The Nineteenth Byte

Sp3000

00:18

@MartinBüttner Ahaha the increment/decrement thing was about the second thing I thought of

Martin Büttner

heh

Sp3000

I'm afraid that there might be a lot of trivial solutions to that problem

Martin Büttner

I guess that's why it's a popcon

Sp3000

Maybe :P

Martin Büttner

you can probably do some really fun elaborate stuff with Marbelous actually

Sp3000

00:20

Heh. I'm still a little confused about what exactly needs to be flipped, personally

e.g. I can do "ABBA"[:2] and claim the reverse is "ABBA"[2:], but what if I have "somestring"[::2]?

Martin Büttner

"The definition of a polarity-reversal is pretty loose, but use your judgment as to what makes sense."

on second thought, this does kinda warrant a close vote

for Marbelous the interpretation was pretty clear

but if we're not talking character-by-character, it's kinda weird

Sp3000

"reversing the order of code-blocks and stand-alone lines" is probably the craziest

Does that mean I need to do except: before try:?

Martin Büttner

lol

Sp3000

Puzzling meta: "[UPDATE - NOW ALL 5 EXCELLENT SCORES ON AREA51]"

:/ I wish people would stop doing that

Sp3000

00:36

@xnor Reading through The Green Llama's chatlog, thanks for raising the issue

hwnd

02:24

@Sp3000 Any luck on the regex challenge?

Sp3000

I've had a solution for a few days, but I can't post it because I was the last poster :P

That said I don't think we'll be able to keep the challenge going for much longer

hwnd

Maybe I am missing something simple that I can't wrap my head around and I teach regex (that is the sad part), this is the closest I've come in 29 chars but it picks up one of the bad strings. regex101.com/r/aF0rB0/5

Sp3000

I think it's not so much missing something simple, it's more trying to fit things together until you get something short enough

That just tells you that you can't use .$ unless you have a way of excluding that fail string

PhiNotPi

I think that the lifespan of the Regex Meta Golf was cut short because many submissions were sub-optimal.

Sp3000

[^?][PG]$|<|PG <-- all this guy's fault

If it was even [GP] instead of [PG] it'd be a lot more manageable

(or even better shuffling it so that the first option became the second option)

PhiNotPi

02:35

I feel like the score should be "your regex length + length of the regex that comes after you"

To penalize people who kill the challenge.

Sp3000

It'd also penalise anybody who had the misfortune of posting before a suboptimal solution, though - and it wouldn't be their fault

(or might not be their fault)

PhiNotPi

Okay then, maybe there is a specific way regexes should be sorted into groups.

Maybe, regexes that match themselves should go in the match set, and regexes that don't match themselves go in the reject set.

With the hope that people only have to make slight modifications to the previous regex to make it keep working.

Sp3000

Ahaha well I've been putting my (one) regex in the opposite set, because otherwise it wouldn't be much of a challenge

Martin did mention the possibility of a growing size limit, that'd probably solve most problems

(working out a good formula for that, though, would be the hard part)

PhiNotPi

For example, [^?][PG]$|<|PG matches itself and was place in the reject set.

We could probably look at how our current regexes have grown over time.

Sp3000

Well you probably don't want to make it too easy, e.g.

^(..[^^].{4,22}\$|[^?]+\w)$|2
^(..[^^].{4,22}\$|[^?]+\w)$|~

PhiNotPi

02:48

That's regex length over time.

FryAmTheEggman

Some of that was definitely my fault, sorry :(

Geobits

@Sp3000 I wrote a prototype ice-sliding game for Android some time back with dynamically generated levels. I'll look over it and see if I can think of anything later.

Sp3000

@Geobits Ahaha thanks. I did actually turn what I posted into a puzzle hunt puzzle once (the difficulty change I did being that rather than one moving object and paint, you had multiple moving objects that could block each other). Turned out to be my worst puzzle though due to unrecognisable letters, so I'm seeing what could have been done better.

But yeah, sorry for leaving the fewest moves part out - that was silly of me

Geobits

Ah, mine had the typical "casual game" three-star rating. If you got fewest moves, 3 stars, less than 150% of fewest earned you two, more was one.

3 stars was a bitch on some levels.

Sp3000

Ahaha k :P Well I thought it'd be interesting to post a question about trying to generate a level with a given solution

(solution being the word, not the moves)

Geobits

03:08

I think the general idea would be to (given a path) simply put a rock at each stopping point to start with. Then you can add more rocks in while ensuring that 1) they aren't on the solution path and 2) they don't form a parallel path with the same number of moves. It might be easiest to add in the "more rocks" randomly over many iterations to see which ones satisfy.

Sp3000

Hmm...

Geobits

It has me wondering if that will work with the coins and paint added in though. The coins are just such a dead giveaway to the intended path...

They're literally signs that say "go this way".

Sp3000

The coins aren't vital to the puzzle - they were just a way of enforcing the solution in the little prototype I have :)

(because the letter n requires going down and back up again, or at least that's what I thought at the time)

Starting with the solution path and adding more rocks seems like a fair idea

Geobits

Also, you should totally add some rocks on the edges. Your prototype clearly says that any path that doesn't lead to a rock is off-limits (endless loop), which narrows it down quite a bit. If there are several rocks around the edges (which may or may not be part of the solution), it gives you more paths to have to choose from.

Sp3000

:P true...

@PhiNotPi Yeah, look like the solutions started spiking after [^?][PG]$|<|PG :/

Unihedron

03:21

At school... Going to try to figure out this solution... How you solved this is still a mystery to me @Sp3000 but I will figure it out o_o

Sp3000

Er... I wrote a script that looked for a bunch of set patterns, essentially

Unihedron

Hmm... ok.

I'm going to try with JS-compatible then, since I'm on chrome

Martin Büttner

@PhiNotPi That's the funny thing. I tried playing it optimally myself before, and the strings grew very quickly. Geobits (who says he's not that familiar with regex) also gave it a try (presumably more suboptimally), and his grow more slowly than mine. So I assumed (inevitable) suboptimal answers would actually keep it going for longer

Sp3000

If anyone's interested, here's the output of my script :) pastebin.com/GPwXErVQ

Martin Büttner

@Sp3000 Yes, I've suggested changing the denominator to 30+N/5 or 30+N/10 (and the size limit such that the scores always remain below one)

Unihedron

03:26

:O

ermeergerd

Martin Büttner

I'm not a big fan of changing scoring, but I think neither change would affect the top 2 or 3 users, and I'd only do it with some consensus in the question comments that people would prefer to keep the challenge going

will post some comments for people to vote on tomorrow if none of you finds another answer until then

Unihedron

I will figure it out o/o

PhiNotPi

@MartinBüttner I think it makes sense that playing it optimally kills it quicker.

The less entropy that is in the patterns, the easier it is to make a short regex.

But by golfing the regex, you increase its entropy, thus making it substantially harder for the next entries.

Unihedron

Yes... It should come to an end soon, 29byte is actually pretty short for most average regexes

xnor

03:49

@Sp3000 thanks for appreciating my efforts on puzzling

i have gotten somewhat frustrated with the discussions

Sp3000

Don't worry, I'm of the same opinion :/

xnor

and the attitude of "if you don't like it, leave"

and "if you don't like it, just downvote"

Sp3000

I've only started posting on puzzling now because the site's content is quite skewed atm

xnor

also, thanks for writing a great puzzle-hunt question

that I can use an example of something that's a puzzle, not mathematical, and of quality

Sp3000

Ahaha no prob - I'd post more, but as is it's not currently a good fit for the site (just look at all the "clarification requests")

xnor

03:51

hopefully things like that can convince people that puzzle != low quality

you posted a new one?

Sp3000

I posted puzzling.stackexchange.com/questions/4964/… yesterday, being a bit closer to the types of questions currently on the site

xnor

ooh, i was quite a fan of the lopsy puzzle

Sp3000

And also puzzling.stackexchange.com/questions/5090/… in an effort to have some puzzle-creation questions

(Basically just my secret mission to skew the site away from brainteasers and spaghetti.)

xnor

oh, i think i've seen a puzzle with a similar mechanism

have you heard of Ricochet Robots?

Sp3000

Yep, I have - funnily, that's the first thing someone else said when I told them about my idea

xnor

03:54

so i've seen a puzzle that gave a series of ricochet robots problems, each with a minimal solution

and if you take the paths of the black robot in all the problems, it spells a message

lemme see if i can find it

Sp3000

Oh? Interesting...

xnor

mathcamp.org/gettoknowmathcamp/residentiallife/hunt/2006/…

Sp3000

:o that looks nice

xnor

is there some sort of state you can include in yours?

maybe switches that open doors when they are passed over?

a hard thing about spelling out a word with a single path is that it's hard to do anything but spell letter by letter

Sp3000

Anything like that could be good. Basically I'm just hoping that the question will generate a lot of different interesting ideas :)

xnor

04:00

i think some of the comments on there could be answers

they seem like solid suggestions that are upvote-worthy

Sp3000

Yeah, same :)

Basically my two goals are 1) I once did this as a puzzle and it turned out to be my worst puzzle that hunt, so I'm wondering how it could have been better and 2) Show that Puzzling can have good answers which aren't predominantly in spoiler tags

xnor

noble goals

Sp3000

I hope people don't mind... I think I'm just about as concerned as you

(Hmm having switches replace the coins could get very interesting...)

xnor

now that I think about it, though, switches might not lead to anything more imaginitive than each letter-path opening a switch that allows the next letter-path to be used

it's the segmentation into letters that makes it straightforward

oh, but you can make some paths paint and not others, right?

what if the required path was totally different from spelling out the word?

Sp3000

Oh? Well anything goes really, as long as a word is the ultimate answer. I mean, you could Morse code if you really wanted

What were you thinking?

xnor

04:07

i was thinking you would still spell out a word

but the path you take would be much more circuitous that the word

for example, instead of spelling out an O by making a rectangle, bouncing against walls for each turn

part of the path makes the upper section of the rectangle-O by passing through a "paint on" square and then a "paint off" square

so, the segment traveled is longer than the segment drawn

then, somewhere later along the path, you make another side of the O, and so on

am i making sense?

Sp3000

Hmmm yeah that's certainly possible :)

That'd make letters like F and n a lot easier, because in the prototype I have there I had to enforce that you go back along the path you came

(making it very restrictive and easy)

xnor

04:22

hey sp3000, if I posted a puzzle now, would you work on it in puzzling chat?

i want to see if i can drum up interest to collaboratively solve a puzzle

Sp3000

I could - why puzzling chat specifically though?

xnor

actually, a separate chat room that i'd advertise

or, what do you suggest?

Sp3000

That sounds good

Separate chatroom might work better

xnor

ok, will do

xnor

04:40

hi, I want to advertise my new just-posted puzzle on Puzzling with the hopes that people work on it in the same time for that collaborative experience

0

Q: Split: A puzzle-hunt style word puzzle

The answer is a six-letter word. You're strongly encouraged to discuss and collaborate on the puzzle in this chat room. I've had this puzzle playtested before, but I still welcome feedback and suggestions. 1) Diamond, to carbon 2) Deflated (#7) 3) That in “this or that” (#9) 4) Eureka pro...

the hope is to raise the level of quality in puzzling

so if you like puzzles but are tired of the low-quality stuff that's being posted on Puzzling, this is a chance to foster a better community

TwiNight

05:59

Ok I tried

I know too little about Java to not use an IDE

Beta Decay

06:58

@Optimizer There didn't seem to be anyone else going to answer, so I just awarded the bounty to you on Manhattan Police

Stack Exchange you cheeky sod: You haven't voted on questions for a while, questions need votes too... I'll vote on whatever I like, thank you! :/

xnor

07:46

so, apparently the way to improve the community on Puzzling SE is the have PPCG solve the puzzles instead :-P

6

Optimizer

@BetaDecay Cool :)

@xnor why should we do that ? PPCG also has puzzles, simply abandon puzzling :P

xnor

let's just write code that solves the puzzles for us!

grc

we can't let them graduate before us though

Optimizer

are they ahead of us ?

are we ever graduating ? do we want to ?

xnor

what does graduation give us?

grc

07:54

not much - it isn't a priority for either site

but it would still be sad if they beat us :P

Optimizer

are we starting a protest ? The questions are turning to be like those asked in protests

xnor

honestly, i think Puzzling needs to worry about surviving now rather than graduation

Peter Taylor compared it to the code trolling era of PPCG

but with a 15x increase in questions

grc

complete with meta posts like this

xnor

let's just say it's been a high-quantity site lately

reminds me of this: meta.codegolf.stackexchange.com/questions/1192/…

Sp3000

Cargo-cult puzzle-writing is almost what I'd want to call it :P

Optimizer

08:10

we are magnitudes ahead of puzzling, in terms of # questions, # answers, # users, % answered and arguably, quality

they just have 400 visitors per day extra

Sp3000

We're lower on questions per day, but that's to be expected

Optimizer

we might be now, but we have 3.5 times more questions

Sp3000

You're comparing 1391 days in beta with 188 :P

Optimizer

08:36

well, then the whole comparison of them getting graduated first is invalid then :P

user2179021

11:29

hi

I had a silly idea for a question I wanted to pass by someone

ah..no one here?

Sp3000

Pleeeeeenty of people here

user2179021

aha :)

well. here it is.. I quite often get sequences of positive integers from work I am doing

and want to work out what formula could produce them

it would be nice, if given some rules for what sorts of formulae are allowed, one could have code to formula-golf the answers

that is give the shortest mathematical formula following the rules that gives the numbers

what do you think?

for example, 1, 2, 6,20, 70

or 1, 8, 96, 1280, 17920

Sp3000

You'd have to say what types of formulae you're allowing

Polynomial? Trigonometric?

user2179021

where you are allowed to use factorial, multiply, binomial(n,k) and positive integer constants

for example

or and raising things to the power of a positive constant

err.. raising things to thepower of other thigns :)

I suppose a grammar would be helpful :)

but is that more or less clear?

I suppose I could allow division too but everything will in positive integers

Sp3000

Grammar would make things more clear I'd imagine :P

user2179021

11:35

yes :)

but first let me see if the question makes sense and/or is interesting

it's basically math-gold

math-golf

Sp3000

Well doing some research, there's already something like this: codegolf.stackexchange.com/questions/3485/what-comes-next

(always good to check what's out there)

user2179021

thanks.. that's a little different

but still interesting

I don't want to golf the code

just the output

(basically I don't like the pure code golf stuff... which I know makes me a bad person :) )

Sp3000

Ahaha well that's not true

user2179021

:)

Sp3000

You'll have to define what you mean by golfing the output then - least operators? How do you count coefficients?

user2179021

11:39

I think least number of bytes might do

what do you think?

I mean characters I suppose, I assume a character is one byte in size

so n^2 is better than n^20

and n! is better than n^2

Sp3000

How does binomial work then? I'm assuming factorial is !

user2179021

(n, k) maybe?

to make it nice and short

still shorter than n!/(n-k)!*k!

err n!/((n-k)!*k!)

Sp3000

Just checking that it's not the entire word binomial(n,k) ;)

user2179021

no :)

overactor

@MartinBüttner I meant a command line input device

user2179021

11:42

I suppose it could be b(n,k) if you think that is less ambiguous

overactor

--
}0
++

Sp3000

Sounds decent - write up a sandbox draft and see what other people think of it?

user2179021

it would be nicer if the rules were part of the input of course :)

Sp3000

Just as an example though, what might be an answer for the sequence 1,2?

n+1?

Or n?

(not sure what we're starting from)

user2179021

well we would have to decide where to index from :)

but we could say n+1

Sp3000

11:45

Well that's something that'd have to go in the draft I guess :P

user2179021

hmm

do you think it might need "if" ?

it would be nice if the rules for the formulae were universal

so you could in principle make any sequence with enough characters

Sp3000

You could put ifs/loops/what not, but then it's just a question of balancing how much work people have to do :P

user2179021

ok so how I can put the rules so that 1,2 is a valid answer?

in other words, just listing the numbers

Sp3000

Ahaha you can make any sequence with enough characters though, that's just interpolation I'd imagine

Oh... if listing the numbers is a valid answer then...hm...

user2179021

how would you make 1,2,6 using my rules?

(forget listing the numbers :) )

Sp3000

11:47

(n+1)!

user2179021

nice :)

how about 1,2, 7

I am not sure you can make any sequence you want can you?

Sp3000

Give me a sec...

user2179021

sure :)

Sp3000

(n+1)!+n*(n-1)/2

You can always specify any number using the n*(n-1)*(n-2)... trick

And golfing that down would just give interpolation

user2179021

you may be on to something :)

Sp3000

11:51

e.g. n*(n-1)*(n-2)*(n-3)/24 would give 0, 0, 0, 1, <blah>

Which is useful if there's only 4 terms and you need to add 1 to the fourth term

user2179021

ok so you have improved my question.. thank you :)

Sp3000

So, just for the sake of fun, I don't think listing the numbers is necessary :P

user2179021

great!

I am much happier

maybe I should allow recursion :)

f(n) = .... f(n-1)...

or maybe that is too much

Sp3000

Ahaha well that's something you'll have to figure out :)

But yeah fun fact about guess-the-next-number questions, you can always interpolate by taking differences

user2179021

oh..er.. how do I find the sandbox?

Sp3000

11:55

18

Q: Sandbox for Proposed Challenges

What is the Sandbox? This "Sandbox" is a place where Programming Puzzles & Code Golf users can get feedback on prospective challenges they wish to post to the main page. This is useful because writing a clear and fully specified challenge on the first try can be difficult. There is a much better...

discussion sandbox

user2179021

thanks!

how are you on formal grammars? Do you think you could give me a hand?>>

Sp3000

Er... not great, unfortunately :/

Martin Büttner

@overactor oh I didn't know subboard input devices acted on ARGV for the mainboard

Sp3000

(also fun fact I just interpolated by hand and for 1,2,7 you can get 2n^2 - n + 1)

user2179021

no problem

nice :)

writing the sandbox question now

Sp3000

12:01

Hmm now that I think about it, what do we need the grammar for again?

user2179021

just to specify what answers are alliowed

this what I have now

A math formula which maps an index n starting at 1 to the relevant value. A formula F can be made up as follows. It can consist of integer constants, *, /, (, ), ^ or (n,k). These are to be interpreted in their normal mathematical sense and (n,k) is to be read as binomial(n,k)

Sp3000

Hmm if anything can be used inside anything else, I think specifying the operators might be enough

user2179021

ok

Sp3000

Might make it easier to write :P

overactor

@MartinBüttner the more you know...

Sp3000

12:04

Just something simple like a*b, a/b, a!, (a,b) (meaning binomial) or something should be fine I think. And possibly things like -a if you want unary minus.

(sorry, for some reason I was thinking people might do something weird if there was no grammar but I can't think of how that might happen...)

Hmm I wonder how well binomial would do - it wouldn't be optimal, but it'd be really simple to write a program for

user2179021

what about 1,2,6,20, 70 ?

can you take a look at meta.codegolf.stackexchange.com/a/2518/9206 ?

Sp3000

1+n+3*(n,2)+7*(n,3)+19*(n,4)

user2179021

nice ! :)

but I think you can do it in less

as it is (n,n/2)^2 for even n

Sp3000

(Take differences between consecutive terms, then take differences between what you get etc until you reach a single number. Then read the leftmost numbers)

user2179021

(2*n/n) I suppose

(2*n,n) I mean

New Sandboxed Posts

12:14

0

A: Sandbox for Proposed Challenges

Math golf This challenge is about writing code that outputs the smallest formula possible for a sequence of position integers. Input A file consisting of a sequence of comma separated positive integers per line. For example 1, 2, 7 1, 4, 36, 400, 4900 96, 1280, 17920, 258048, 378...

Sp3000

Ahaha well maybe - not sure how I'd write something to recognise that though (maybe someone else might have better ideas)

user2179021

maybe I should specify the order of precedence. I mean... are all the brackets needed in 4^n*(2*n)!/(n!)^2 ?

Sp3000

Yeah precedence might be good to specify

user2179021

the issue with n!^2 isn't exactly that though is it

?

what do you think?

Sp3000

I'd imagine n!^2 to be (n!)^2

You do remind me of something though - is a^b^c = a^(b^c)?

Also I'm not sure how your third example works, I don't get 96 if I plug in 1

user2179021

12:22

oop

try 4^(n+1)*(2*(n+1))!/((n+1)!)^2

i.e. plug in 2

@So

@Sp3000 what do you think?

Sp3000

12:43

Oh, right sorry

Hmmm only thing I'm not sure on is the scoring

Martin Büttner

@Doorknob sandbox got unfeatured

user2179021

13:11

@so

@Sp3000 go on.. what are you thinking about the scoring?

Sp3000

I can't remember where but there was a discussion about using hidden test cases for scoring

Martin Büttner

6

Q: Winning conditions not measurable until question author decides to reveal test data set

This question has an interesting property, in that judging a winner is dependent on unknown data: Recognize handwritten digits This is actually not an uncommon way of things working in puzzle-space. In programming competitions, you often are given simple test data and then you submit your progr...

discussion policy tags

user2179021

I'll read it thanks!

Sp3000

Ah there we go, thanks Martin :P

user2179021

the problem is that if I don't do that you can just hard code the answers

which is boring

Martin Büttner

13:14

no problem

@user2179021 about that meta.codegolf.stackexchange.com/questions/2506/…

use a very large number of test cases

and put a hard limit on the code size that makes hardcoding impossible

alternatively:

user2179021

hmm

Martin Büttner

score on a fixed (still large) set of test cases (which is disclosed), but disallow optimisation towards the test cases. if you suspect anyone of doing something like that, you run their code on a different (randomly generated) set of test cases, and if they perform considerably better, they're disqualified.

(that is, your disclosed set must be generated by the same code)

user2179021

ok.. well first I need to improve the question

@MartinBüttner the problem is how many brackets do people need to include

Martin Büttner

@user2179021 brackets?

user2179021

@MartinBüttner 4^(n+1)*(2*(n+1))!/((n+1)!)^2 Are all those brackets needed?

Martin Büttner

13:19

I don't know about the outer one in the denominator

establish clear precedence rules

but it should be unambiguous to leave those off anyway

also, you might wanna call them "parentheses"

brackets are technically []

user2179021

so ((n+1)!)^2 == (n+1)!^2 is one question

the order of precedence isn't exactly the same

Martin Büttner

right

also, how long will the input sequences be?

user2179021

ah yes.. thanks

Martin Büttner

your examples range from 3 to 7 elements

user2179021

ok fixed

any more?

Martin Büttner

13:23

so the minimum is 1 element?

user2179021

yes

fewer seems silly :)

Martin Büttner

well, but more might seem reasonable

but that's fine

user2179021

more than 20 you mean?

Martin Büttner

no

user2179021

oh I see... a bigger minimum

Martin Büttner

13:24

a larger minimum that 1

yeah

don't worry

what do I do if I can't find any formula?

I guess I can resort to some algebraic form of a cyclic sequence of the given numbers

Sp3000

@MartinBüttner A formula is always possible

user2179021

@MartinBüttner well @Sp3000 has a general method that always works :)

Sp3000

(Huzzah for interpolation)

Martin Büttner

right

@user2179021 "A special rule applies to the factorial function ! which requires brackets if it is to be raised to a power. I.e. it is (n!)^2 and not n!^2." why?

user2179021

because n!^2 looks weird to me :)

Martin Büttner

13:29

okay ^^

user2179021

basically it should be !(n)^2

n! is some short hand

Martin Büttner

what about n!!?

user2179021

Isn't !! somethign specific

I mean not !(!(n))

or (n!)! I should say

Martin Büttner

what?

I'm asking should n!! be treated as n!! or as (n!)!

(n!! being n*(n-2)*(n-4)*...)

oh

I read "!! isn't something specific"

so yes, it is something specific, so what should it be? should n!! be allowed?

user2179021

ok tried to fix it again

!! is not allowed

Martin Büttner

13:36

maybe mention it explicitly

challenge idea: given a number determine what kind of factorial it is

user2179021

but !! is a new function I don't list

Martin Büttner

built-in factorisation disallowed

@user2179021 I'm sure you'll get arguments about it because "but you allowed that character"

user2179021

grr

I made another mistake

how do I say what I mean about brackets

I want n! to have brackets if any operator is applied ot it I think

ok try now

Martin Büttner

even if 3+n! ?

In that case, just call the operator (...!)

user2179021

hmm

could you help with the phrasing please

I could just list the operators I am worried about :)

Martin Büttner

13:42

as I said, just call the operator (...!) and it's fixed

user2179021

ok :) I hope my new version is fixed too

Transcript for