The Nineteenth Byte

Zacharý

00:17

In RAD, what should 1 2 3+4 5 do? I don't want it to error.

FrownyFrog

fill the short one?

5 7 3

same as Jelly

Zacharý

I was thinking something along the lines of (1 + 4 5)(2 + 4 5)(3 + 4 5)

FrownyFrog

@Zacharý that's +/ in J

I really like the Jelly behavior

Zacharý

Maybe I can make it a ⎕something controlled behavior?

@FrownyFrog What use cases would that have?

Adám

@Zacharý Why that and not (1 2 3+4)(1 2 3+5)?

Zacharý

00:27

@Adám That could be done as 1 2 3∘+¨4 5

Adám

@Zacharý As opposed to 1 2 3+⊂4 5

FrownyFrog

@Zacharý 1 2 3 +⍨ 4 5

should be doing this

Zacharý

@Adám On functions that vectorize to depth 0, your idea seems to work.

Adám

@Zacharý K has each-left and each-right for that.

Zacharý

@Adám RAD will as well, I'm just thinking of what to do with just + (or any other function like that)

Adám

00:30

@Zacharý How about falling back to ∘.+ if the arguments are mis-matched?

@FrownyFrog Wouldn't work for non-commutative functions.

FrownyFrog

i see

Zacharý

⍃⍄ or ⍅⍆ for each left and right?

I think I'll use the ⎕ idea, I'll just have to decide on a default.

Adám

@Zacharý Oh no! Not more primitive-changing ⎕-names! Plz.

″‶

ASCII-only

@Pavel what if the type system doesn't let it...

Pavel

Does " mean anything in APL?

@ASCII-only Sure it does. Booleans are ints.

Adám

00:37

@Pavel No, but ¨ does.

ASCII-only

@Pavel You do know it's invalid in most C-derived languages (i.e. the ones with more strict typing) right?

Pavel

@ASCII-only But in Python, booleans are ints.

ASCII-only

@Pavel Python says otherwise

bool is a subclass of int

int isn't a subclass of bool

Pavel

00:52

Right, and if checks ints

Python didn't even have bools for a long time

Zacharý

@Adám I'll generalize it: make it a monadic operator (which the user can change) which determines what certain functions do upon a length error

Pavel

Calling __bool__ on ints is A) pointless and B) super slow

Zacharý

Also, because I don't want to have to think of an order for those things

I'll also generalize ⎕IO as well: ⎕IO←2 will be possible.

@Pavel I have it planned as a ⎕UCS alias for RAD

Adám

@Zacharý Like a re-assignable system operator? ⎕MISMATCH←{⍺⍺/↑⍺ ⍵} for the fill option?

@Zacharý ⎕IO←1.5J¯2?

Zacharý

@Adám I have to stop somewhere, so only integers (or I might just make sure it's a number and do the required math on it)

Adám

00:58

@Zacharý I'd rather get rid of ⎕IO and settle on a default, but then have a monadic operator {⎕IO←0 ⋄ ⍺←⊢ ⋄ ⍺ ⍺⍺ ⍵} for the other option.

Zacharý

@Adám Isn't it {⍺⍺⌿↑⍺⍵}?

@Adám That's the one primitive-changing thing I'm keeping.

⎕ML is GONE though

Adám

~⎕IO∊0 1 is rarely useful. You'd better allow … to be a range generating function, so 0…9 is (⍳9)-⎕IO and …4 is 0 1 2 3 4, for example.

@Zacharý Yes, typo.

Zacharý

@Adám Knew I was forgetting a codepoint.

Let's assign → to be that: there is NO way I'm programming gotos.

Adám

@Zacharý Why, though? If Ⓞ←{⎕IO←0 ⋄ ⍺←⊢ ⋄ ⍺ ⍺⍺ ⍵} then {⎕IO←0 ⋄ blah blah} can be replaced with {blah blah}Ⓞ.

Zacharý

@Adám It's just a pointless generalization, just like the number system.

Adám

01:02

@Zacharý Have you settled on a character set already‽

Zacharý

If you look at the bitbucket page, RAD_IO (the ⎕IO internal variable) is in a file called RAD_Defaults, so I'm hoping I can get away with what APL does.

Adám

@Zacharý It is interesting how newbies seem to like the settable ⎕IO. APL old-timers universally hate it.

Zacharý

@Adám Yep. bitbucket.org/zacharyjtaylor/rad/src/master/RAD_document.txt

And ‽ is a codepoint.

I ran out of APL in Miscellaneous and technical.

ngn

@Adám I've hated it since the time I first saw it

Zacharý

I am neutral when it comes to just programming (uses of it both ways), but like it for code golfing (since either is accepted without cost, and both can be useful)

But ⎕ML is definitely pointless if starting from scratch

Adám

01:07

@Zacharý What will monadic ∊ do?

Zacharý

@Adám Flatten. (Monadic ⍷ will do similar, just with differences with strings)

Adám

@Zacharý Will you have a prototype system like APL?

@Zacharý Strings?

@Zacharý Have you decided whether monadic f/ should use the "insert" system or the "reduce rank" system?

Zacharý

Wow, Dyalog does not like vectors of characters distinct from strings

@Adám ∊'aa' 'bb' will be like Dyalog, and ⍷'aa' 'bb' will just leave the vector unchanged

@Adám R AD. So, obviously the first one

Adám

@Zacharý OK, "reduce depth" then. I.e. will ,/'abc' 'def' give 'abcdef' or ⊂'abcdef'?

Zacharý

@Adám Yes, I might even have an atom for getting the prototype.

Adám

01:13

@Zacharý Like monadic ∊ in ⎕ML←0

Zacharý

'abcdef'

Yeah, maybe one of these in the monadic function case as a prototype atom: ∍⋾

Adám

@Zacharý So monadic ⍷ flattens everything except leaf character vectors? Why shouldn't it do the same for leaf numeric vectors?

Zacharý

@Adám In case one wants strings to be treated separately from numerics. I remember running across either the ∊ and ⍷ cases in different languages and esolangs, but not both in one.

Adám

@Zacharý In my imaginary APL where ⎕IO is fixed to 1, array[0] will give the prototype. (And array[¯3] gives the 3rd last element.)

Zacharý

@Adám Might add that as well (definitely adding in the "last element part")

Adám

01:18

@Zacharý Sure, I can see the need for both, but why make monadic ∊ and ⍷ the same on numbers?

Zacharý

@Adám They are. (⍷(1 2)(3 4))≡∊(1 2)(3 4)

@Adám If this ever is a thing: ÁPL

Adám

@Zacharý But how can that work with ⎕IO←0? Which number would you use for the prototype, and would the last element be ¯0‽

Zacharý

No, just generalize it, even for weird results. ⎕IO-1 is prototype, ⎕IO-2 being last element.

I'm not expecting RAD to be like ngn/apl or Dyalog.

Adám

@Zacharý Uh, array[¯2] for the last element. You're kidding, right?

@Zacharý So ⍷(1 2)'abc' is 1 2 'abc'‽

Zacharý

@Adám Yeah.

@Adám No. The main reason for ⎕IO in RAD is for golfing purposes

I know! I'll use á as an always (⎕IO←1) indexing!

Adám

01:23

I like the symmetry in 4⌷array: 4th major cell; 4↑array: 4 leading major cells; ¯4⌷array 4th-last major cell; ¯4↑array: 4 trailing major cells; 0⌷array: prototypical major cell; 0↑array empty prototypical array.

Zacharý

0↑array definitely won't work in RAD: that'll always be ⍬, or [] in Pythonic notation

Adám

@Zacharý Or '', no?

Zacharý

Whoops, yeah. But it won't work with any kind of higher dimensional stuff.

Adám

@Zacharý What does it even mean to support prototypes if you don't support empty arrays?

Zacharý

@Adám I made a mistake

@Adám have ↑ work as expected

Adám

01:27

@Zacharý This is why I dream of folding depth into rank, not the other way around. DAD.

Zacharý

._.

Adámic Depthless APL-ish Masterpiece

@Adám By mistake, I mean typographical error

Adám

Getting late. ○/

Zacharý

wait ... what symbol is that?

Holy crap. It's ○. dang it looks weird

I'll start programming in actual evaluation logic tomorrow, still haven't decided on a default ⎕M/⎕MISMATCH

ngn

@Zacharý how are you going to represent arrays?

Zacharý

(1 2)(3 4)

The parser can already do that

ngn

01:38

@Zacharý I mean what does that parse as? what does the underlying Python object look like?

Zacharý

[[1,2],[3,4]]

ngn

@Zacharý so APL array = Python list

Zacharý

RAD ~~array~~ vector = Python list (I call them vectors for RAD since there's no rank)

ngn

How will you distinguish between ⊂1 2 and ,⊂1 2 ?

Zacharý

... pardon me for saying this: there is no ⊂

ngn

01:40

@Zacharý oh, ok

Zacharý

It's not meant to be an APL, J, nor K dialect. It's more like an unintelligible variety of all three (mostly APL and K) :D

@ngn Are those different in K?

ngn

@Zacharý in k ,1 2 is like APL's ,⊂1 2

there are no enclosed scalars

only lists that can contain other lists

Zacharý

@ngn Yeah, RAD has surprisingly alot of stuff in common with K...

ngn

however, simple scalars do exist: 1 is distinct from ,1

Zacharý

@ngn Same in RAD (probably even same notation)

APL is to J as RAD is to K ... pretty much

ngn

01:46

@Zacharý to a large extent k is discovered, not invented :)

Zacharý

But of course: RAD might win a few golfing competitions just by removing {} :P (I won't do that for a submission, unless it's trivial_

ngn

@Zacharý yeah, I must add those in every ngn/k answer and that bothers me too...

Zacharý

@ngn That's why I'm making it semi-esoteric, that way it's still understandable, but is competitive within APL/J/K.

α:1
ω

Is perfectly legal

(And RAD fills up a codepage with potential atoms)

PhiNotPi

Hello peoples

Zacharý

Hello φ~π

PhiNotPi

01:52

What language are you talking about?

Zacharý

A Rankless APL Derivative (a language I'm developing, in the vein of APL / J / K)

AKA: RAD

ngn

@Zacharý or φπn't according to the latest meme :)

Zacharý

RAD is definitely already better than MY

@DJMcMayhem, I ask your opinion: should this be moved to a room?

PhiNotPi

02:08

Unrelated: I'm about to buy a new domain name and hopefully (eventually) turn it into a cool website: neuroscience.space

Zacharý

02:52

Okay: I'm going with the "fill" that @FrownyFrog gave

user202729

03:46

@Anush And that's why we need MathJax.

Can we have MathJax back (part II)?

FrownyFrog

I just discovered that 1 2⍪ 1 3⍴3 4 5 doesn't work

Zacharý

._.

Length error, mate!

FrownyFrog

That's how you do the fill thing in J

It appends with a 0, then you sum the rows

seriously, you can't do that in APL?

Zacharý

APL > J

FrownyFrog

L is for lame

Zacharý

04:01

At least APL knows how to be original

FrownyFrog

and for LENGTH ERROR

Zacharý

... that's why RAD'll be a thing

(AKA APL + K's love child, pretty much)

FrownyFrog

LAPK

PLAK

Zacharý

Rankless APL Derivative.

FrownyFrog

:)

Zacharý

04:04

How many bytes does it take to do {⍺=2:3⋄⍵∊1 2:4⋄⍺+⍵} in J?

RAD:

FrownyFrog

I can't read that

Zacharý

⍺=2:3
⍵∊⍳2:4
+

FrownyFrog

what's :

Zacharý

branch, sorta like if

If left argument is 2,return 3. Otherwise if right argument is 1 or 2, return 4. Otherwise, return the sum

FrownyFrog

thanks

Zacharý

04:06

How many would that take in J?

FrownyFrog

I don't know

less than a kilobyte

Zacharý

._.

Does J even have an if-like structure?

quartata

yes

Zacharý

What is it?

FrownyFrog

yes, I'm firguring it out atm

it's if. do. end.

Zacharý

04:12

APL wins again, one byte suckas (or three)

I probably should learn J sometime

FrownyFrog

tio.run/##y/r/…

Zacharý

...

FrownyFrog

It's 45 bytes in one line

Zacharý

APL > J

user202729

@Anush A possible way is to use stack snippet. Although I have not yet figure out how to use markdown inside.

Also, you can view markdown source of wikipedia pages.

FrownyFrog

05:13

tio.run/##y/r/P81WT0E7wcQqwdjKQU9Dw0jLyDZa004vNlXPUMfISvN/…

tacit, 27 bytes

I guess ((2*2=[)>.]e.1,2:){+,4,3: is the golfiest

Zacharý

05:28

Still sad

FrownyFrog

only the 27-byte one truly branches, the other 2 always do the addition

Zacharý

J > Perl tho.

Dennis

3+4ḟØ½$?@2Ƒ?

Adám

06:08

@FrownyFrog ↑(1 2)(3 4 5)

FrownyFrog

thanks

Adám

@FrownyFrog Also, APL has a full-fledged prototype system, so it will use the top left element as fill but with all numbers set to 0 and all characters set to space. Oh, and mixed-type arrays are allowed.

FrownyFrog

got it

New Main Posts

0

Q: To find 3 nearest values that are fully divisible by an user given value

Code challenge - i want to get 3 nearest values to an value given by user but that 3 nearest values should also fully divisible by the user given value. Eg: Input=30 Output=60,90,120. Also, i want to get 3 more values but which are nerest and lower than user value but which can divide the user va...

code-challenge

Adám

@FrownyFrog In fact, it goes all the way: If the top left element is (or contains) an instance of a class that has a niladic constructor, it will fill with (arrays that contain) new instances of that class!

Leaky Nun

06:19

hi @Adám

Adám

@LeakyNun Hello Kenny.

@FrownyFrog 4[^:(1 2 e.~])3[^:(2=[)+ works as well.

FrownyFrog

nice one

Leaky Nun

@Adám do you say oygn or ayin?

אויגן / עַיִן‎

Adám

@LeakyNun oyg (singular), oygn (plural).

FrownyFrog

@Adám it doesn't work

the arguments to ^:(...) are not the original ones

Adám

06:25

@FrownyFrog Oh, right, of course.

FrownyFrog

f=. 4 :'3[^:(x=2)4[^:(y e.1 2)x+y'

Adám

@FrownyFrog bet that requires pre-set vars.

FrownyFrog

@Adám costs 5 bytes

4 means dyadic only, 3 means you have to specify both cases

or something like that

you can specify only the monadic case, but if you don't, you need a : in the beginning, I think that's how it works

actually I have no idea

tio.run/##y/r/P81WT8FYwcqAy4rLODrOSqPC1kjTBMSoVEjVM1Qw0qzQruTS/…

yes, a single : lets me omit the monadic case

but it doesn't work in a one-liner

Adám

06:45

@FrownyFrog J's explicit notation is horrible, imho. Dfns are so neat in comparison, and even tradfns are not that bad.

FrownyFrog

can't argue at all

Adám

07:11

@FrownyFrog +`4:@.(1 2 e.~])`3:@.(2=[) truly branches

FrownyFrog

amazing

+`4:@.(e.1,2:)~ saves a byte

New Main Posts

0

Q: X Steps Forward, 1 Step Back

Here the first 100 numbers of an easy sequence: 0,1,0,2,1,4,3,7,6,11,10,16,15,22,21,29,28,37,36,46,45,56,55,67,66,79,78,92,91,106,105,121,120,137,136,154,153,172,171,191,190,211,210,232,231,254,253,277,276,301,300,326,325,352,351,379,378,407,406,436,435,466,465,497,496,529,528,562,561,596,595,63...

code-golf number sequence integer

Adám

07:37

@FrownyFrog OK, but only because plus is commutative.

New Main Posts

1

Q: New Password Idea: Word-walker

I thought of a new way to generate my passwords, and even though it's probably not very clever in the long run, it could still make for a fun code-golf. Taking a string of words, the password is generated thus: Pick the nth character in the nth word If n is larger than the word, continue count...

code-golf string

ASCII-only

@NewMainPosts :| I can directly port this but IDK if I should

Erik the Outgolfer

@Adám yeah, I did edit that in just before the time limit, finals really do corrupt your memory

@Adám the last 6 bytes are redundant...

user202729

I'm reading grammar.y (SPL source code) to see what's inside.

Adám

07:53

Ah, right. But good practice, imho.

user202729

@ASCII-only Explanation added (hopefully)

(too bad that SPL doesn't recognize */ as one token)

New Main Posts

09:02

0

Q: Alphabet Position Finder

Inspired by the Codewars Kata. Your goal is to take an input string such as this one: "'Twas a dark and stormy night..." and return a string containing the position of each character in the alphabet, separated by spaces and ignoring non-alphabetical characters, like this: "20 23 1 19 1 4 1 1...

string number

Anush

@user202729 I would love to know how to use a stack snippet for this

hi all

oh dear.. maybe I set a challenge too hard for ppcg?!

user202729

@Anush I just have other things to do...

Anush

@user202729 even worse! :)

@user202729 are you busy for a while now?

user202729

Busy with other challenges.

Anush

ah.. that's like saying you are busy with your other girlfriends/boyfriends :)

(that was a joke of course!)

user202729

09:16

Another problem with the stack snippet is that the content can only be displayed inside the snippet.

Just like the leaderboard snippet.

Anush

@user202729 hmm ok.. in principle the process of translating the wiki to ppcg format should be automatable I think

I mean all you do is download the svgs, convert them to png and add them .. but you need to work out the right scale and I don't understand how to fix the vertical centering

user202729

You can get the source Markdown by clicking "edit" on the wikipedia page.

Like this?

ASCII-only

@user202729 it's not Markdown...

Anush

I see abc <-- render as mathjax
**markdown does not work**
you have to use html instead

ASCII-only

It's WikiMark

user202729

09:21

@ASCII-only Somewhat.

Fine, but what's inside <math>..</math> is latex.

So just find a markdown/something to html compiler.

@ASCII-only You know any wikimark to html compiler?

Anush

I wonder if this could be turned into a PPCG challenge

that actually also helps PPCG

user202729

Doesn't sound very interesting...

Anush

some people get a hit from being helpful :)

user202729

Let me try... First, copy the HTML source from Wikipedia.

user202729

09:38

for(var a of $('.mwe-math-element'))a.textContent='$'+a.firstChild.firstChild.getAttribute('alttext')+'$';copy($('.mw-parser-output')[0].innerHTML)

(that converts every math image into latex formula with $ wraparound then copy the page)

ASCII-only

@user202729 inspect element...

user202729

@ASCII-only I end up using that.

New Sandboxed Posts

0

A: Sandbox for Proposed Challenges

Hexarun! king-of-the-hill javascript work-in-progress Hexarun (stylized as Hexarun!) is a simultaneous game with complete information. Hexarun is intended for a minimum of three (3) players. Overview Up to twelve (12) players start on a regular hexagonal toroidal board with a number...

user202729

09:56

@Anush This should look good now (except the font)

@Anush Besides, we have serious contender rule. "Answers must be a serious attempt at the winning criteria"

2^(2^(n-b))×2^(2n-b)×poly(n) doesn't sound very serious for a fastest-code challenge.

Anush

@user202729 It is a serious contender if it's the fastest, surely

@user202729 Nicely done!!

user202729

@Anush Although the previous SPL quine is the smallest, nobody would call it a serious contender for code golf.

Quote Mego:

@Anush (fastest possible, not fastest posted)

still finding the quote

Anush

I don't fully understand that.. the words "serious contender" must refer to being a serious contender for winning under the stated rules. So all that matters is the relative speed, not the absolute speed

user202729

> [...] if the only way a submission could win a challenge [...] is if no other solutions were posted, it's almost certainly not a serious contender [...].

Yes, the relative speed matters.

Anush

@user202729 could that snippet have been made automatically do you think? That is did you have to apply any judgement to make it?

user202729

10:03

@Anush I just have to select the correct part.

Anush

or could a script have done it in principle

user202729

But... having a very slow solution makes people doubt about whether their is a serious contender.

Anush

@user202729 may be but also might make other people more optimistic about their chances

user202729

@Anush Also that.

(last time I posted the second worst solution, and after that, people all post their better solution. Can't remember when)

Anush

there may be a smart solution for b = 1 of course

@user202729 oh that's interesting

maybe I should have asked a code-golf question for b = 1 first, to warm people up

or just a question about b = 1 :)

user202729

10:11

Still very hard...

Anyway. I will just post the first answer to encourage others.

Anush

thanks!!

for b=1 the sequence for n=4...9 seems to be 4, 6,12,20,40,70 which is oeis.org/…

@user202729 will your code work for n=10, b = 1?

user202729

Generate all distinct substrings of a length is hard enough...

Still coding.

This is fastest code, I have to find a way to make the code as fast as possible, to my ability.

Mayube

11:42

@cairdcoinheringaahing did we ever actually find a use for BrainFlump? I don't think it's ever actually been used, looks like the TIO version doesn't even work

user202729

11:54

If (a) is a subsequence of (b), then (b) is a [...] of (a)?

ASCII-only

@user202729 supersequence

apparently

user202729

That sounds really weird.

Anush

I think it's right though

user202729

12:10

Debugging just takes forever...

Anush

true!

that's why we are all supposed to code in Haskell so it takes forever to write the code in the first place :)

user202729

Only if you're not familiar with Haskell.

(yes, I'm not familiar with Haskell too)

Mr. Xcoder

Not planning to post yet, but review request: codegolf.meta.stackexchange.com/a/16451/59487

Anush

typo "And example of a concave matrix:"

user202729

@JoKing (what have you done? If that's on the constants part I'd rather working on that myself, you may post your own answer if you want)

And I have not read it.

I don't want spoilers.

@Mr.Xcoder I think the specification has bugs.

Mr. Xcoder

12:27

@user202729 Could you please be more specific?

user202729

Your specification implies that for a concave matrix,

all elements in column 1 are larger than all elements in column 2, while 4<5.

Mr. Xcoder

Indeed, you are right.

I'll fix that

user202729

For n=2, any matrices that are not flat are both concave and convex.

ASCII-only

@user202729 spoilers - lower char codes

user202729

Wait, but...

Last time I check ! gives more bytes than ..

Wait...

I think I misremembered?

ASCII-only

12:32

@user202729 who knows. ! does have the lowest ASCII code though (excepting space)

user202729

Wait but...

Well...

I think things changed when I subtract everything by 31.

Nope.

Now I have absolutely no idea why I used . instead of !.

Must be some silly mistakes.

Jo King

13:05

@user202729 Mostly changing the print sections yeah, but as pointed out, ! is smaller than . and : (for scene declarations)

Also, there were a couple of superfluous bits like You be you in there

user202729

13:46

And that's after I remembered to s/you be a/you/.

Adám

@Dennis Can you grant 354613 explicit write access to 52405?

Dennis

@Adám Done.

user202729

@Anush My program doesn't even work for n=7,b=1.

Adám

@Dennis Thank you so much.

user202729

(work = run fast enough)

I need better algorithm.

Anush

13:57

@user202729 ah...

user202729

The problem is I can't think of one.

If each sequence of length n-b is the subsequence of exactly two sequences of length n then it's an instance of the vertex cover problem (which is already NP-complete)

Anush

hmm... I must have been much cleverer in 2013 :)

@user202729 can you use a library to do the np-complete parts? Some of them are quite fast

e.g. igraph

user202729

@Anush I still could not reduce it to a well known classical problem.

That is "if".

That is not just true.

Transcript for