawa001 i know

Nathan Merrill

@Sparr ok, the read/write to the same variable should be fixed

Sp3000

11:26 AM

@MartinBüttner For the Visual Multiplication challenge, "You may print trailing whitespace but it must not exceed the diagram's bounding box", do you mean the axis-aligned bounding box or the bounding box rotated 45 degrees?

Martin Büttner

axis aligned

will clarify

Sp3000

Damn :P k

Sp3000

11:43 AM

Hmm kinda surprised nobody's CJam'd that challenge yet. Current leader is... well...

Agawa001

@Sp3000 because not everyone can learn chinese ?? well notice that my method cud be useful for you (one single iteration with geometric coordinates of both lines y=x , y=-x)

Sp3000

Yeah, I've got something like that right now. Just trying to golf down a few of the longer expressions

Agawa001

if u jam this method , u ll be the effective winner

Sp3000

If I'm bothered :/ Don't think I will, tbh

Unless I can get what I want working

Agawa001

btw i didnt use 92/~i +i ascii twiddling , i used three variants and i know u get plenty of tricky magic up ur sleeve

Sp3000

11:52 AM

Well for Python I'm just indexing into a string...

I guess being able to multiply lists and stuff make things a bit easier compared to C

Agawa001

12:03 PM

someone knows how to set whole memory buffer to 0 in C with one command ?

Martin Büttner

linux.die.net/man/3/memset

Agawa001

thats eerie cuz compiler must do it automatically as first consequent procedure after declaring it

well (according to my last update)

well thanks @MartinBüttner

i must extend my next C codes by dozen of bytes henceforth

Sp3000

Hmm actually now that I look at it, I wonder why nobody's done visual multiplication in J yet, if it's got a take-diagonals operator

Martin Büttner

12:21 PM

@randomra ^

New Main Posts

12:41 PM

0

Q: Fastest gun of the west, king of the hill challenge

This is a king of the hill challenge everybody vs everybody. Its a battle the find the best gunman of the west! To be able to compete in this contest you need to make two functions, the first one determines the stats of your gunman and the second is the main logic function. Stat function funct...

king-of-the-hill

Sp3000

@MartinBüttner Hm... is it just me, or is 1*45 the other way around?

Martin Büttner

@Sp3000 fixed

Sp3000

:) thanks

Sparr

1:11 PM

@NathanMerrill did you see my mention that one thread is able to overwrite a line that another thread has locked?

@NathanMerrill also, outstanding bug from my earlier bot, one thread is able to copy from another thread's locked flag line

Doorknob

1:27 PM

@aditsu so remember when we were talking about implementing the newest xkcd? I started to implement it in Rust, but then I built such a versatile parser that I'm tempted to abandon the xkcd idea and just make a normal language/math evaluator tool... >_<

(and you were definitely right in that it is not golfable at all and would make a terrible code-golf challenge :P)

randomra

@Sp3000 You still need leading padding and gaps, and J is terrible with string manipulation. I would rather compute the type ('/\x ') of each position independently.

Sp3000

Damn :/

randomra

thought a bit about the question, but it was very unappealing for me to write it in J

it seems inconvenient in general, I'm surprised there are so many submissions

@MartinBüttner I can see a Retina solution though, that looks interesting

but no internet/PC for me next week

VisualMelon

2:02 PM

I have recently been introduced to the joy of Infra-sound by my brother... we though it would be a good idea to play a 17Hz sine wave from my computer, which has a subwoofer, but I don't know if it works, or even if I have drivers for it. Anyhow, after about 2minutes, my computer appeared to start smoking, the smoke seemingly disappeared after we turned the sound off. Now my computer smells of burning.

Which SE is the right SE for "Can Infrasound set fire to my PC?"

(I mentioned the sub-woofer, because it's rather hot, and appears to be the source of the burning smell)

(my computer does not have a history of being on fire)

Lembik

@Agawa001 I have included a sample input in the question. You can assume that the inputs won't be much longer than that

@MitchSchwartz posting a new answer seems like a great idea. Why might it be a bad idea?

Mitch Schwartz

@Lembik well just in case the convention would be to include it in the old answer for whatever reason, but i think, it makes a lot more sense to make a new answer in this case

btw, the formula is significantly faster, with similar straightforward implementation

BrainSteel

Hey all, I would like some more feedback on this challenge. I'll likely post it on Monday if there are no glaring errors.

Lembik

@MitchSchwartz that's great! I just added a large bounty to the more difficult question

@MitchSchwartz I am trying to extend Min_25's method to j = 2 but it's not at all obvious how to do it. His method seems really specific to j =1

maybe your method will be easier to extend?

Mitch Schwartz

2:20 PM

@Lembik it seems doubtful

Lembik

@MitchSchwartz ah ok.. well maybe j =3 can be solved up to 50 in any case :)

@MitchSchwartz I still don't understand how Min_25's solution deals with the wrap around. Do you?

Mitch Schwartz

it disallows an odd number of A[i] != A[i+1] instances, as feersum wrote in his explanation

Lembik

oh!

@MitchSchwartz so I was wondering something. We can look at A and B as being wrapped on a circle on top of each other. So my j=1 question asks the probability that two specific consecutive rotations both give zero inner products

@MitchSchwartz what happens if you consider two non-consecutive rotations? Is it any different?

Mitch Schwartz

hmm, i expected it would be different, but when i modified my brute force for a shift of 2, i got the same results

Lembik

@MitchSchwartz right! I am wondering if it really is always the same.. which would be really interesting

Mitch Schwartz

2:31 PM

for n = 1..6

Lembik

I find this whole area fascinating and it seems completely unexplored by math

it is ours to play with :)

Mitch Schwartz

sorry, i didn't look carefully enough :)

Lembik

oh is it actually different?

Mitch Schwartz

n=4 gives a different answer, 53/256 , but the others match

Lembik

how different is 53/256 ?

Mitch Schwartz

2:34 PM

89/512 vs 106/512

Lembik

hmm

can you try it for n = 8, say?

Mitch Schwartz

my brute force is written in python

n = 8 will take too long

well, i'll try it

Lembik

thanks!

maybe pypy will help with this code

Mitch Schwartz

n = 7 matches

Lembik

can you try n =7 with other shifts than 2?

please :)

Mitch Schwartz

2:38 PM

n = 8 did not match; 86249/1048576

i am using solve1a() of the code i linked you to earlier, i just modified in small way ideone.com/kwU01G

(the A*3 is wasteful for larger n)

aditsu

@Doorknob yeah.. I guess you've seen my java implementation

aditsu

2:58 PM

Theme song for the chat: youtube.com/watch?v=hT_nvWreIhg

BrainSteel

3:22 PM

Imagine how much trouble they went through to get that alligator in the music video.

Lembik

3:33 PM

@MitchSchwartz all intriguing.. maybe it works for odd n :)

Lembik

4:05 PM

@MitchSchwartz very nice new answer!

Mitch Schwartz

@Lembik thank you :)

Lembik

new score added

now it's time to focus on j= 2, 3 :)

Lembik

4:50 PM

@feersum hi

@Min_25 I have tried to extend your great technique to j = 2 for codegolf.stackexchange.com/questions/51624/… but it seems really hard. Do you think it is possible?

Min_25

5:20 PM

@Lembik I'm not sure ...

Lembik

5:34 PM

@Min_25 at least I didn't ask a question that was too easy :)

@Min_25 out of interest.. did you get a chance to look at "The Joy of X"?

Min_25

5:56 PM

I saw "The joy of X". I'm sure that this problem is also hard !

Lembik

@Min_25 ah :)

Nathan Merrill

@Sparr I fixed the locking bug I believe. Regardless, the changes moved LockedStationaryDoubleTap to the bottom.

Sparr

6:16 PM

yeah, if his attack thread can't read the locked flag then he needs more flags :)

@NathanMerrill are your changes on github yet? fixing the locking and random increment problems should make Copy and Self Flag pretty awesome, I think

Lembik

@Min_25 The Joy of X turns out to be a variant of a famous coin weighing puzzle

Abdou Abdou

7:11 PM

@Lembik i think i can do ur challende in O(n*max appearance of any character in string)

Optimizer

that's basically O(n*n)

Abdou Abdou

no

Optimizer

yes

Abdou Abdou

abcdaefg

O(8*2)

is not

Optimizer

aaaaaaa

Abdou Abdou

7:13 PM

O(8*8_)

well thats worse case

Optimizer

this is O so you have to take worst case

duh

Abdou Abdou

:/

Optimizer

well, I can do any kind of sorting in O(1) then

Abdou Abdou

wait wait wait

its still O(n)

Optimizer

I think you need to read again how worst case time complexity works

Abdou Abdou

7:15 PM

aaaaaa , abcde (b and c and d and e are empty , no need to access into array)

Optimizer

then your first statement above is incorrect

how about aaaaaaa, bbbbbb ?

Abdou Abdou

still O(n)

Optimizer

then this statement is wrong

6 mins ago, by Abdou Abdou

@Lembik i think i can do ur challende in O(n*max appearance of any character in string)

Abdou Abdou

aaabbbbb aabc is 2*3+1*5+0

i cant translate this

properly

Martin Büttner

@AbdouAbdou you're clearly using your sockpuppet to circumvent a chatban now. that does count as sockpuppet abuse.

Abdou Abdou

7:18 PM

oh againù

i have admirers now

Optimizer

he was chat banned here again ?

Martin Büttner

it's a general chat suspension across the network

Lembik

7:37 PM

for some reason my longest common substring challenge attracts answers of a very different quality to my other challenges

something about strings does this to people? :)

randomra

@Lembik I would guess it's the "no answer yet" part.

randomra

7:55 PM

@MartinBüttner how do you plan to add newline support for single file in Retina? I don't see a simple, efficient way.

Lembik

@randomra you mean answers get better in quality once there are >=1 ?

randomra

@Lembik you will get more lower quality answers if no answer yet and the requirements are not clear for everybody

Lembik

@randomra ok that makes sense

it's just funny because the mean standard of the answers I get is... "awesome" :)

Nathan Merrill

8:44 PM

@Sparr I forgot to commit, but the scores posted were correct

Sparr

cool, thanks

glad to see my later entry move up. I've got a plan for a much better one now that those bugs are gone

TheNumberOne

9:00 PM

Wow, Locked Scanner Bot did pretty good. Banana bot was squished though.

Martin Büttner

9:13 PM

@randomra it will always cost bytes, but I'll probably just parse the replacement strings for \n or $n or something like that before passing them to Regex.Replace

Sparr

not too surprised to see locked scanner bot move up so high, some of the other bots were abusing bugs

I need to go back and re-write some of my entries.

Doorknob

@aditsu I have not, actually. Is it on Github?

randomra

9:46 PM

@MartinBüttner if it costs extra bytes, no one will use it for golf, and we still have to give an explanatory paragraph about -s and newlines

@MartinBüttner you could use a rarely used char instead of $n like N, and if you want a literal N, you would have to write something like \N or $N

so you only need extra bytes if the rare char is being used, not when the newline

Martin Büttner

10:47 PM

@randomra I'm not sure I want -s to be on par (golfing wise) with multifile mode... you'll always have to mention multifile mode anyway, because otherwise you'd incur a 3 byte penalty for -s. I'll think about it though. maybe something weird with using \r instead of \n or keeping $n but making the n implicit if the next character doesn't generate a valid $ escape sequence.

randomra

@MartinBüttner oh, I thought you are planning making the -s mode the default

I remembered incorrectly then

if -s won't be default, there is no much reason to make newline small, the $n or \n version is fine then

Martin Büttner

11:18 PM

I don't see a good way to make -s efficient for golfing in all cases, unless I change the delimiting character to something absurd that will get stripped or converted by SE anyway.

Transcript for