The Nineteenth Byte

3:38 AM

@user202729 Let's take this to chat :)

I don't understand your question: If there is no range specified, the question is closed

if a bound is specified, then a O(1) answer is given, which is dumb

if it is specified to be unbounded, then the answer is invalid

Most of our code-golf challenges have no bound. Which is both good (people can use their native integer type upper bound), and bad (standard loophole "Abuse native integer data types")

For fastest algorithm, you have to have a bound

Practically there is usually some wishy-washy language like "you must implement an algorithm that theoretically works for arbitrary large inputs"

As far as I see, no [fastest-algorithm] has a bound. For example this.

With codegolf, people are pushing the bounds of the source code. This is why we have lots of rules around what counts as source code.

Anonymous

3:41 AM

@feersum Either that, or language like "all inputs and outputs will be within the representable range of your language"

however, with fastest code, the goal is to push the bounds of N

@feersum Yes, the "algorithm" works if the assembly instruction is extended to larger bit-width.

so you have to have rules around it

@Mego Yeah, but I find that eminently bogus.

Actually people also generalize it to "all intermediate values will be within the representable range".

3:43 AM

@user202729 Right, for arrays, the bound is basically infinite, as is common with arrays

Anonymous

@feersum What do you mean by that?

I think most people assume that arrays can go infinitely

I think Beatnik multiplication will overflow for A=24,B=2 and produce wrong result, although 48 is within integer range.

It's bogus because the challenge should not have different requirements for different programming languages.

Anonymous

@user202729 Sometimes, sometimes not. If there's multiple ways to solve the problem, then not giving that guarantee can affect which approaches are valid in a given language, which may or may not be desirable

3:44 AM

@NathanMerrill But it isn't true. As arrays in most languages are not lazy (not Haskell) or linked, their length must be stored. And if integers are not infinite width by default ...

Anonymous

@feersum While that's true, integer bounds are one of the least problematic things to allow to be dependent on the programming language.

@user202729 oh, for sure, practically, speaking, but in terms of challenge specs, I think it's fair to assume that we consider lists to be infinite

Anonymous

@NathanMerrill I think you mean unbounded, which is a different thing entirely

IMO most code golf challenges should have input bounds.

Anonymous

Most challenges don't require infinite lists, but do require unbounded lists

3:45 AM

right :)

sorry, forgot that infinite lists are a thing

Then do we "assume list length to be infinite" too? ... (remember "list length" === "integer")

It's only awkard to have bounds for things that involve asymptotic complexity (as then, everything is O(1)).

Anonymous

@feersum The problem with that is, either the bounds are tight to accompany all languages (opening up avenues for abuse), or some languages (like bf) get excluded.

then let's exclude them?

(this meta post is related to @msh210 's last question and I don't know how much does reading an integer take. link)

3:47 AM

@Mego I don't see the problem. Can you give an example?

Anonymous

@user202729 Not infinite, but unbounded.

@Mego Ouch.

@feersum BF cell is usually 8 bit.

So? It has "unbounded" number of cells.

Anonymous

@feersum Say we go for option 1. We restrict the integer bounds to one byte to accompany brainfuck. That causes a lot of sequence challenges to be able to be hardcoded, since many sequences with challenges have only a few values under 127 (or 255 for unsigned).

Anonymous

@feersum But its native integer type is 8-bit, so we would restrict the bounds to 8-bit integers.

3:50 AM

Technically they're not excluded, but we would make the work much harder for them. Implementing bignum is hard, ~~regardless of language~~.

It's silly to say that integers have to be represented as single cells in brainfuck.

@user202729 what's so ouch about it

A chain of N nonzero cells to represent N often works better anyway.

@feersum Then it would be hard to store extensible 1D array.

@user202729 you don't need bigints to have unbounded lists

3:51 AM

(which is as hard as storing 2D array of bytes)

Anonymous

Either that, or you completely defeat the purpose of setting integer bounds (making it fair across languages), because the bounds force at least one language to do much more difficult computation due to native limits.

@user202729 easy. a bunch of numbers with precisely one zero cell in between each nuber

@ASCII-only (wrong reply target)

@user202729 nope

I would suggest something around 10^9 for a typical integer bound.

3:52 AM

Why not just do 2^32-1?

@feersum Bf users will complain.

@user202729 let them?

@NathanMerrill unsigned > int...

@user202729 awesome. I've given you more room to golf

Anonymous

1 min ago, by Mego

Either that, or you completely defeat the purpose of setting integer bounds (making it fair across languages), because the bounds force at least one language to do much more difficult computation due to native limits.

3:52 AM

@user202729 I don't think so. It's not that popular.

@user202729 most languages have unsigned though. and long for that matter

@Mego since when are we trying to make it fair across languages

@feersum Much more popular than Dodos or Malbolge. dodos...

@NathanMerrill "All languages should be treated equally." – Mego

We should cancel PPCG because pretty much every challenge is more difficult in Brainfuck.

@user202729 right. Treated equally. But we don't do cross-language competitions

Anonymous

3:53 AM

@feersum That's an absurd reductionist strawman. You can't just dismiss the criticisms like that.

I treat all languages the same: They all get the exact same challenge

^

I don't give certain language benefits because they can't handle it as well

Hm. Fair point. If it's harder for BF, all solutions would be longer.

Still, it kinda detracts from the main challenge. The main challenge is "do X", not "implement bignum and do X".

Don't we permit unary I/O?

This makes implement bignum not required.

3:55 AM

@user202729 just use a BF variant with unbounded cells if you're too lazy

@feersum yeah, which is something I voted against

@ASCII-only ... well that will make clearing a cell harder... [-] no longer works.

Anonymous

@NathanMerrill What that actually means is, you give every language the same text describing the challenge, but some languages must jump through extra hoops. Which is fine, but that's not the way I prefer to write my challenges. I prefer to try to make the challenge as fair as possible across languages, by eliminating stuff that can cause significant additional challenge in some languages (which is the same ideal that guides our flexible I/O policies).

@user202729 why not

@ASCII-only Anyway, that's not the main point...

3:57 AM

@user202729 why not

Anonymous

@user202729 I agree with you there

@Mego Yeah, I do agree with this sentiment. I just think that there really aren't hoops for many languages

@user202729 it's like saying "implementing char is too difficult in boolfuck so we're just going to use bools"

for mainstream languages

Anonymous

@NathanMerrill The most significant hoop in many languages, across most challenges, is integer bounds, which sparked this discussion.

3:58 AM

I completely agree with golfing languages: you'd be completely gimping some of them

@NathanMerrill "not many mainstream" is still a lot in PPCG

Doesn't PPCG consist of mostly recreational languages...

@NathanMerrill >_> don't golfing languages usually have more features than mainstream languages

Anonymous

But golfing languages (and esoteric languages in general) are central to this community

@user202729 nope. mostly golflangs and mainstream

3:58 AM

@Mego I think most mainstream can handle integers up to 2^32-1 really well. If a golfing language's default integer isn't that big, then yeah, that's a big hoop

so yeah, I think golfing languages are hard to deal with

@NathanMerrill what kind of golflang would do that intentionally though :P

perhaps 2^31-1?

@NathanMerrill Esoteric, not golfing.

Huh? how could something without ints claim to be a golflang?

also basically all modern golflangs use a language with bigints

3:59 AM

@feersum Retina!

stressed out pokemon trainer

@feersum My specific limit is unsigned ints.

I do think that signed ints is more practical

orlp

►

Anyone think CPU draw most power should be reopened? It has objective winning criteria, but oddly it is a popularity-contest.

Also, currently Integer square root (older) has 4 unclear close votes.

@user202729 It doesn't seem good that the specific processor in not specified.

@user202729 there's a hard cap on power so you'd end up with a too-many-way tie

4:14 AM

> or measuring personally i use AIDA64 on Windows, Computer/Sensor/Power/CPU Package wattage reading; My configuration is i7 3770k @4.2Ghz @1.21V, with all 4 cores and HT enabled. Room temperature is 25C, i have ~1kg Noctua NH-D15 cooler. And Prime95 v294b8 shows 85W in that particular configuration.

Otherwise you could win by owning a processor with the most scores rather than writing special code.

or programs that just break the processor instantly

It says any AMD64-based processor is allowed.

@ASCII-only I don't think there is any program like this...

Why? Breaking stuff doesn't help a malware developer make money.

4:15 AM

^

@feersum Right. To make people annoying perhaps?

@ASCII-only Actually all challenges have hard cap. Some are easier to reach than others.

4:36 AM

For breaking the processor instantly (though without permanent damage): en.wikipedia.org/wiki/Pentium_F00F_bug

5:18 AM

What command line program should I use for password-protecting files?

Is there anything better than openssl?

Anonymous

OpenSSL is kinda the gold standard for that

Anonymous

Most other things you'll find are just wrappers around OpenSSL