The Nineteenth Byte

Martin Büttner

1:00 PM

you guys are arguing that without code there's no proof of correctness. with code there isn't a proof either, so I don't see the point.

feersum

there's no evidence of correctness

Nathan Merrill

I initially agreed with you Martin, but the evidence is a really good point

Martin Büttner

Doorknob said "verify". a verification sounds more like a proof than evidence.

I agree that code will in general make it more likely that the answer is correct, but I think an answer containing a rigorous proof is better in that regard than an answer containing only code.

Nathan Merrill

Correct. It wasn't Doorknob that convinced me.

feersum

Doorknob isn't a computer science professor..probably not what he meant

academic proofs belong on a CS stack exchange, not here

Nathan Merrill

1:07 PM

For simple challenges, it is trivial to come up with an algorithm. If the algorithm is difficult, then it is tough to get it right and a small oversight can make your solution entirely unuseable

When you have code, that oversight will make it impossible to code

feersum

Good algorithm challenges are unlikely to occur

There's not that many things where there is a fast solution that isn't obvious or well known

Nathan Merrill

I think that that is irrelevant

a good one is the one Martin and I talked about

which I still don't believe ever got a solution

feersum

the tree thing? is that an actual post?

Nathan Merrill

nope

but it would certainly be a good fastest-challenge

Doorknob

Ah, @MartinBüttner when I said "verify" I meant "make sure it actually works."

Nathan Merrill

1:10 PM

I know its not np-hard, because facebook had it as a challenge

feersum

If the complexity of an answer is uncertain, one possibility is to put an accumulator into the program, and increment it each time memory is modified

since actual runtimes have little to do with theoretical complexity

VisualMelon

1:24 PM

there is a reason they call it asymptotic complexity ;)

Martin Büttner

@Optimizer just saw your comment on meta... that is by far the weakest point of the answer

Optimizer

yeah, but it was funny

Martin Büttner

"so far, almost no answers are free of code. therefore, we can't possibly accept algorithm-only answers."

Optimizer

"This is called Programming Puzzles and Code Golf, not Algorithm Puzzles and Complexity Golf"

Martin Büttner

if designing algorithms isn't part of programming I don't know what is.

VisualMelon

1:33 PM

off by one errors, mostly

Martin Büttner

and naming things

feersum

debugging the standard libraries

header guards

Martin Büttner

@Optimizer fixed the question

Optimizer

ok, but I think my comment still provide more info , so I will let it be there.

like it highlights the fact that generally, its a democratic decision

Lembik

my answer got downvoted alreadY!!

-1

A: Do all challenges have to be answered with code?

No not all answers need to contain fully working code Even with working code you can't necessarily check that an answer is correct. It might just work on a few cases. So it's not true that having code solves the problem of checking correctness. We currently allow answers which would need to be ...

well.. I tried :)

Optimizer

1:42 PM

yeah, it just appeared for me and was -1 already :D

Lembik

@Optimizer I am just trying to contribute :)

Optimizer

you did right

Lembik

is my answer really bad?

@Optimizer thanks

Optimizer

@Doorknob it should be "Algorithmic Puzzles"

and may be "Complex Golf"

Martin Büttner

@Lembik honestly, I think your points apart from the first and last are equally weak :/ ... I don't even know myself how to make a good case for either side of the argument.

Lembik

1:44 PM

@MartinBüttner well two good ones is good :)

@MartinBüttner although I am not sure I agree that the other points are weak. Why is "Pseudo code can be easier to verify for a non-specialist than an answer in code, especially if the code is in some minority language." weak?

@MartinBüttner and do you think the name issue really is important?

Martin Büttner

@Lembik the code doesn't have to be readable to a single user (that is impossible), as long as the community can verify its correctness

Lembik

@MartinBüttner well ok but if the community that can verify it is bigger that must surely be a good thing. What is the total number of people who can verify an APL solution and that read answers here?

compared to the number who could verify an answer in clear unambiguous psuedo code?

Nathan Merrill

ok. Here's a question: why do sites limit their scope

Optimizer

are you doing a retrospective ?

Lembik

@NathanMerrill a nice general question...

Nathan Merrill

1:49 PM

well, if I'm an asker, and I'm willing to deal with the pitfalls of algorithm-only answers

Lembik

@NathanMerrill we could have one SE for everthing.. and the user could choose subsets of tags

Nathan Merrill

why shouldn't that be allowed?

feersum

They don't want a bunch of unsightly garbage piling up everywhere...decreases property values.

Nathan Merrill

so, the debate is "Algorithm is unsightly garbage"?

Lembik

@feersum I think the question wasn't about garbage but more why don't we allow great questions about biology on codegolf.se

feersum

1:50 PM

I was replying to your general question...not the subsequent statements

Lembik

if I understood it correctly

Nathan Merrill

@feersum correct, but I'm applying your answer

feersum

algorithm-only answers are answers, not questions

and they are untestable...which makes them more likely to contain garbage

Optimizer

@NathanMerrill then I can also ask a question in which I am willing to deal with the pitfalls of image only answers where the image is found on google and the winner is the image appears the least number of times [...]

Lembik

@feersum I don't agree they are untestable. Why do you say that?

feersum

1:52 PM

@Lembik Do you have a pseudocode compiler?

Lembik

@feersum yes.. me!

Optimizer

My algorithm : 1 ) Take product of all even index pairs, (2) Do unicorn magic (3) Get inverse of the matrix . Time complexity : O(n)

Lembik

@Optimizer that's a terrible algorithm

@Optimizer there are good and bad algorithms :)

Optimizer

@Lembik doesnt make it an invalid answer, does it ?

just that I might not win or might get many downvotes

Lembik

@Optimizer well it makes it a terrible answer

Optimizer

1:54 PM

still a valid one though

Lembik

@Optimizer which I think is so bad it should be deleted

Nathan Merrill

Answers already should be verified by the poster (or community) to make sure they follow the rules

Optimizer

this is how garbage starts pilin gup.

Lembik

@Optimizer not really.. really terrible answers of any sort should be deleted

Nathan Merrill

I could answer every question with HQ9+

Lembik

1:54 PM

@Optimizer I mean I could paste irrelevant code as well

Nathan Merrill

that doesn't mean its a correct answer

Sp3000

@MartinBüttner I'm having trouble putting this into words, but I'm of the opinion "Almost always include code, unless the question is highly algorithmic/proof-based in nature". I'm thinking back to Book Stack Sort here, for example (grc's post has code, but most of it is the algorithm). Also there have been many questions with code but not likely to terminate any time soon, like the one for Graham's number.

Lembik

@NathanMerrill snap :)

Optimizer

we have no right to delete valid but "terrible answers"

Lembik

@Optimizer of course you do! If I paste lots of random C commands as an answer its just spam

Nathan Merrill

1:55 PM

oooh, that's a good point

what about answers that never terminate

we don't actually test those

Lembik

and I can happily write guidance for what a decent algorithmic answer is

Optimizer

@Lembik so how do you categorize that my algorithm is a spam or not ? (See the slippery slope point kicking in ? )

Nathan Merrill

we just assume that the algorithm is correct

Optimizer

@Lembik dude, that is impossible.

Lembik

@Optimizer well only impossible in one sense. Let us take this one at a time. If I just paste random C commands as an answer it will be deleted, right?

Sp3000

1:56 PM

Yeah, the closest thing we can do to testing it is trying very, very small inputs/changing the code slightly

Lembik

you would argue it wasn't really an answer or not really code I suppose

feersum

speaking of answers that take a long time to run...my Boggle-grid spam generator has currently produced 21.5 GB

Sp3000

Ahaha still running it?

Optimizer

"spam generator"

Lembik

@Optimizer it's no different. We make value judgements and an algorithmic answer which is completely unclear and/or makes no serious attempt to explain its steps is not acceptable

unless you want to define for me what "code" is :)

which is also impossible

it's no easier to define code than define algorithm.. but in practice there is no problem recognising either

I do agree that we don't want answers that only contain long paragraphs of English text supposedly setting out what an algorithm is meant to do. This can be dealt with either by requiring clear unambiguous pseudo code or by just downvoting I suppose

ok.. important question.. shall I offer 50 or 100 points in my bounty?

Sp3000

2:02 PM

Reminder: Bounties aren't magical :)

Lembik

@Sp3000 don't tell me that !

Nathan Merrill

What is the difference between refusing pseudocode and refusing code that supposedly works, but we can't actually test (because it'll crash our computer or take infinite amount of time)?

2

Lembik

@Sp3000 but are they monotonic? I mean does more points mean more/better answers?

Sp3000

@Lembik Not necessarily. If nobody wants to give an answer, then it wouldn't change much (a little pessimistic, I know)

Lembik

@Sp3000 :( Could it make it worse I wonder?

bounty added

Sp3000

2:05 PM

That was fast

Lembik

@Sp3000 you mean soon after the deadline for when i could add a bounty?

Sp3000

That, and the fact you added it a minute after you started this convo

Lembik

@Sp3000 ah.. very speedy :)

Sp3000

Anyhow, I'd like to bring in exhibit B: questions where it's impossible to execute the code because halting problem

Not saying anything like algorithm-only answers should be okay for that question, just that we do have the occasional theoretical question

Lembik

to be fair.. I understand the wish to see code in answers.. Let use say we say all answers should have code, how do we specify how much of the answer should be code?

feersum

2:09 PM

Too small of a reward can be detrimental

Lembik

@feersum disaster!

feersum

See section 2.4 of this review

Lembik

@feersum very interesting.. I often feel that myself

to be honest, I wouldn't mind that much if all answers had to contain fully working code that could just be copied and pasted without editing or adding anything. It would just be quite a different site .

it is nice what they have done at tex.se with that rule

feersum

What did they do? Everyone was wanting to write pseudo-TeX?

Lembik

@feersum answers have to contain full LaTeX documents that you can just copy and paste and compile

@feersum the aim is to provide helpful answers for people who don't understand LaTeX at all

Geobits

2:13 PM

Example of great question/answers with no code: Chessboard Compression

4

Lembik

well I don't know if they have to but they do and when they don't people complain

which is much the same

feersum

Dunno..that doesn't seem good necessarily

Sometimes if you are working from a full example, you're not sure which parts you actually need, and what can be removed

Lembik

@Geobits great example! Maybe someone can upvote my meta answer :)

@feersum the answers are meant to be minimal and the people are really helpful. It works really well

overactor

@MartinBüttner This might be up your alley:

Lembik

but then tex.se is perhaps the most helpful place on the Internet :)

Geobits

2:14 PM

@Lembik I'm still trying to decide if I agree with all of it. A clearer, simpler post with an example or two would be easier to upvote.

Lembik

@Geobits true

@Geobits please go ahead then I will delete my answer and upvote yours :)

overactor

Is there way of approximating the drag experienced by a person of certain size running at a certain velocity?

Lembik

@overactor you need shape and clothing

@overactor naked? :)

Martin Büttner

@overactor yeah the shape is pretty important

if you model them as cylinders it's probably pretty trivial

Lembik

@overactor thenakedscientists.com/HTML/questions/question/1843

overactor

2:16 PM

@MartinBüttner Just a rough estimate will do

They can be assumed to be wearing aerodynamic clothing

Lembik

@overactor onlinelibrary.wiley.com/doi/10.1256/wea.29.02/pdf

can you read that?

overactor

@Lembik I'll look through it

Martin Büttner

apparently the drag coefficient of a man is between 1 and 1.3

overactor

drag force = 0.36AV^2

So for similarly shaped people, drag is roughly proportional to the cross-sectional area and the square of their speed?

Or am I misinterpreting here?

Geobits

That makes a lot of sense, really.

overactor

2:24 PM

It probably only really holds in rather limited cases though

Martin Büttner

@NathanMerrill ummm... should your answer start with "No"?

you seem to be arguing that algorithm-only challenges are fine

feersum

Oh, that formula is only for people

Martin Büttner

@feersum yes, that's the 1 to 1.3 I'm referring to

@overactor that's correct.

@Lembik whether an answer is a code snippet or a full program is completely irrelevant. a person familiar with the language will know the necessary context for a snippet to make it run.

Lembik

@MartinBüttner I disagree a little depending on how much work is needed to test the code snippet. One of the complaints about algorithmic answers without code is that they can't be tested. Of course you can test them by implementing them or by running the algorithm in your head or by following any provided proof. If a code snippet can't actually be tested without some extra work then it seems to require no less expertise to test.

@MartinBüttner but you are right that it is not the most important point

Geobits

You could also argue that a well-proven algorithm description is clearer (and more verifiable) than code. See the [underhanded] tag for example.

2

Lembik

2:35 PM

@Geobits I agree completely

feersum

The underhanded tag demonstrates how code that looks okay can be flawed

Lembik

we also have no definition for code do we?

VisualMelon

I take it logical expressions don't count as code? (codegolf.stackexchange.com/questions/37648/…)

feersum

But if you run the code, the flaw is exposed

Martin Büttner

@Lembik we have a definition of programming language

I think anything that can be executed in what we consider a programming language is code

Lembik

2:36 PM

@MartinBüttner ok but is "for" code?

Martin Büttner

@feersum except for underhanded code that only fails on very specific inputs...

Geobits

@feersum If it's an obvious flaw, sure. If the flaw only surfaces 1/1000 runs, it's not so obvious (I'm trying to remember what challenge made me think of that...).

Martin Büttner

@Lembik sure, it just doesn't do anything

Lembik

@MartinBüttner ok so only answers that can be executed (without crashing) should be allowed?

Martin Büttner

in the right context, yes

Lembik

2:37 PM

@MartinBüttner you may be the wrong person to make this point to :)

Geobits

And take a finite amount of time?

Lembik

this isn't how codegolf.se works! we have plenty of answers you would never want to run

feersum

All solutions must run in a finite amount of time, where 'finite' means less than 10 minutes.

Lembik

ones that do something malicious for example

Martin Büttner

@feersum is that so?

Geobits

2:39 PM

@feersum Eww, no. Some code-challenges explicitly state "x hours" for time limit, and some are open-ended "highest n found".

Lembik

I just don't think it's true that we actually want all code to be runnable

feersum

This is a type of disclaimer I would try to remember to put in my challenges

Geobits

Oh, well that's fine. I thought you were saying that should be a general rule for all challenges ;)

Lembik

I used to put 10 minutes but got bored and reduced it to 1 :)

Geobits

I think one of the best arguments (to me) against broadly requiring code is that most people vote without running the code at all, on a regular basis.

2

feersum

2:43 PM

Good point...you can easily get 5 or even 10 votes with something that fails half the cases

Nathan Merrill

@Geobits I added that to my post

I hope you won't sue me

Lembik

I am going to delete my answer soon as everyone hated it :(

Martin Büttner

no please leave it

Lembik

it makes me feel bad!

Martin Büttner

you don't lose any rep from it on meta, and it's important to show this opinion as well

Lembik

2:44 PM

@MartinBüttner ok

Geobits

@Nathan Nope, my lawyer says it's a shaky case at best. You're off the hook (this time).

Nathan Merrill

hehe

trichoplax

@Lembik I like your answer and downvoted because I disagree with the conclusion, not because I disapprove of the answer. It's important that your view remains visible so we can see how many upvotes and downvotes it gets. A meta discussion with only one side is less useful for judging how the community feels.

Martin Büttner

@Doorknob so, do I have your permission to rename the Weekly Challenge to Fortnightly Challenge? ;) (I find "biweekly" ambiguous... it also doesn't sound as cool)

Lembik

@trichoplax ok thanks.. I just feel like the unpopular kid in school now

trichoplax

2:49 PM

+1 for fortnightly

5

@Lembik votes on meta are different - I know it feels weird though :)

feersum

Meta questions are often viewed as polls

Lembik

@trichoplax what do you think about requiring all code to be copy and pastable?

@trichoplax rather then just being snippets

Martin Büttner

@Lembik by default, snippets are not allowed

submissions have to be functions or full programs

Rainbolt

@MartinBüttner I'm a little concerned that Nathan Merril is updating the Java controller to your challenge two minutes ago while the challenge was posted 2 days ago.

Lembik

@MartinBüttner I may be using the wrong words.. see codegolf.stackexchange.com/questions/17312/… for example

@MartinBüttner ok so I meant not just functions

@MartinBüttner but a complete piece of source code one could run

Martin Büttner

2:51 PM

@Lembik the place to discuss this consensus is

8

Q: Default for Code Golf: Program, Function or Snippet?

One of the things many (new?) users forget to specify in code-golf challenges is whether the answers should provide a full program or a function (or even just a snippet). For most such things, we have defaults in the tag wiki (like scoring by bytes), but not for this. The conclusion in chat was ...

Lembik

@MartinBüttner thanks!

Martin Büttner

@Rainbolt when the challenge was posted, it explicitly said that the Java controller was not ready yet

Lembik

"This has my downvote. – Martin Büttner" :)

trichoplax

@Lembik Although I've cast my votes to encourage disallowing answers without code, because I don't want it encouraged, I'm not too worried about the finer details. In general, the community has a strong desire for objectivity, which tends to deal with any grey areas.

Martin Büttner

@Rainbolt also, I don't see further changes to the controllers as problems, as long as the spec and API don't change. valid changes might include: bug fix, performance improvement, new diagnostic feature/graphical output

Lembik

2:54 PM

ok I better go..good chatting all

trichoplax

Bye @Lembik :)

I think a good way to generate votes (up or down) for the code/algorithm meta discussion is to start posting questions that explicitly do not require code.

feersum

hopefully in the sandbox?

Geobits

Arguably, some of the image compression answers barely count as code (same with kolmogorov). If your entire "code" is print(bigassbase64string), then yea... We only make them wrap it in actual code because "that's how we do it", not for any intelligent reason I can think of.

Rainbolt

3:11 PM

@MartinBüttner Oh, I see. I didn't check the revision history of the problem. Whoops.

trichoplax

@feersum lol yes

Lembik

3:28 PM

I need to write code to run other people's (java) code and check they give the right output given some input

is there a lovely and clean way to do this? I am currently just hacking python as if it was bash

feersum

that's what it's for

Geobits

In general, I believe you just run it with subprocess.

Lembik

@Geobits ok thanks.. It's all very ugly. I also need to kill the process if it takes too long I suppose or if it gives to much output

Geobits

That's a bit more... fun. I haven't played with that much in python, but I recall some chat about it being a pain in the ass (specifically the timeout).

Lembik

ah

I ideally I would also prevent it from reading and writing to disk :)

I wonder where I can ask this question online

I don't want the code to read my ssh private key :)

Geobits

3:33 PM

answers.yahoo.com

Lembik

:)

what is answers.yahoo.com? Is it all written by yahoo?

it just looks like spam to me

Geobits

Look around. It will make you feel better about yourself and worse about humanity.

6

Lembik

:)

I still remember youtube comments!

globby

@Optimizer , or anyone else with adequate reputation, would you mind removing The Great, White, North! from the sandbox op, and adding Rise and Shine?

Martin Büttner

@Doorknob whatever your opinion on the weekly challenge was, you've been overruled by the stars on trichoplax's message ;) ... I'm going ahead with the change

Geobits

3:38 PM

@Martin I thought fortnightly was cool before it was cool to think it was cool :P

Martin Büttner

I know you did ;)

why though? is it particularly archaic usage or something?

(like "thrice", say?)

Geobits

It is in the US at least. I don't know about the rest of the English-speaking world.

feersum

It's slightly archaic, like thrice

Martin Büttner

kk... any other unambiguous alternative to biweekly?

Lembik

fortnightly? That's not archaic where I live

but we only have 60 million people :)

globby

3:41 PM

@Geobits Thanks for the edit.

Geobits

@globby No problem :D

Lembik

from the wiki " Fortnight and fortnightly are commonly used words in the English-speaking world"

so the wiki agrees with me :)

feersum

If it was really common, it wouldn't be necessary to say it was common ;)

Lembik

@feersum :)

I claim it is the only accepted word in British English

for the concept it represents

and every single English speaking person in the UK knows it

who is over 10 :)

Geobits

Ngrams shows it roughly 2-3x more common in British English than American.

Lembik

3:44 PM

thanks ngrams :)

feersum

It's superior to biweekly, which is ambiguous

Geobits

@Lembik American speakers also know it, we just don't use it as often. There's a difference.

Lembik

yes and also implies you should buy something weekly

@Geobits there is indeed a difference

@Geobits Brits know it and use it :)

Martin Büttner

changed the meta post

trichoplax

Where I come from people use the word "fortnight" every other week

Lembik

3:47 PM

or every 15 days in France!

which always confused me

I suppose it is correct

trichoplax

yeah how about we make the challenge once every metric fortnight?

Lembik

cubic parsec?

dcs.gla.ac.uk/~pd/Numbers/metric-imperial-rant.html

Geobits

FFF system

The Furlong/Firkin/Fortnight (FFF) system is a humorous system of units based on unusual or impractical measurements. The length unit of the system is the furlong, the mass unit is the mass of a firkin of water, and the time unit is the fortnight. Like the SI or metre-kilogram-second system, there are derived units for velocity etc. While the FFF system is not used in practice, it has been used as an example in discussions of the relative merits of different systems of units. Some of the FFF units, notably the microfortnight, have been used jokingly in computer science. Besides having the meaning...

Nathan Merrill

hey, do you guys see any problems with this sandbox post: meta.codegolf.stackexchange.com/a/4686/20198 ?

trichoplax

@Geobits That surely needs a PPCG question...

feersum

3:53 PM

idk, the breeding/killing mechanism seems a little artificial

needs some more visceral method of killing

Nathan Merrill

what do you mean?

feersum

idk, just saying random bs

trichoplax

@NathanMerrill I've tidied up my comments that you have already addressed, and left the ones that still seem pertinent

Nathan Merrill

ok, thank you

trichoplax

I still think that the instability of the population size needs thinking about. At present I'd expect a population to die out below a certain size and explode above a certain size, and I'm not sure the range in between is non zero.

Martin Büttner

3:57 PM

@Rainbolt nice meta answer, thanks. except "Feature requests belong in a new question." I don't think we should use meta for feature requests for random stack snippets. that's just clutter. the feature-request is for feature requests regarding SE. but feature requests in comments or github issues seems fine to me.

Nathan Merrill

@trichoplax I'm considering have 3 configurations, 2 of them being death

trichoplax

@NathanMerrill I've just realised that the death configuration will moderate the population as they will be killing their own cells too - ignore me :) I'll delete my comment...

Rainbolt

@MartinBüttner Feature request is by definition beyond the scope of the original problem, so it belongs in a new question. In what way does that "clutter" meta?

Transcript for