The Nineteenth Byte

trichoplax

5:00 PM

@NathanMerrill Wow - I thought you hadn't commented on that answer yet...

Martin Büttner

(it decodes the integer as base 115, converts each integer to a character and evaluates the resulting string)

@trichoplax I think he's referring to your comment on the question.

quintopia

@MartinBüttner I don't know CJam, but I don't think I would want to disallow such a neat program. How does it work?

Nathan Merrill

sorry, I have a life, can't instantly respond :D

Martin Büttner

> (it decodes the integer as base 115, converts each integer to a character and evaluates the resulting string)

quintopia

huh

Martin Büttner

5:01 PM

when I say "decode", I really just mean "it gets the base-115 digits of that number"

quintopia

I feel like there should be some one true code restriction that would make atomic-code-golf work in the maximal, most general fashion (even if it isn't as simple as your preferred scheme) according to my above definition. For instance, just disallowing operating on code as if it were data.

I don't know if that would be well-defined enough

quartata

It's not.

The only way atomic code golf works is with a toy language

I tried wrestling with these problems a while back

There's no satisfying solution it seems

Martin Büttner

@quintopia I think there are languages where there is no distinction between code and data.

think SMBF

quartata

Like GolfScript

trichoplax

@NathanMerrill Ah I see - no I knew you'd get a ping for the comment on the question. I was just letting you know that I made a comment on the answer too, which won't ping you since you aren't in that comment conversation

quartata

5:03 PM

Blocks are specialized strings essentially

FlagAsSpam

@flawr Not really. I just know that do{}while(); does something while the bit in the while is true. Past the syntactical difference, I have no clue of the difference.

Nathan Merrill

gotcha :P

trichoplax

(no rush to respond - just letting you know I made a comment you might find relevant)

Martin Büttner

@FlagAsSpam do{}while() executes the body once before ever checking the condition. while(){} already checks the condition before the first loop iteration.

quintopia

@MartinBüttner And Purple. Nonetheless, the one writing the program would know if they were operating on code as if it were data. For instance, because they know that eventually that data will be evaluated as if it were code.

FlagAsSpam

5:04 PM

@MartinBüttner Okay, that's cool.

quintopia

(If it is never evaluated as code, it's not a problem.)

Martin Büttner

@quintopia restrictions on the semantics of a program usually run into Rice's theorem at some point.

(linky)

quintopia

@MartinBüttner Rice's Theorem says there is no automated way to check whether any given program does some certain thing without regard to how it actually functions. It doesn't really apply here, since we have humans here to explain what their code does.

Martin Büttner

an explanation is no formal proof though

quartata

^

Martin Büttner

5:06 PM

and I don't think you want to require that from your participants

(starting with a formal definition of "code" and "data")

quintopia

"don't submit a program that you can't prove doesn't eventually execute code it created itself in the process of running" problem solved

quartata

There also comes the point where you have so many rules that no one wants to participate in the challenge

The problem with humans

quintopia

I'm trying to come up with only one rule

Rainbolt

@quartata Have you seen the rules for Magic: the Gathering?

3

Martin Büttner

@quintopia "use this language" :P

quartata

5:08 PM

@Rainbolt But usually you don't need to know them all

(except for judges)

@quintopia What's your definition of a token for this?

Rainbolt

-10

Q: How many rules does Magic have?

More specifically, how many subrules exist in the September 25, 2014 revision of the Comprehensive Rules, located here? 1 Game Concepts Not a rule. This is a header. 100 General Also not a rule. This is a subheader. Just to clarify, CR 100.1 and CR 100.1a are two different subrules. ...

magic-the-gathering

One of the most downvoted questions on BCG

quintopia

@MartinBüttner That sounds far more complicated than mine, as it requires participants to read and understand the language you have defined.

Rainbolt

Whoops, sorry. It is not one of the most downvoted, it isthe most downvoted question lol

I deserve a badge for that

quartata

@quintopia Could be something existing like GolfCPU

quintopia

@quartata would that not discourage people from participating because they have to learn GolfCPU?

@quartata ah, wait. it can be compiled from C?

Martin Büttner

5:12 PM

wooo, Hexagony won a challenge :) (in a tie, but hey!)

@quintopia okay, let me ask a different question: what's your goal in using atomic scoring in the first place?

Optimizer

hey hey hey

oscars nominees are out

ETHproductions

@ETHproductions s/31/26, I miscounted the original

TimmyD

@MartinBüttner How the devil was I a computer science student for >5 years and this is the first I've encountered Rice's Theorem?

ETHproductions

@ETHproductions s/26/55, I golfed off another 29 bytes

132 -> 77

Cᴏɴᴏʀ O'Bʀɪᴇɴ

s/\//\\/g

quintopia

5:14 PM

@MartinBüttner Perhaps to have an eventual scoring method where the optimal python program and the optimal pyth program have roughly the same score. I had no immediate intent to use such a thing (as I'm not dissatisfied with bytes), but I do enjoy exploring ideas.

Martin Büttner

@quintopia Pyth will still win, because in Python you've got parentheses, and brackets, and delimiters, and fewer built-ins.

Atomic code golf might reduce the range of scores, but it will very rarely do much about the the advantage golfing languages have.

Nathan Merrill

@MartinBüttner I made the posts community wiki, so people could edit them with their specific rationals

New Meta Posts

-1

Q: Irrational Short Code Rationals

Many questions nowadays include a ridiculous rational as to why the code needs to be as short as possible. The frequency of these rationals apparently is starting to annoy some people (although, I am personally unbothered). After a discussion in chat, and on this question, I decided that a meta...

discussion

Martin Büttner

@NathanMerrill Right, but will they?

Nathan Merrill

I thought that was typical

Martin Büttner

5:15 PM

"oh there's already an answer with my opinion, voting, moving on"

Nathan Merrill

I can remove them

don't mind :P

quintopia

@MartinBüttner hence the "roughly". a multiple of 2 is close enough to be interesting.

Martin Büttner

as opposed to "hey, I disagree with this, let me make a case for the opposite side!"

@quintopia is it? o.O how do the actual numbers matter if the ordering stays the same? you could take the square (or cube) root of code golf scores, if that's what you're after.

ETHproductions

@ETHproductions s/77/67

quintopia

@MartinBüttner The number will be a better estimate of the complexity of the program than an arbitrary function applied to its length. Cyclomatic complexity would be better for seeing that, but it is very hard to compute. It'd be nice to have some single simple thing that still manages to yield estimates of the same order as cyclomatic complexity.

ETHproductions

5:22 PM

I have now golfed 49% off of my old answer. O_o

quintopia

Looks like Polyiamonds is a very difficult question. How does one coax Ell out of hiding?

Martin Büttner

@quintopia pray.

@quintopia but that means you're not even interested in an actual implementation. you want to score an algorithm. then why get the score from the source code in an arbitrary language? defining a toy language still seems like the cleanest way to achieve that goal.

quintopia

are people actually annoyed by the silly rationales?

Martin Büttner

yes

TimmyD

@quintopia Some are, some aren't.

quintopia

5:25 PM

@MartinBüttner I don't know how I gave the impression that I'm not interested in implementation...

Martin Büttner

@quintopia I was considering implementing it in Mathematica, but it wouldn't have been a very interesting solution, because I would have just converted the thing to a graph with the appropriate edges for adjacent triangles and then called some built-in to check whether the graph is connected.

@quintopia you're trying to make the score independent of whether it's implemented in Pyth or in Python.

quintopia

@MartinBüttner ah, I see. I didn't mean to imply that.

@MartinBüttner I'd probably end up doing it the same way, although that is insufficient of course. You need also to check for holes.

@MartinBüttner I'm trying to make the score independent of whether the tokens are 1 byte or 5.

Martin Büttner

@quintopia good point. in that case, two rounds of flood fill might be easier.

RikerW

@quintopia Yes.

Transcript for