The APL Orchard

elliptic00

00:04

↑ v1¨ (v1 m)
┌→────────────────────────────────┐
↓ ┌→────┐ ┌→────┐ ┌→────┐ ┌→────┐ │
│ ↓1 2 3│ ↓7 4 1│ ↓3 6 9│ ↓1 4 7│ │
│ │4 5 6│ │8 5 2│ │2 5 8│ │2 5 8│ │
│ │7 8 9│ │9 6 3│ │1 4 7│ │3 6 9│ │
│ └~────┘ └~────┘ └~────┘ └~────┘ │
│ ┌→────┐ ┌→────┐ ┌→────┐ ┌→────┐ │
│ ↓3 6 9│ ↓1 2 3│ ↓9 8 7│ ↓3 2 1│ │
│ │2 5 8│ │4 5 6│ │6 5 4│ │6 5 4│ │
│ │1 4 7│ │7 8 9│ │3 2 1│ │9 8 7│ │
│ └~────┘ └~────┘ └~────┘ └~────┘ │
│ ┌→────┐ ┌→────┐ ┌→────┐ ┌→────┐ │
│ ↓7 4 1│ ↓9 8 7│ ↓1 2 3│ ↓7 8 9│ │
│ │8 5 2│ │6 5 4│ │4 5 6│ │4 5 6│ │

got it.. nice.

It seems to me not possible to use outer product ∘.∘

Adám

It doesn't make so much sense here. Outer product is all about combining two elements of arrays at a time. Here, you want to combine two functions at a time. It can be done using one of those methods I linked to, but it only complicates matters.

elliptic00

Yep, make sense. Now, we can see the power of APL, image try generate such tables on Java or Python..

Adám

@elliptic00 Like this: n←⎕NS¨⍬⍬⍬⍬ ⋄ n[1].F←⍉ ⋄ n[2].F←⌽ ⋄ n[3].F←⍉ ⋄ n[4].F←⊢ ⋄ ∘.{⍺.F ⍵.F m}⍨n

elliptic00

@Adám, I need time to digest it.. first... :).. awesome :)

it is not that bad..

⍬⍬⍬⍬ what is that?

Adám

Four empty lists.

elliptic00

00:16

⍴ ⍬⍬⍬⍬
┌→┐
│4│
└~┘

⍴ ⍬
┌→┐
│0│
└~┘
      ⍴ (⍬⍬)
┌→┐
│2│
└~┘
      ⍴ (⍬⍬⍬)
┌→┐
│3│
└~┘

where is 1 ?

⍴ (⍬)
┌→┐
│0│
└~┘
      ⍴ (⍬⍬)
┌→┐
│2│
└~┘
      ⍴ (⍬⍬⍬)
┌→┐
│3│
└~┘
      ⍴ (⍬⍬⍬⍬)
┌→┐
│4│
└~┘

Adám

Parentheses do not enclose.

So ⍴⍬ gives the shape of the empty list, which is 0.

elliptic00

⍴ ⍬⍬
┌→┐
│2│
└~┘

⍴ ⍬⍬
┌→┐
│2│
└~┘

⍴ 1
┌⊖┐
│0│
└~┘
      ⍴ 1 2
┌→┐
│2│
└~┘

rak1507

@elliptic00 tio.run/##PYzLCsIwEEX3/… not so bad :P

elliptic00

Ic,

@rak1507, nice, Python has very similar functions like Haskell. But Java, it might a bit more code, there is no zipWith in Java Lambda..

rak1507

yeah, doing it in java sounds painful

zip(*foo) is nice but ⍉foo is nicer

elliptic00

00:26

when come to this kind of question, APL is unbeatable :)

rak1507

yeah, especially if it was then extended to 3d or something

Adám

@elliptic00 Except by BQN: Try it!

elliptic00

wow, BQN is much more cleaner than APL..

Adám

Well, yes, the whole goal of BQN is to be a cleaner APL, based on half a century of experience.

rak1507

having an array of functions there is useful

Adám

00:30

And being able to switch role so easily.

elliptic00

cleaner and easier to understand.. I think Dyalog should take note.. what can they learn from BQN

do you guy think Dyalog is too conservative for the development APL ?

Adám

@elliptic00 Uh, much of BQN was originally developed in-house at Dyalog.

@elliptic00 You may want to read this blog post which discusses Dyalog's stance on moving quickly.

elliptic00

They should run the company like Google, FB, or spaceX, move fast, if something breaks, fix it..

Adám

@elliptic00 Are you serious?

elliptic00

00:47

:)... I mean they should build more tools around the array language ecosystem, so new user can try and test.. without any programming experience...

Adám

What kind of tools do you have in mind?

elliptic00

e.g. running apl as script language is no braining important for development. I'm not sure why take so long to do so. It might be technical challenging that I don't not understand obviously..

18.2 will be ..:)

Adám

Early 2022.

elliptic00

Yep..

BQN is completely open source language?

like C++, or Python?

Bubbler

Yes, there are multiple implementations and they're all open source so far

elliptic00

00:54

the best way to expose your language to more people is to let people to have fun with your language.

Adám

OK, so how do you suggest we do that with APL?

elliptic00

I remember no long ago, I try to download Dyalog APL (I have no idea what is RIDE. what is APL, what is array language at that time), But Dyalog ask me to register an account in order to download them.. I just give up... and late..... Dyalog allow ppl to download RIDE.. without any account... I think..

Adám

OK, so we fixed that, as you no longer need to register to download the unlimited product.

elliptic00

The document is too out dated in Dyalog website.., It is very hard to understand it for people like me with no experience.

xpqz.github.io/learnapl/tacit.html

Right now I use above site to learn APL.. it is very easy to understand..

The fact is there is not too many resource online.

Adám

@elliptic00 Anything in particular you found to be out of date? (I do recognise that what's on our website is neither welcoming, nor good learning material – our new website is in the works.)

@elliptic00 Wait, are you aware of the learning materials page on APL Wiki?

elliptic00

01:08

Yep, the APL Wiki is pretty good actually, I use all the time

APL is "hard" language for ppl who are from C++/Java/Python world at the beginning.

2

Adám

Yes. Much easier to teach to non-programmers.

2

elliptic00

Maybe..

APL is the language "make your think differently"

Adám

Yes.

PyGamer0

04:57

^ what? (on RIDE, pressed Ctrl+UpArrow in a session)

Jonah

06:39

@Marshall I was checking out BQN. The docs are nice and I especially appreciate this page: mlochbaum.github.io/BQN/doc/fromJ.html. One piece of info I'd love to see: "things BQN is better at" -- examples of specific problems that are awkward to handle in J but which BQN handles gracefully. If you have anything off the top of your head you can share here, that would be helpful too.

I did notice having ‿ instead of spaces for stranding allows you to avoid parens, which eg are required in J when a numeric rank operator is adjacent to a list. That one feels minor though, and I know there are deeper things BQN fixes.

milia

07:02

@elliptic00 it should be an easy language to anyone that had some training in mathematics, because I find that the language is like applied abstract algebra.

@Adám mlliarm.github.io/apl-in-bcn

RikedyP

07:29

@PyGamer0 Can confirm I can reproduce - probably some crud getting interpreted as Unicode when generating the error message, but I reckon must be related to Edit → Preferences → Shortcuts

Indeed, TL is "Toggle Localistation"

which is only relevant in the editor

I'll log an issue

Fawn Locke

I can also reproduce, Ctrl-Shift-J (JBK) produces a similar error

RikedyP

yeah and Ctrl+Alt+C

logged: github.com/Dyalog/ride/issues/802

Fawn Locke

Also big fan of your videos, Rikedy :)

RikedyP

@FawnLocke Thanks - I think I thould make more but combination of not sure what to cover and lack of time

or maybe I'm just making excuses ;)

advent of code seems like low hanging fruit at first, but it'll just be embarassing when it comes to the later problems

Fawn Locke

Haha, yeah. Content production is tough! I remember hearing about how John Scholes would labour over a single video for months.

And totally, the majority of us struggle with those later problems :)

Plenty of challenges out there to complete with the "ancient runes" lol. I preferred your more educational videos, on modern features & optimisation in APL (Love the font you use btw, currently using it system-wide)

xpqz

08:02

Advent of Code 2021 has begun, and it kicks off with an APL-banger. Right up our street.

Bubbler

+1 if you didn't use 3+/ in part 2 :)

milia

@xpqz looks interesting, thanks for the shoutout :)

not sure if I feel comfortable enough to join with APL. But then, it'd be a good opportunity to get better.

xpqz

@milia It's a perfect way to learn, trust me -- the ramp-up is gentle.

milia

@xpqz thanks :) I'll go with APL then, having your book side by side as a reference :D

xpqz

@Bubbler does 3+⌿ count? :P

Bubbler

08:06

Many people try using a new language for AoC to learn it

xpqz

@rak1507 can we use your leaderboard?

PyGamer0

@Bubbler so i dont get +1 :/

actually i feel that my program is dumb

PyGamer0

09:40

whats the best way to do something like (1 <op> 2) (2 <op> 3) (3 <op> 4) when i give <op> and 1 2 3 4

Fawn Locke

⎕←2,/⍳4

TryAPL

@FawnLocke
┌───┬───┬───┐
│1 2│2 3│3 4│
└───┴───┴───┘

Fawn Locke

Windowed reduce

aplwiki.com/wiki/Windowed_Reduce

⋄ ,{2⍺⍺/⍵}1 2 3 4 ⍝ Does this answer your question?

TryAPL

@FawnLocke
┌───┬───┬───┐
│1 2│2 3│3 4│
└───┴───┴───┘

PyGamer0

ah i am dumb :/

⋄ <{2⍺⍺/⍵}⍳10

TryAPL

09:45

@PyGamer0 1 1 1 1 1 1 1 1 1

rak1507

@xpqz hm? what? ofc

code is 656501-67f7cded if anyone else wants it

@Bubbler what'd you use?

Fawn Locke

You can use ⌺ and ⌽

Couldn't think of anything else

rak1507

presumably it's using the fact a+b+c<b+c+d is the same as a<d

Fawn Locke

yeah

rak1507

+/<⌿a[↑¯3 3↓¨⊂⍳≢a]

Fawn Locke

09:57

Nice :)

rak1507

1-⍨+/<⍥⊃∘⌽⍨⍤⊢⌺4⊢a

Fawn Locke

Bubbler's: +/3∘⌽>⊢

rak1507

doesn't work in general

Fawn Locke

Yeah

Bubbler

Yeah, I said it's cheating in the other chatroom

rak1507

10:01

ah

Bubbler

10:28

julialang.org/blog/2021/11/julia-1.7-highlights

Package manager, newline in string literal, ndarray literal, apple silicon support

Julia 4 - 0 APL

rak1507

yeah... I wonder how julia is funded

milia

@rak1507 maybe, MIT ?

@Bubbler :'(...APL will get there eventually...

PyGamer0

10:56

should i share my apl solution here?

xpqz

Share as a link,

milia

@PyGamer0 for the advent of code ?

rak1507

@PyGamer0 yeah as a link

PyGamer0

ok

lemme make a gh repo

xpqz

I ported my APL solution to numpy. Iverson must be turning in his grave: gist.github.com/xpqz/151a8aa81f5333da9d3d75bb2bb48fee

2

Fawn Locke

10:57

lol

rak1507

*a,=map(int,open(0))
print(sum(i<j for i,j in zip(a,a[1:])))
print(sum(i<j for i,j in zip(a,a[3:])))

xpqz

Cheating.

PyGamer0

here

xpqz

My apl solution.

Fawn Locke

mine., There's an obvious couple character solution to part 1, but I'll just leave up my initial thoughts

rak1507

11:02

github.com/rak1507/Advent-Of-Code-APL/blob/main/2021/…

xpqz

My k version

rak1507

+/'3'd can be 3+/':d (n'x is removed in the latest version)

milia

@PyGamer0 for small things also gist is nice :)

Adám

12:39

@Jonah A recent example was an outer product ("table") of functions. See from this message and until I show the BQN solution.

Marshall

12:49

@Jonah Right, I've been thinking about extending the why BQN? page with some language-specific comparisons. Relative to J you get garbage-collected closures and namespaces, painless Unicode handling, the based array model (comparison), and somewhat better trains and combinators (no @ versus @: annoyance; 2-train is atop; ⊸⟜ are really nice).

Adám

@Marshall And proper, non-hacky, first-class functions.

Marshall

Right, I left that out since you'd already mentioned it.

And multi-line array notation is a nice benefit too.

klao

13:18

Hello all! I just started learning APL and want to do Advent of Code in it (as I see many ppl here are doing :)). I think I have a good understanding of the basics, but one thing I'm totally missing is how do I read in and parse a file?

I'll look at others' solutions, but what resources should I look at first? Say, how do I read a file with a bunch of numbers per line into a vector of (boxed) vectors?

Adám

@klao Welcome! Which APL are you using?

klao

These are my solutions for today's AoC, btw:

day01a ← +/2</⊢
day01b ← +/2</3+/⊢

@Adám Dyalog

Adám

@klao If each line has the same number of numbers, you should be able to use ⎕CSV.

@klao Btw, when you have some minutes, have a look at apl.wiki/APL_Orchard#Features for some tips on this room, including how to format code properly.

klao

@Adám Will do, thanks!

Adám

@klao Isn't there only one number per line?

klao

13:25

@Adám In today's problem yes, but who knows what tomorrow will bring. Want to be prepared :)

Adám

Ah, I see. For general file reads, have a look at ⎕NGET. Also, are you familiar with APLcart?

klao

@Adám I have discovered the APLcart, yes. It is a bit intimidating, though. :) But yeah, I had some fun decipher a few things from it. I haven't though to look for i/o thing in there…

Adám

Well, now you have :-)

rak1507

13:45

@klao ⍎¨lines

klao

@rak1507 Thanks, I found it in your gist above!

@rak1507 This wouldn't work for negative numbers, right? I have to replace minuses with high minuses first

rak1507

yea, because ⍎ is exec and '1 2 3 4 5' is a valid array it'll work on that too

@klao true, for that ⎕CSV might be better

xpqz

14:04

@klao This is how I read and converted the AoC data today: ⍎¨⊃⎕NGET'../d/1'1

klao

@xpqz Thank you! Same as @rak1507

Adám

But it is literally numbers in a text file, right?

Fawn Locke

This is the input

Adám

@FawnLocke You can't link like that.

xpqz

@Adám Yes.

klao

14:07

@xpqz Btw, why do you use 1⊥ instead of +/ in your solution?

Fawn Locke

I can't?

klao

@Adám Thanks, found it now!

Adám

@FawnLocke No, it gives Puzzle inputs differ by user. Please log in to get your puzzle input.

Fawn Locke

Ah okay.

xpqz

@klao Habit. But it's faster, and easier in tacit phrases, as it's not using an operator.

klao

14:09

@xpqz Interesting. And it has a left argument, which makes trains easier. Will remember it.

Patrick Villela

17:14

@Adám Hey @Adám! I'm sorry I didn't reply earlier, I'm actually not familiar with the StackExchange Chat.

I am indeed interested in APL. I'm currently attempting AdventOfCode in APL and in C (mostly to avoid being extremely frustrated hahaha)

xpqz

@PatrickVillela I can recommend xpqz.github.io/learnapl :)

Fawn Locke

Yes, xpqz's book and APL Cultivation are the ticket to success :)

Patrick Villela

17:30

@xpqz Thanks @xpqz and @FawnLocke. I've actually started reading both resources, as well as Mastering Dyalog APL. I'm tending to continue using yours, @xpqz. I found it works better for me.

xpqz

@PatrickVillela That makes me happy to hear, thank you.

@PatrickVillela I gave a presentation at the Dyalog '21 conference talking about it: dyalog.tv/Dyalog21/?v=qrWCYPFfsCU

Patrick Villela

@xpqz Oh, that's you! That's awesome! I'm actually halfway through that presentation. I'm somewhat drinking from a firehose here, but that's really fun! Thank you and the people here for keeping APL alive!

Is there some place where we can discuss AdventOfCode improvements without spoiling it?

Fawn Locke

There's the Discord server, we've been posting solutions there.

xpqz

We used this chat last year, certainly.

Fawn Locke

https://aplwiki.com/wiki/Forums

Feel free to ask questions or for improvements and you'll get a hoard of people willing to show you how good at APL they are :P

Patrick Villela

17:42

Hahaha thanks a lot y'all. I'll go there!

Adám

18:03

@PatrickVillela Welcome. No pressure. If you're not so familiar with SE Chat, have a look at aplwiki.com/wiki/APL_Orchard#Features

You'll find that the crowds in the APL Orchard and on the APL Farm have some overlap, but the culture differs.

kimmolinna

20:37

@xpqz I don't understand your solution to the part two. Why don't you need to take 3-windows sum?

kimmolinna

20:55

@xpqz Now I get it. Two element are the same ones so you can compare just the different one. This is a really nice way to learn from wiser people. ;)

xpqz

21:53

I can’t really take credit for that insight - @rak1507’s Python code above does the same.

rak1507

22:40

have to say I'm surprised how few people have used the fact a+b+c<b+c+d is equivalent to a<d: in APL 3+/ is so easy I just did that anyway, but in other languages I've seen people building up a list of slices with for loops etc when it would be much simpler to do it the other way

Transcript for