Can someone explain what precisely is happening with the encoding behavior of box characters here? When I get them from a. by index, 2 2 shapes them correctly. But if copy/paste that output back, I need to shape by 2 6 to get the same result.

Presumably something to do with the number of bytes in the characters, but I don't understand exactly what.

@Jonah J is using control characters to encode box drawing characters. But these characters don't display as boxes in the IDE (because they're not), so it prints out different characters. J string literals are UTF-8, and the actual box drawing characters displayed are three bytes long.

@Marshall "so it prints out different characters" So J is internally doing conversion from control characters representing the boxes to the actual box characters? Where in J does this actually occur? Also your list sentence implies that J strings are basically just byte arrays and not "character aware" -- is that correct?

@Jonah I'm not going to dig through the J source for that conversion—it might even be in the IDE. As for J's relation with Unicode, see u:. It's not good.

@Marshall Thanks.

<moon-child> anybody have thoughts on a good glyph for a cap (as in capped forks)?

<moon-child> overstruck ⊣ seems appropriate, but I don't think unicode has it

<moon-child> (or crossed out|inverted ⊣, indicating the absence of a left argument. Maybe there should be a colourapl a la colorforth.github.io)

@DyalogAPL you mean a monadic fork right

hmm

Was trying this for the t square thing

<moon-child> @Razetime yeah, like jsoftware.com/help/dictionary/d502.htm

<moon-child> there are ⍅ and ⍡, which are somewhat warm, so I'm holding out hope

<moon-child> actually ⍆ might be good, showing the (uni-directional) direction of application

⍆ does look good, but it's very detailed

<kritixilithos> for this bounty codegolf.stackexchange.com/q/217690, i think a port of chrispn's k answer might prove shorter than the recursive apl one

@DyalogAPL razetime.github.io/APLgolf/… this was my original attempt, pretty similar

maybe k is just more optimized for it

<kritixilithos> @razetime, can you shorten ⍸¨↓⍵∘.=⍺? what about (⊂⍬)? try also for ⍸~(≢⍺)↑⍸⍣¯1

<kritixilithos> @razetime what apl does that website run - extended?

okay, i'm finished with my load balancer for isolates

bal←{{(¯1↓0,+\⍵)+⍳¨⍵}⍣(⍺>0)⊢(1↓q),⍨(⍵-+/q)+⊃q←(|⍺){+/↑(⍺⍴⌈⍺÷⍨≢⍵){(-⍺)↑¨↑∘⍵¨+\⍺}⍵,(⍺-⍺|≢⍵)⍴0}({⍵↓⍨-⊥⍨0=⍵}(⌊⊢+2÷⍨0.5≠2|⊢)⍵×⌽v÷+/v←⍺⍺¨⍳⍵)}

if i have a function with the O(n^2) complexity, and i want to run it in 4 different threads on the dataset consisting of 30 elements in ascending order with the same gap between them, then

      ¯4(×⍨bal)30
13 9 5 3

computes the dataset layout, and after i pass a positive core number to bal, then generates indices of element to be passed to the function

      ¯4(×⍨bal)16
7 5 3 1
      4(×⍨bal)16
┌─────────────┬────────────┬────────┬──┐
│1 2 3 4 5 6 7│8 9 10 11 12│13 14 15│16│
└─────────────┴────────────┴────────┴──┘

i used it with my PE72 solution:

* Benchmarking "+/{_←'pco'⎕cy'dfns'⋄+/{⍵××/(1-1÷⊢)¨∪3pco⍵}¨⍵}IÏ 1+16(×⍨bal)999999"
             (ms)
 CPU (avg):  1256
 Elapsed:    8677

@DyalogAPL you can choose from 17, 18, dzaima, extended in the options

@DyalogAPL I'm not sure how to shorten those yet

If I have any more ideas I'll post links

It's a bit hard to find where to split on for explanations.

APLgolf now has a simple explanation assistant.

@Razetime nice solution to t square

@Razetime ⍸¨↓⍵∘.=⍺ → ⍸¨⍺∘=¨⍵; no need for the ⊢ after ⍸⍣¯1; in extended (⊂⍬), could be ⍬⊂⍛,

⍸~(≢⍺)↑⍸⍣¯1 → (⍳≢⍺)~

@dzaima woah

@rak1507 I decided to actually read the question lol

lmao

@dzaima this saves a ton

@dzaima 43 bytes: {'*'@((⍳≢⍺)~⊃{⍵,⊃⍺/⍨⍺>⊃⌽⍵}/⌽⍬⊂⍛,⍸¨⍺∘=¨⍵)⊢⍺} in extended

not sure what "blank char" suffix is aaaa

@rak1507 why answer in dyalog classic?

Oh for some reason it didn't work in extended

Couldn't be bothered figuring out what the issue was

Can someone explain to me why enclose returns an array with shape zilde? (Dyalog APL 18.0)

      m ← ⍳5
      m
1 2 3 4 5
      ⊂m
┌─────────┐
│1 2 3 4 5│
└─────────┘
      ⍴m
5
      ⍴⊂m

sorry for the monospace on the question, never really used SE chat before ^^'

It's a scalar

@Martmists The result is a scalar. Maybe the APL Wiki page on the concept makes it clearer?

See also bound.

but why would it have an empty shape rather than simply a shape of 1

@Martmists A shape of 1 (,1 really) indicates that it has one axis of length one, but the result of Enclose doesn't really have any axes.

One strange result you would get if you made it a vector instead of a scalar is that the outer product (⊂'abc') ∘., 'ef' 'ghi' 'jk' would have shape 1 3 (the shapes of the two results together) instead of just 3. So you have this extra axis that appears and can get mixed into other arrays where you don't want it.

@Marshall i've never heard "bound" before. does it come from j?

@ngn Dyalog implementers used it regularly. I think it's an old APL term but I don't know when it originated.

@Marshall maybe it just happens to be the name of an internal function or field with no significance outside of dyalog's implementation?

@ngn I believe I've seen it in documentation for some old APL dialect, but I have no hope of tracking it down now. In any case, it's featured prominently in my outer product talk so there's some precedence now.

:)

a follow-up question about scalars: How does it decide when to apply certain functions and when to error? i.e. ⍳⊂10 works but ⍳⊂1 2 does not.

@Martmists ⊂10 is the same as 10 but ⊂1 2 is not the same as 1 2

@Martmists ⊂10 doesn't enclose 10

See floating arrays.

@Martmists try ⍳⊂,10

Alright, I think I get it

time to overhaul my code again lol

beware of ⍳,10 which will never be fixed

⋄ (⍳⍴∘1)¨⍳5

bot is taking the day off?

@rak1507 could be jewish :)

lol

@ngn oh no what's wrong with that

@rak1507 use userscript

effort

      (⍳⍴∘1)¨⍳5
┌─┬─────┬───────┬─────────┬───────────┐
│1│┌───┐│┌─────┐│┌───────┐│┌─────────┐│
│ ││1 1│││1 1 1│││1 1 1 1│││1 1 1 1 1││
│ │└───┘│└─────┘│└───────┘│└─────────┘│
└─┴─────┴───────┴─────────┴───────────┘

It's more profound with ⎕IO←0

      (⍳⍴∘1)¨⍳5
┌──┬─┬─────┬───────┬─────────┐
│┌┐│0│┌───┐│┌─────┐│┌───────┐│
││││ ││0 0│││0 0 0│││0 0 0 0││
│└┘│ │└───┘│└─────┘│└───────┘│
└──┴─┴─────┴───────┴─────────┘

yay userscript

@Razetime it should return the ,¨ of what ⍳10 returns

@ngn oh that does make sense

@Razetime ⍬⊂⍛, do you really need it to be a ⍬? is there anything else that could take its place?

maybe it's worth porting the newer ngn/k solution? idk, feels weird seeing a recursive solution win over in apl

I want to make a function that simply maps a given value a to a given value b, but it needs to be invertible

so when I give it b, it needs to give back a

what is the easiest way to do this?

it doesn't matter what it gives for other outputs that are neither a nor b

@rak1507 what are a and b - some values you've chosen? arguments?

some values that I've chosen

are they arrays? numbers?

Ideally it should work for anything, but in particular character vectors

this works

@dzaima -2

That's smart, thanks

came up with imo a pretty decent answer to problem 57 (@KamilaSzewczyk) dzaima.github.io/paste#009avftS71e5R71KNRz1dhgaPeuc/…

@rak1507 challenge: solve #66 without web searching :)

Challenge accepted :)