and on the y axis I hope

so they are... I'm sure I check that I decided they were not xD

I am so glad that I couldn't figure out handling the underflow right away

I never would have found those beautiful striped patterns if the last version would have worked right away :D

they are beautiful

might have to get some A3 prints to put up my house

nice :)

@VisualMelon haha I was thinking about that too ^^

it's slightly disappointing that my earlier submissions (especially the pretty lame rainbow) are slowly catching up with the random walk one...

voters are fickle, regardless of context

yeah, bloody voters :P

far too generous

@Sparr would your CAs improve by applying periodic boundary conditions?

(if you have the spare characters)

Finally got down to 140 characters...

looks neat :)

Thank you!

fabulous

:)

I'm half expecting someone to come along with a ray traced reflective sphere in 140 characters.

@MartinBüttner I've got a few characters to spare, but not enough for more functionality I think :(

@githubphagocyte I'm expecting to see creative use of the three functions calling each other so someone can do a ~400 character greyscale function

that is, when green is called with legal coordinates, just return red for those coordinates. when it's called with illegal coordinates, do some math/processing for red

so red looks like { blah blah green(i+DIM,j+DIM) blah blah blue(i+DIM,j+DIM) } and then green and blue both start with "if i<DIM return red(i,j) else do some math for red"

you can all feel free to steal that idea :p

@Sparr unfortunately that's not possible unless the OP adds forward declarations to the framework

oh, drat

I asked him to add them

otherwise, port it to JS and do it there :D

weird... I just got my revival badge... although someone else got the 2 upvotes before me (although they posted later)

gonna see if I can get a little rep with a horribly bad answer to the missing letters challenge

ahhh, interesting

"No other answers posted earlier than your answer have a score of two. "Earlier" refers to the time the answer was originally posted (first revision)."

if nothing else, now there's a reference implementation

hehe

which runs perfectly nicely on word lengths of 3 or less :)

@Sparr I was heading down that route too until Martin Büttner pointed out the forward declaration problem - I was so disappointed...

can you just make a green-scale image, then?

err, blue-scale?

have red and green return nothing of importance for legal input

and have blue use red and green to do computation

I guess so

yes

So does that mean you could have a cyan image where blue and green both make use of red_fn?

there would be a lot more nifty options if we could put code above red, at a cost of bytes*3

@githubphagocyte I think so

I'd settle for just being able to call any colour function from any other (forward declarations) and shortening the function names - r(), g(), b() instead of red_fn()

functions can call themselves...

yes they can

indeed

cut my runtime in half for same words...

stupid constants, not improving my big-O

as long as the constant is big enough, that's still sufficient for a fixed word list ;)

word list is bigger for longer words

I'm still toying with the 3 and 4 letter word lists

The OP must have a very good algorithm to be providing mid-range length results as examples.

or he's running for a long time

The very long words give short dictionaries (probably shorter than the short word lists) but the mid range word length totals are huge.

I've got my runtime for 4,2 down to 22 seconds, a bit over half my first try

my longest word list is about 24k 8 letter words

iirc

28k 9 letter words is the peak, actually

however, more damning than the word length is the number of permutations of the missing letters

1 or n-1 missing letters is pretty easy

sqrt(n) missing letters is horrible

(maybe that's ln(n)?)

let's see how long 9,1 takes now...

I would guess n/2 missing letters is the worst case but I'm uncertain...

I can't remember how the combinatorics work out

Ah no I was thinking it was symmetrical but it's not

it's certainly a curve with low points at 1 and n-1 and a peak in between

the shape of the curve is moot if I can't actually run any tests on values other than 1 or n-1

2 and n-2 both go longer than I've been willing to let the script run

You're right - I think it is symmetrical - the letters themselves can't move - I was getting muddled

5,2 and 5,3 are tests I haven't let run to completion

I think I'll take maths graphics framework now, ditch all the restrictions, and see where I can take my random walk approach.

There are n ways of having one missing letter or (n-1) missing letters

Then n(n-1) for the next step in

and then sell repeating t-shirts with the patterns on them

Then n(n-1)(n-2) and so on

@MartinBüttner I'm impressed with the random painting. I still can't understand how it's possible to look in all 4 directions, when the rest of the picture hasn't been painted yet, and the main program generates them in raster order.

@githubphagocyte it's a recursive function ;)

and I'm memoising the pixels

9,1 is thrashing my cpu pretty hard

I tried culling my comparison list by checking hamming distance between words ahead of time. That didn't help much at all for short lists, but I think it's helping for longer lists.

That possibility made me wonder about Buddabrot but I couldn't understand how to treat pixels that haven't been dealt with yet

ah I see

might be hard to do within 140 characters

although now with complex.h it might be possible

then again, declaring a complex double alone is quite a mouthful :D

@MartinBüttner the memoisation is still making my head spin. Do you mean that when a given pixel looks in a direction that the main program hasn't reached yet, it calculates it anyway and memoises it, including recursively spreading over most of the rest of the image if necessary, before the main program has even considered it, and then returning the already calculated pixels later?

yes exactly... I'm computing the value the first time it's asked for, regardless of whether by main program or through recursion

That's brilliant

Should open up a range of possibilities that can't easily be calculated in raster order

thanks ^^

yeah I'm sure there are lots of possibilities that haven't been explored in the answers yet

I'm thinking a monochrome buddabrot might be possible with memoisation, just very slow (as buddabrots inevitably are)

Yes it's going to be interesting to see what arrives over the next few days - and that's before we get forward declarations...

it's unfortunate that this has been posted on the weekend

I just hope it doesn't leave the HNQ before Monday

@MartinBüttner since you did the initial work on the Mandelbrot images, do you want to try for a Buddabrot or shall I give it a go?

Do you know what determines what gets into the hot network questions panel?

Is it just activity or something specific like votes or new answers?

afaik it's an unknown mix of upvotes, answers and number of votes on answers

@githubphagocyte feel free to try it yourself. I want to look a bit more into random painting without having to golf now :D

I thought they were questions blessed by Jon Skeet when he didn't have time to write anything original of his own

guys

lol

guys...

we're getting forward declarations!

woo hoo!!

ray tracers and buddabrots and everything then!

:D

bailing on 9,1 after 22 minutes

and I can make my Hofstadter sequence submission valid if I can be bothered

@MartinBüttner HNQ?

@Sparr the computer I'm using at my internship is running a mathematica script all weekend for this challenge codegolf.stackexchange.com/questions/35457/…

@Sparr Hot Network Questions... that sidebar

oh, appearing on non-codegolf sites?

Will you be posting your random painting results anywhere if they are not actually a fit for the question? I'm curious to see...

@Sparr yup

@githubphagocyte I might just upload them all into an imgur gallery, so as not to spam this chat or anywhere else

You could then link from the painting answer to further examples without the restrictions

yeah good point

@Sparr have you tried starting with a random top row?

nice, just leaving out the blue channel gives something that looks very much like an autumn-y canopy

@MartinBüttner I have not. good idea.

I will try that if I mess with it again

@MartinBüttner just had a better look at your painting code ready to try a memoised approach myself. You have nested ternary operators?? I didn't even know that was legal... No parentheses or anything?

@githubphagocyte yeah it works in most languages with a ternary operator

you can also nest them on both sides

it's quite amazing :D

I use it a lot when golfing in ruby

both sides??

you mean you can put them in the true section and the false section?

I suppose it is always unambiguous I just didn't expect the compiler to do the work of deciding which bit belongs to which operator

yes, both in the true and the false section

I should have realised that "both sides" is ambiguous for a ternary operator :D

Hey, I'm so surprised at it being possible I'm ready to believe you could nest in the 1st of the 3 sections too...

that would probably be less useful and much more confusing

that's basically an if switch to determine what value you want to test for in your next if switch

I shall keep it in mind in case it becomes useful one day... ;)

well at least it can be used as a golfed if/elseif/elseif/else

(if you always nest it in the false part)

I'm going to need to rethink my memoised random walk attempt - I just got a stack overflow...

well you need a reasonable chance to terminate ^^

oh god... I wanted to try out a circular effect for the brush strokes

I couldn't have dreamed how insane that looks