The Nineteenth Byte

thejonymyster

6:00 PM

maybe split it into teams of teams

Adám

Making a language that has the worst from every language is way easier than the opposite.

thejonymyster

perhaps :) maybe there is a lesson to this :?

Adám

Like: require both braces and whitespace, while line ends are significant but you still need to terminate statements… with ⋄

There's line continuation, but it requires a symbol both at the end of the previous line, and at the beginning of the next line.

Even here, you'll end up in contradictions: You can do tacit programming but you still have to mention the arguments by name‽

thejonymyster

well that implies tacit programming is bad :o

am i out of touch

am i wrong, or does tacit programming already have "you can name arguments or have them be implied"?

i dunno, im wading a bit out of my depth thinking of implications and options

Adám

Tacit programming means you don't mention the arguments.

It is both good and bad. It can lead to really neat code and it can lead to really obscure code (or both).

thejonymyster

6:08 PM

i guess i was confusing it with something else, i was under the impression that the arguments were optional to specify, rather than strictly required to be absent

Adám

"Explicit" (non-tacit) code is both good and bad. It can lead to really clear code and it can lead to really verbose code (or both).

thejonymyster

if that makes sense, i said it weirdly

Adám

@thejonymyster Not just "can be absent"; it simply isn't tacit if they are present.

thejonymyster

ah, i understand

ty, hm

oh, i was getting confused with defaulting arguments

like f(a=1){return a}

Adám

Tacit programming

Tacit programming, also called point-free style, is a programming paradigm in which function definitions do not identify the arguments (or "points") on which they operate. Instead the definitions merely compose other functions, among which are combinators that manipulate the arguments. Tacit programming is of theoretical interest, because the strict use of composition results in programs that are well adapted for equational reasoning. It is also the natural style of certain programming languages, including APL and its derivatives, and concatenative languages such as Forth. The lack of argument...

Also apl.wiki/Tacit_programming

thejonymyster

6:18 PM

hm, so in the example

from functools import partial, reduce
def compose(*fns):
    return partial(reduce, lambda v, fn: fn(v), fns)

example = compose(foo, bar, baz)

this used to be a function example(x)

how would you now call this function?

Adám

You'd still call it like that. It is the definition that's tacit, not the usage.

thejonymyster

ahhh

i now understanding the everything

ty

Adám

I think it is easier to understand in APL. Explicit version: F ← {2 - 1 + ⍺ × ⍵} Tacit version: F ← 2 - 1 + ×

Notice how the mention of arguments ⍺ and ⍵ are gone.

So × now stands for the product of the arguments, rather than simply multiplying the explicitly mentioned-by-name arguments.

Similarly F ← {G H ⍵} can be written as F ← G H

So now H stands for "H of the argument" rather than mentioning the argument by its name, ⍵.

thejonymyster

that makes sense :D

would that be similar to, say, js array mapping?

Anush

any comments for codegolf.meta.stackexchange.com/a/23973/98590 gratefully received (I sent the wrong link earlier)

thejonymyster

6:24 PM

hm, maybe not now that i think of it (in re: myself)

Adám

@thejonymyster Huh, I don't think so. This has nothing to do with mapping. It is juggling functions.

thejonymyster

ill try to come up with a better example of what i was trying to get at in a bit

user

@Adám That sounds like the worst of both worlds

Adám

With various combinators, you can do awesome stuff. E.g. F ← {(F ⍺) + (F ⍵)} can be written tacitly as F ← +⍥F

Anush

@Adám does APL have an easy way to do this?

Adám

6:25 PM

@user That was the point.

@Anush Do what?

user

Ah

Anush

@Adám the challenge I linked about solving equations

user

APL can solve polynomials, right?

Or come close, I think

Adám

@user It can.

It can also easily solve linear equation systems (that's a built-in symbol: ⌹).

Anush

@Adám how about over the integers?

Adám

6:27 PM

However, I'm not well versed in number theory and restricting things to domains, so I'm not sure how one would do it. Maybe there's an easy way. I don't know.

Anush

ah ok

user

Might want to ask in apl orchard

Anush

I would ideally like to post it tomorrow so please do let me know of anything that is unclear

thejonymyster

@Adám ok i figured out what i was thinking of, and sort of why it was probably a bit off

arr = [1, 2, 3, 4];
arr = arr.map(x => x * 2);

Adám

See, the arrow function there is explicit, as it has to name its argument.

thejonymyster

6:31 PM

right, a tacit version would be like

arr = arr.map(*2);

right?

Adám

Exactly.

thejonymyster

:-) i understad

Adám

Although you might need to be explicit about the currying.

thejonymyster

whats this got to do with food :P

Adám

Currying means binding an argument to a multi-argument function.

So in APL, +∘1 and 1∘+ are both tacit increment functions.

Decrement is -∘1 and double is either 2∘× or ×∘2

thejonymyster

6:34 PM

oh, since * needs something on both sides

Adám

Exactly.

pxeger

@Adám do people not like it?

emanresu A

Scala has the default variable _, so your example would be a.map(_*2)

Adám

@emanresuA Hm, but isn't that explicit?

@pxeger I'm sure many don't.

emanresu A

@Adám Yes, although you don't need to name the argument.

thejonymyster

6:41 PM

@pxeger i dont like the idea of it, it seems scary that your code is being graded on how well you indent lol. plus maybe i have my own indenting ideas...

Adám

That's no different from APL's usage of ⍵ as default argument name.

Seggan

6:57 PM

how would i convert a string to a list in jelly?

i couldnt find any builtins for that

hyper-neutrino

wdym, jelly strings are lists

Seggan

ah, ig my testing wasnt thorough enough. ty

hyper-neutrino

@Seggan the atom ŒṘ will convert any object into the string representation of the actual object in Python itself - the return value is a list of characters since that's how Jelly does strings, so you can print it out and see what it actually is

since jelly smash-printing and bracket omission can make it hard to tell what you're actually working with

emanresu A

o (I still haven't gotten my arms back)

Seggan

7:17 PM

@hyper-neutrino yeah, all my char arrays were printed as strings. ty

Redwolf Programs

7:27 PM

Ooh I changed some settings and the markov chain song thing actually makes sense in some parts

E.g. the first line is "who walks among the broken away from your skin", which sort of makes sense

The grammar, at least

thejonymyster

i could hear that being in a song yea

Neil

7:42 PM

@AaroneousMiller I rely on that behaviour in many of my Charcoal answers

Anush

8:16 PM

Great to see the chatroom busy

Any way to count how many different accounts have posted to the main site today?

Redwolf Programs

Doesn't take too long to do so by hand

Anush

@RedwolfPrograms I mean including questions and answers

Redwolf Programs

There's not more than 20 or so answers in a day are there?

Seems like writing a script would be overkill

You could use the API though

Anush

I don't know.

On the topic of the sandbox. The problem is that no one says anything quite often.

Unlike when you post a question

Redwolf Programs

We're working on a solution, sort of

Anush

8:27 PM

@RedwolfPrograms solution for what?

Redwolf Programs

The sandbox feedback issue?

Anush

For the lack of comments on sandboxed questions?

Oh cool

What is your idea?

I wouldn't even mind giving bounty for useful comments

I have tried giving thanks for that doesn't seem to do it :)

Redwolf Programs

Guild of Reviewers, which should have a bot added to it eventually that does the "any feedback on xyz" automatically

And I'll add other tools to it

It could possibly show which users have done a lot of sandbox reviewing recently, which would make it more rewarding

Anush

Oh so high rated people will join the guild?

Redwolf Programs

No, that's just the name of the room :p

Anush

8:34 PM

Some reward is definitely needed

Redwolf Programs

Yeah. Maybe bounties at the end of the year.

Anush

Why not more immediate bounties?

A year is too long

Redwolf Programs

Well because +50 for every comment would be too much :p

hyper-neutrino

well it doesn't have to be that much you can just accumulate it, so like if you want +10 per comment just make it so every fifth comment they immediately get a bounty

Redwolf Programs

Yeah good point

Anush

8:50 PM

Can you give bounty for upvoted comments? That might make more sense

Otherwise there is a risk that idiotic comments will get a lot of bounty o

Redwolf Programs

I think it'd need to be more human than that

Anush

What do you have in mind?

Redwolf Programs

Just have voting or something, or use the more traditional system of people giving bounties to people who deserve it on their own accord

Wezl programs redwolf

just use my verbal blockchain cryptocurrency

Anush

Ok. I still think reward for upvoted comments on the sandbox could be good

People like immediate rewards

Redwolf Programs

8:54 PM

Seems like that'd require support from SE

Which isn't an option

Anush

They don’t like to help?

hyper-neutrino

9:18 PM

they're not going to implement something to incentivize a system pretty exclusive to very few sites; especially when it comes to reputation, I see no reason for SE to implement new rewards and totally change the nature of rep itself

"they don't like to help" is a pretty big misinterpretation but getting dev support for insignificant things like this is pretty much never going to happen

Bubbler

10:26 PM

@Anush You can write a sandbox post and then bring it to chat for more feedback

which is what all of us do regularly

hyper-neutrino

ok here's the thing right

comments are good for leaving individual suggestions or pointing our simple issues

comments are not for extended discussion, as the prompt literally tells you if you try to use them for extended discussion, because they add a lot of clutter and are intended to be temporary (however consistent or reasonable you think that statement is)

Bubbler

And it definitely helps to avoid being closed for unexpected or nitpicky "unclear"

hyper-neutrino

so I see literally no problem with people bringing their proposals to chat and having more in-depth discussions with reviewers in chat because it's just a much better way of discussing things

Bubbler

Like, I know a few users would cast a close vote if a challenge is written in math notation with no plain English explanation

user

11:07 PM

@RedwolfPrograms I think the $2^{n-1}$ rewards more alleles too much

Redwolf Programs

Idk, once you get past a few the penalty is basically zero and it doesn't matter. That was my main goal with it.

user

Answers like Wheat Wizard's could just keep on adding identifiers forever because the average byte count matters very little. It might do to make the numerator bigger

Fair enough

Redwolf Programs

@user No, at some point it's optimal. The exponentialnessness grows faster.

I think

user

@RedwolfPrograms Exactly, so you need to balance out the exponent

lyxal

@NewPosts oh boy that's bringing back memories from doing high school biology

Even though it was only a year ago, it feels like it's been longer since then

user

11:10 PM

Biology was a fun class. There was a snake, fish and these cute little chicks

hyper-neutrino

you're old now :p

Redwolf Programs

@user Nah, once you're above like 10 programs, nobody cares about the extra 0.0000000001 bytes you save by adding more programs

Even if you could keep going forever and getting a more optimal score, the limit is just the average byte count.

user

Oh yeah, that's a problem with the average

Max byte count instead of average wouldn't help too much either

Redwolf Programs

What is?

user

That at one point, adding a zillion byte program won't matter if you had a gazillion one-byte programs

*For some definition of zillion and gazillion

Redwolf Programs

11:12 PM

I'm not sure what you mean

There's only so many short programs possible

Bubbler

but (at least theoretically) there're exponentially many short programs

user

Yeah, but say you have a 50-byte base allele. once you add multibyte identifiers to the mix, there's a ton of 50-60 byte programs you could make

Redwolf Programs

I designed this specifically so that getting the lowest possible average byte count was the primary goal, with getting a few extra alleles being a small side goal that rewards you with a small decrease in your score.

Allele count isn't supposed to be a major part of the challenge because it's pretty trivial to increase

user

As it's defined right now, the alleles kinda make the average byte count useless

Redwolf Programs

TBH I'm not really sure I understand what you're saying

"The alleles kinda make the average byte count useless" doesn't make any sense to me

user

11:15 PM

Meaning that I could write alleles with a 1000 byte average but if I have enough of them, I can get a score as good as anyone else's

Redwolf Programs

No you can't

You can never get it below 1000

user

@RedwolfPrograms I mean that the average byte count becomes quickly weighed out by the denominator

@RedwolfPrograms Wait, why not?

Redwolf Programs

Are you sure you're reading the scoring right?

user

Oh I'm such an idiot

Redwolf Programs

You don't divide the byte count. It's a penalty on top of it.

lol

Grain Ghost

11:17 PM

The scoring is a little weird.

Like the penalty seems to pretty instantly become 0 as soon as you achieve a working solution.

user

I didn't see that it was the fraction plus the average bytecount

I literally thought the point of this was to make the score 0

Grain Ghost

I misread it that way my first time too.

Redwolf Programs

The main point of the penalty was that I didn't want people to just make two programs and that's it. I wanted a small reward for adding an extra few programs. I didn't think about the fact that they'd all already be payload capable quines anyway, though, making that trivial.

Grain Ghost

I mean, these sorts of things are basically impossible to score well ahead of time, so I wouldn't blame you, but I think it could be better.

Redwolf Programs

I'd probably just remove the penalty altogether if I redid the scoring

lyxal

11:19 PM

should the scoring section be edited to make it clearer that it's a penalty, not the final score?

because I accidentally read it as the fraction part being the score too

Redwolf Programs

I did edit it, I guess it needs more clarification

The final sentence/paragraph was intended to clear that up

user

Never trust people to read what you say, no matter how clear it is :P

lyxal

@RedwolfPrograms I more meant adding the scoring formula in a $$ block

Redwolf Programs

How would that help?

lyxal

just to make it extra clear the fraction is being added

e.g.

(I haven't actually submitted an edit, that's just a preview)

Redwolf Programs

11:23 PM

I have taken your advice and ignored it but ended up in about the same place :p

I didn't use the $$ though

lyxal

lol

user

If you didn't use it, give me the money

Redwolf Programs

I added a bold "This is a penalty, not a multiplier: your score will never be lower than your average byte count." at the end, and clarified "plus an amount determined by b2n−1 (making your total score b+b2n−1)"

lyxal

that's much better

Redwolf Programs

Anyone wanna do some of my calculus homework? I have like five pages of stuff on related rates and it's boring.

(Hot really :p)

Uh...*not* really

hyper-neutrino

11:25 PM

i can tell you that d/dx x^2 is 2x

user

@RedwolfPrograms For money? Sure

Or a bounty :P

lyxal

@RedwolfPrograms you, uh, sure that wasn't intentional. That wouldn't be the first time you've "leaked" more private stuff ;p

Redwolf Programs

Oct 11 '20 at 2:02, by Lyxal

Ight how much monopoly dollars you want?

lyxal

@RedwolfPrograms haha no I have my own related rates to deal with

user

@RedwolfPrograms The integral of e^(e^x) from -999999 to 123pi

No approximations, please, I like my monopoly dollars unrounded

@RedwolfPrograms Better offer: You do my calculus homework and I do yours (don't worry, there's no catch, I'm just offering to help you out)

Redwolf Programs

11:29 PM

I know the stuff on mine it's just way too boring to do it 40 times in a row

But I gotta go to ~~deal with the math police~~ get dinner

Adám

Challenge: Determine the x to maximise the value of f(x)=ˣ√x

user

I don't know the stuff on mine, but it's very exciting. Can we please trade?

@Adám 2.718? idk

Adám

@user Close.

user

CMC: Determine the derivative of x^(1/x) (totally unrelated to Adám's question)

hyper-neutrino

hmmm

user

11:31 PM

@Adám ¯\_(ツ)_/¯ I just put it into Desmos and looked at the grey thing

lyxal

@Adám that's just e, isn't?

It is.

Ohhhh

thought so

When I discovered that, my high school mathematics teacher gave me his TI-86 calculator with the reason that I'd make better use of it than he would.

lyxal

11:32 PM

because the first derivative has a root at x = e, and stationary points only happen at roots of the first derivative

Adám

@lyxal Wait, how did you determine the first derivative?

hyper-neutrino

it's been too long since i've done this lol; i do not remember how to do this

i want to say chain rule but idk if that'll work

user

There's an exponent rule that comes first, I think, but the chain rule is needed. Should be quite gnarly

Adám

iirc, it is a tricky one.

lyxal

@Adám symbolab.com

user

11:37 PM

I think it's (1-lnx)(x^(1/x))/(x^2)?

So either x=0, which is impossible, or lnx=1, which means x=e

@lyxal Coward :P

Grain Ghost

I just saw a slightly harder version of this problem on a practize math gre.

Adám

@lyxal That's very neat.

@GrainGhost Integrate ˣ√x ?

Bubbler

y=x^(1/x)
ln y = (ln x)/x
(1/y)dy = (1/x^2 - (ln x)/x^2)dx = ((1-ln x)/x^2)dx
dy/dx = ((1-ln x)/x^2)x^(1/x)

Grain Ghost

No it was something about infinite sums and you ended up having to find a fixed point and you could convert it to finding the maximum of x^(1/x)

A very good problem for a lot of the test problems is to just guess the most likely constant.

It's usually pretty easy to check your answer is right.

Adám

@GrainGhost How many kilometres in a mile?

Bubbler

11:45 PM

@Adám Golden ratio

Adám

Yup ;-)

Bubbler

That said, any feedback?

Adám

@GrainGhost e^π-20?

user

Or maybe the avogad constant

Bubbler

avocado constant?

user

11:50 PM

yes

Adám

Actually, better: ln(20+e)

user

Misspelled Avogadro because of avocad :P

@Adám Which problem are you guessing this for?

Adám

Also, I hear the 4th root of (9^2 + 19^2/22) is pi.

Transcript for