cmc: shortest apl expr to find how each element of a matrix ranks among its distinct values? in other words, replace all occurrences of the smallest value with 0+⎕io, the second smallest with 1+⎕io, etc

the matrix can be assumed to be numeric and real (not complex)

@ngn So something like ⍴⍴,⍳⍨⊂∘⍋∘∪∘,⌷, but shorter?

@Adám I was hoping for something shorter that I might be missing. Currently I have c[⍋c←∪,b]⍳b←⎕, it's for this challenge

that doesn’t work

@FrownyFrog why?

4  9 4  9
4  9 4 11
6 12 8  6
---
1 2 1 2
1 2 1 7
7 7 7 7

the tio link

Yeah, mine doesn't work.

It should be ⍴⍴,⍳⍨∘∪⊂∘⍋∘,⌷,

I had the ∘∪ in the wrong place

in that challenge we can be sure that each distinct element has 5 copies in the matrix - maybe I can take advantage of that...

haha! used ⍋⍋ floor-div by 5 and stripped another 3 bytes :)

@Adám any word on a fix for my ws issues yesterday?

@nathanrogers Ah, no I forgot I wasn't going in to the office today, but hang on, let me IM him.

@nathanrogers Try with HKEY_CURRENT_USER instead.

awesome

thanks @Adám

@nathanrogers Thanks forwarded to Chief Architect.

for the record

there should really be more media pertaining to apl/j/k

it seems the only media out there are either teleconferences or just really really brief short videos.

and generally with poor audio quality

@nathanrogers that's.. 2004. More recent APL videos can be found here, though I don't know what exactly are you expecting from APL videos

anything interesting, and with better audio than cranking the volume, can still barely hear what the speaker is saying, and also more than just the same presentation over and over. I'm sure many new comers to APL stumbled accross krombergs life in APL, which was exciting and interesting

then there's the interesting history of the APL family of languages, and there's media of the history or people involved

i hadn't seen this page though

what's more is, like videos of ken, I can't find a single video with Arthur whitney. I'd love his thoughts on the langauge, and his ideas that lead to K, and just some general commentary on the design decisions that were and are involved in these languages.

There's also this which gives an overview of APL

seen that already

presented several times at different events

more stuff like that, or youtube.com/watch?v=a9xAKttWgP4&t=75s

@nathanrogers You know about dyalog.tv?

@nathanrogers one interview of arthur's: queue.acm.org/detail.cfm?id=1531242

yes about dyalog.tv

i recognized you from there :P

@nathanrogers How about APLTrainer?

neato

i think I have seen that before, but didn't watch anything there.

@nathanrogers Btw, we're sometimes struggling to come up with good subjects for our webinars. If you have any subjects you'd like to see, tell us: webinar@dyalog

@nathanrogers On video?

yeah, stories. why this word, why this symbol, conversations had between this guy and that, people who made decisions about design of the language that had an impact on its use, proliferation, stories and tales and history

and anything for non-domain experts

I'm a software engineer, and I find any discussion I have on the subject of apl family languages to be completely wasted, as most engineers are taught that succinctness = code smell

so demonstrating practical solutions to software problems would be great, as most of the content I see are academic problems or domain problems, i.e. linear algebra, capital E engineering type stuff

web servers/services, ui, other data structures

I'm sure all that stuff is there, as you guys just linked a lot of things that I haven't seen before

but transliterating solutions in some other language to APL would be beneficial to non Engineer, more computer sciency software developers

@nathanrogers simply transliterating things to APL from "regular" languages will almost always result in very bad APL code, giving a very wrong impression

@nathanrogers Wow, great feedback! I'll make sure the powers-at-be see this.

there are also great demonstrations of APL, but not many explanations. Like the life video. when I first saw it, it seemed as though he was blazing through the solution. as I understand more and have watched it since, it makes sense and it doesn't seem so wild. but making videos aimed at people who don't get it with more explanation would be gravy

@dzaima I think he means re-implementing the APL way. Notice his liking the jugalbandi.

@dzaima, but the video I linked above

this one, they do exactly that

and show "this is how we would solve this problem in APL

it's a fun video with good humor, but also powerfully demonstrating how the tools we use to think encourage different natural solutions in a given language

and that a more flexible powerful language reveals more direct solutions

@nathanrogers ah I googled "transliterate" and it gave me "write something using the closest corresponding things" :|

@Adám yes, exactly :) didn't see your comment there

the jugalbundi was great. a not-so-simple OOP, even FP problem that takes some thinking, but that still has a direct solution in APL. I'm understanding APL, I can read most of what I see, even trains since I was trying to understand tacit J for a bit, and I think tacit APL is much easier to look at... but I'm having difficulty in solving problems. I'll see a problem,

like "bracket balancing"but think to myself, "well this uses looping and stack pushing and popping... that doesn't seem like a problem well suited to APL" because I don't know the tactic that would be used in APL for this kind of problem

Got to go — the sun is setting -⊖-

haha :P

J trains are pretty much the same

if there is an odd number of terms, they are equivalent in both monadic and dyadic case. For even number the monadic case saves you exactly one byte, J: (%+&1), APL: (⊢÷+∘1)

the dyadic case is like this, APL: ÷∘(2××⍨), J: (%2**~)

only the last term is actually dyadic out of the whole "dyadic" train

the leftmost, that is

@FrownyFrog isn't one of the *-s dyadic too? so, in J x(f g h i)y ←→ x f g h i y?

how does one convert char to int?

specifically ascii char codes

@nathanrogers ⎕UCS

would anything of value be hurt if I made dyadic ∪ and ∩ return unique elements? (so like (∪∪)/(∪∩))

how do you preserve the shape of an array while filtering it?

I want values where true, retaining the shape

er wait.

0@? probably?

@nathanrogers what would that even be like? What would you expect from filtering out odd numbers from 2 2⍴⍳4?

⍸4=list

so there's a list where some are 4

and I a list of indices where 4=

I want those to be 0

and retain the rest of the list

@nathanrogers oh so not filter, but replace with 0

without modifying the list

yeah

{0@(⍸⍺⍺¨⍵)⊢⍵}

wow

splain plox?

@nathanrogers ⍺⍺¨⍵ - execute ⍺⍺ for each of ⍵; ⍸ - get places where that was truthy; 0@ replace those places with 0; ⍵ in ⍵. That ⊢ was out of laziness, that expression could as well be {(0@(⍸⍺⍺¨⍵))⍵}

as a side-note, I've found way too many functions that work fine on multidimensional arrays that I didn't expect to while testing stuff for implementing my APL (including that ⍸)

Laziness = ingrained code-golf.

  s←'{()}{[()]}{}{}{}{}'
     open←'({['
  close←')}]'
  {g←{0@(⍸4=⍵)⊢⍵} ⋄ 2,/(g open⍳⍵)+g close⍳⍵}s

I got this going on

what is the next step

and indexing off of that

but then I kept getting wonky results when recursively applying this

@nathanrogers what's the goal - check if parentheses match?

becausyes

i don't want the solution

I just don't know where to go next

in the pairwise enlist

perhaps I can get outer values for where cell 0=

o=+/cell

er... no 0=/ cell

because if 1 0 1 then it holds true that I can 1 take left cell and -1take right cell

but I'm not sure how to do that

and in another case where 1 0 0, then I can simply take 1 on the left cell, and 0 0 1, take 1 on the right cell

perhaps there is a way to do a 3-wise reduction and only return where ~0¨=/each cell

i'll keep trying

no, because in the case of 2 1 1 2 3, the second and third cells would be 0 0 1, 1 1 1

and I still get that stupid 1 back

{g←{0@(⍸4=⍵)⊢⍵} ⋄ (2+/(g open⍳⍵)+-g close⍳⍵)~0}s

i mean, this looks really promising, but then it doesn't work recursively X(

  {g←{0@(⍸4=⍵)⊢⍵} ⋄ ∊(-/¨a)↑¨a←2,/(g open⍳⍵)+g close⍳⍵} s

getting there, but I'm not sure where the boolean expression comes in. hmm

how do you take the sign of something?