The APL Orchard - 2021-01-11

« first day (1447 days earlier) ← previous day next day → last day (1504 days later) »

code_report

02:20

APL Video Solution Walkthrough of AdventOfCode Day 1 Problem A: youtu.be/0RQFW6P1Tt0

Feel free to leave comments of code-golfed solutions

rak1507

I like the logo!

huh, I never actually noticed the input box in tryapl, I just always typed into the main bit

do you have a link to the paper you mentioned? @code_report did not know about that!

btw the dfn you used can be tacitified like this ⊃∘⍸2020=∘.+⍨

so the whole thing could be ×/⊢⌷⍨⊃∘⍸2020=∘.+⍨

I prefer ×/⊢∩2020-⊢ though ;)

code_report

@rak1507 jsoftware.com/papers/fork1.htm

rak1507

thanks!

code_report

This is the section that mentions it:
> Discussion of Hook and Fork

Hook. In combinatory logic one of the most useful primitive combinators is designated by S [Sch24]. Curry defines Sfgx in prefix notation to be fx(gx) [CuFeCr74]. In common mathematical infix notation this would be given by (x)f(g(x)) , which one can write in APL as xfgx , and this is the hook form (fg)x . The combinatory logician appreciates this form because of its great expressiveness: it can be shown that S , along with K , the constancy combinator, suffice to define all other combinators of interest [Ro50]. (The const…

rak1507

interesting that he uses (fg) not (⊢fg)

I can't get that continued fraction thingy to work

Ah it works with +∘÷

1 hour later…

Bubbler

04:02

@rak1507 Hook (2-train) is differently defined in that paper (and therefore J).

(f g)y there is equivalent to f∘g⍨y in Dyalog and x(f g)y there is x f∘g y.

Razetime

@code_report cool video

I suggest having a separate tutorial video so that you can focus more on how each solution is implemented rather than what the functions and operators do

rak1507

04:20

Eh, most of his audience probably don't know APL so it's probably a decent idea to explain stuff as he goes along

@Bubbler Right, got it

3 hours later…

Romilly Cocking

07:02

I've seen some discussion about implementing Dual (g^-1 f g). Is there any interest within Dyalog for implementing it? I need it lots.

Dyalog APL

07:14

<moon-child> it's in dzaima/apl and apl extended as 'under'

Adám

07:44

@RomillyCocking Absolutely. We just need a C developer who has the skills. Meanwhile, you can use my APL model.

@RomillyCocking Alternative simpler versions each supporting one of the two modes, but not both.

Adám

08:00

@rak1507 Nope, since ⌷ needs an enclosed list of indices :-( you need 1↑∘⍸ or ⊣/⍤⍸ or similar.

Bubbler

@Adám The first one doesn't look right. Shouldn't it be ⍵⍵⍣¯1⊢?

Adám

@Bubbler Oh wow, good catch. Pushed fix, this time with example to prove it.

Adám

08:57

@rak1507 The most beautiful thing about that solution is that you can then break out the 2020 as a left argument: ×/⊢∩-

rak1507

That is nice

xpqz

09:58

@MortenKromberg Even for us on MacOS?

Didn't think that could work, given how the interpreter is hidden away in a non-standard location.

Romilly Cocking

@Adám Perfect! Thanks.

Adám

@xpqz It can, if you can find the interpreter :-) Didn't you report that particular issue?

ngn

@code_report something doesn't add up here. the guy in the video says that a 3-train (fork) is an S combinator. the paper mentions a 2-train (hook) but a different kind of 2-train from dyalog's (atop).

Adám

@ngn See Bubbler's comment.

If I understand right, Dyalog's ⊣ f g and f∘g⍨ are S combinators.

rak1507

Sounds right to me

Adám

10:12

I don't know this combinator stuff, but I'd love to add such search terms to APLcart. Anyone willing to help?

ngn

@Adám yes, when applied monadically. easy to see when you write it down. but afaict a fork in general has nothing to do with S

xpqz

@Adám I did (and even if you can find the interpreter, it still doesn't work, due to how it is launched, as I recall).

Adám

@xpqz Ah, then your reply to Morten is appropriate. He's in a meeting right now, though.

xpqz

@Adám However, I was encouraged to see the demos of v19 at Dyalog -20, which seemed to suggest that the upcoming 'dyalogscript' will also take MacOS into account.

Adám

@ngn Not in general, no. Just this specific fork.

ngn

10:21

@Adám the fork in the video wasn't of this form

Adám

@ngn True, so I think @code_report confused S→specific fork→general fork→his specific AoC solution.

@xpqz Yeah, and in principle it should all be platform agnostic. Unfortunately, I doubt anyone at Dyalog actually uses a Mac often or thoroughly enough to notice when something ends up awkward there. We "dumbed" down installation on Mac, with bad consequences, it seems. Maybe we should have both dumb and full install packages for Mac?

RikedyP

@rak1507 that's pretty tasty

rak1507

It is kinda, his solution is (dfn ⌷ ⊢) which is like the S combinator

xpqz

I know I'm a small minority -- but loud :). My wish (as I've said many times before of course) is that dyalog sees itself as any other (non-Apple) dev tool chain on the mac (python, perl, node, ...) -- basically: treat the Mac as you already treat Linux. Separate the install of the interpreter itself, and RIDE. Allow Mac users to leverage the command-line in a natural way.

Adám

@xpqz Don't give up!

xpqz

10:29

Oh, I'm having too much fun to give up.

xpqz

10:46

@RGS you're a maths-geek -- is there a special name for a monotonic sequence of integers increasing by 1? E.g 3, 4, 5, 6, ..

rak1507

arithmetic sequence?

xpqz

Maybe yes. "Arithmetic progression with a common difference of 1" Doesn't quite roll off the toungue, does it.

rak1507

No, not really

Adám

@xpqz Sequential integers?

RGS

11:04

@xpqz I know no names other than the things proposed already. I'd call them consecutive integers starting at X.

chrispsn

11:49

@code_report nice! intrigued about whether 'where-at-depth' would work in k (right now it only works on flat lists)

ngn

@code_report a comment about terminology: primitives are not algorithms. primitives can be implemented using particular algorithms, or they themselves can be used to express higher-level algorithms.

for instance: "⍋ uses counting sort for small-domain arguments and radix sort in the general case", or: "⍋⍋ is the algorithm for inverting a permutation"

(correction: ⍋⍋ is the algorithm for "ranking", i.e. inverting the sorting permutation of something)

Tom Cockram

12:05

How can I remove the Dyalog keyboard? I uninstalled the program but the keyboard still exists. It doesn't appear in the list of languages when I go to settings > languages

dzaima

@TomCockram assuming Windows, it might be under some language as a layout

Tom Cockram

Yeah sorry, on windows.

dzaima

haven't used windows in a while, but iirc from the language selector on the taskbar you might be able to go to some separate settings page for keyboard layouts

Tom Cockram

@dzaima There isn't a 2nd language to remove... i.imgur.com/zzP373a.png

ngn

@rak1507 hm, maybe. it's like S with swapped x and y. to be fair, ⌷ is one of those primitives begging for an arg-swap :)

dzaima

12:11

@TomCockram that seems to be about languages/translation. there should be a separate settings page for keyboard layouts somewhere

Tom Cockram

@dzaima keyboards is the same situation i.imgur.com/e4t1zUW.png

dzaima

@TomCockram that's the same screenshot

Tom Cockram

ops i.imgur.com/uPM3NoS.png

ngn

do it the windows way - reinstall the os :P

dzaima

@TomCockram there should be a full settings page instead of a popup somewhere

Tom Cockram

12:16

I ran the uninstaller again and ticked uninstall IME and that worked.

dzaima

¯\_(ツ)_/¯

ngn

12:38

@chrispsn what is "where at depth"?

something like whereatdepth (0 0 1;0 1 0) -> (0 2;1 1)?

chrispsn

@ngn yeah

Feels a bit like odometer

ngn

12:54

@chrispsn ah, right, so the flipped form would be more efficient, (0 1;2 1)

chrispsn

@ngn what makes it more efficient?

ngn

@chrispsn less space wasted on headers, more opportunities for simd

just like odometer prefers (0 0 0 1 1 1;0 1 2 0 1 2) over (0 0;0 1;0 2;1 0;1 1;1 2)

chrispsn

Ahh I see - data will tend to have more points than depth levels

ngn

yeah, that's a better way of saying it :)

chrispsn

13:29

However it's arranged, it would ideally be fed directly to dot-indexing

ngn

14:20

@chrispsn x@/y should work too (@-fold with an initial value, where x is the n-dim array, y - the result from deep where)

ngn

14:31

@chrispsn i think dot-indexing is always rectangular (think of a[i;j;k] where i j k could be lists), so it would require an "each"

Adám

15:27

@Lyxal We had an email to you bounce. Can you check that your email account is working (e.g. inbox not full)?

Morten Kromberg

15:51

@xpqz There is no question that we started off on the Mac trying to follow what we perceive to be Apples view that their products are "appliances" and not "computers". The only early internal user was John Scholes, and he was allergic to complexity youtube.com/watch?v=qSVR4Z3DA24. We just agreed to hold a meeting to review our mac strategy.

xpqz

16:01

@MortenKromberg I can only speak for the places I work, but here at IBM, for example, all our software developers' machines are Macs (by default; you can opt out). The reason for this is that they are unix desktop machines that can be centrally managed and deployed. Whilst we deploy software we write as containers via kubernetes, we all write software locally in python, go, (or whatever), rarely leaving the command line.

Razetime

Trying to make an interpreter in js right now

Really missing the APL privilege

xpqz

@MortenKromberg It seems to me that Dyalog fits perfectly well into this pattern: the interpreter is analoguos to my /usr/local/bin/python. RIDE is analoguous to my ~/Applications/VSCode.

Which is what you do anyway on Linux, right?

Adám

@Razetime Like ngn/apl?

Razetime

16:26

@Adám nah, it's a 2D esolang I'd been wanting to make for a while now

Razetime

16:42

I could try it in dzaima/APL

Romilly Cocking

@Adám The Perlis Idiom list is here cpsc.yale.edu/sites/default/files/files/tr87.pdf - I doubt there is much new in it, as I think the Finns will have been familiar with it.

4 hours later…

Lyxal

21:06

@Adám that's odd, I'm able to send and receive emails normally on that account, and the previous emails came through just fine.

Adám

@Lyxal I'll try one right now. Maybe it was a temp glitch. Sent. Let me know if you get it.

Lyxal

@Adám It seems Vivaldi mail is acting strangely: I can get emails from my outlook account but not from my gmail account

Perhaps I should just provide an alternate email

Adám

@Lyxal Sure, you can respond directly back to Fiona with that.

Lyxal

@Adám do you mean from the vivaldi account or the new email?

Because Vivaldi mail is having difficulties sending things too

Adám

Just send from your alternate account, quoting the Vivaldi email address, for a tad of security.

Martin Janiczek

21:27

@RomillyCocking Wow. I'm reminded of the history of the language once again. I find it fascinating that these little tidbits have been there for ~50 years and are still relevant today. Not what usually happens eg. in Python world, seems to me.

rak1507

not much python was being written 50 years ago

Martin Janiczek

@MartinJaniczek Sometimes I'm sad I didn't get to work or even live at the time eg. Oberon, Smalltalk, Plan9 and so on were at their top. (EDIT: also Common Lisp)

@rak1507 That's the surprising thing to me (and something to get used to). That APL predates so many other technologies but is still relevant. (I guess that's due to it being "extended mathematical notation"?)

rak1507

Yeah, that surprises me too, and how ahead of other modern languages it is in some ways

Adám

If APL had only stayed ahead…

Martin Janiczek

Tangent: googling for the wiki page about APL, I saw this image and have the strongest G.A.S. right now for some ⍋ metallic trinket on my keychain

user image

(G.A.S. = gear acquisition syndrome :D )

Adám

21:33

That's not a bad idea. We should start selling metal APL glyphs on the website.

I don't like ⍲⍴⍸ though, but I wouldn't mind using a high-gloss steel ⍋ for my keychain.

Martin Janiczek

21:46

Hm not knowing better I found thestudio.com, and 100 of such trinkets would be $325.44. And probably not that high-gloss steel. Some kind of silver-like finish. I think I'll sleep on it :) But it's a neat idea, together with APL laptop stickers. (Yeah I'm one of those people)

IIRC, Aaron Hsu also has some T-Shirt designs with some APL idioms and snippets on them.

rak1507

yeah he does, I was tempted to ask for one for Christmas but I didn't in the end

He has a ⎕IO←0 tshirt, I'm tempted to make a ⎕IO←1 one just to annoy him if I ever saw him in real life

Adám

We always (except 2020, of course) hand out Dyalog "goodies" at the user meetings, but maybe we should start making "Powered by APL" and "I'm an APLer" laptop stickers.

⎕IO←0 and ⎕IO←1 and "⎕IO delenda est" are good too.

I have a giant ⊃∨∧⌊○⊂ orange-on-clear sticker on my silver laptop. Looks pretty neat.

Maybe some of RikedyP's ⊢>+⌿÷≢

Martin Janiczek

That's one of the T's bonfire.com/store/arcfide

Took me a second to understand the reference on your sticker :)

rak1507

I prefer things that have actual meaning to just how the symbols look

I like ⊢>+⌿÷≢, that one is clever

Adám

APL old-timers used to have I❤️APL. At the office, there's an ⍺*⎕ umbrella, and my father had one too.

How about ⌷≡⌽⌽≡⌷

Or >⍝ my only hope

Morten uses ⍒⍋ as his avatar.

user image

@MartinJaniczek I had a bunch of these ^ made at the local bakery.

Martin Janiczek

21:57

@Adám Ah I love it

rak1507

I don't get >⍝ my only hope

Adám

Star Wars: R2D2 projecting Leia

@MartinJaniczek It was in celebration of adding TAO (apl.wiki/TAO) which I was very involved in.

CMP: If Dyalog was to sell APL merch, what would you want to see the most?

rak1507

@Adám Ah, I thought it would be something like that, but I couldn't quite picture it

Adám

@rak1507 Have you not seen the APL puns?

rak1507

no!

Adám

22:03

Oh, hey we should have RO⌊L or RO⌊LOL T-shirts.

Or maybe ?O⌊L / ?O⌊LOL

Martin Janiczek

BTW before I forget it. News about the APL vs Go race at our company: I've prepared an end-to-end benchmark for the various implementations, measuring mem usage, CPU usage, the response times under load and not under load.

My colleague made his Go impl compliant with that spec (exposed a HTTP endpoint), so we finally were able to compare meaningful numbers (not his "3ms!" on his shiny Ryzen vs my "3ms!" on my laptop i7). Turns out his 3ms is 10-25ms on mine, and he has some contention in the goroutines so under load he has hundreds to thousands of ms (median 400ms) while 8x APL + nginx loa…

Adám

Any metric Go wins in (besides being respectable)?

rak1507

price

Adám

Fair, though I guess you pay for advantage on all the other metrics.

Martin Janiczek

Measurable metrics: not that I know of. I believe most of these can be fixed (startup time and memory usage by preprocessing similarly to how I do it; response time by fixing the goroutine contention), so I'm glad it at least forces the team to strive for better results now that they know they're possible

But I'm hoping they start thinking about sticking with the APL solution.

Adám

22:08

You could ask the people in charge how much they think such performance improvements are worth.

@MartinJaniczek Another interesting question is developer hours to get there (and how many spent already).

rak1507

Another would be if Martin's colleagues could pick up APL, I'm assuming it would be easier for him to work on the Go code than vice versa

Or maybe they'll just all get fired, you seem to be doing the job single handedly

Adám

Wouldn't be the first such story.

Martin Janiczek

@rak1507 I knew^H^H^H^Hstarted learning APL two days before I started writing the prototype and had it mostly ready two weeks worth of evenings later. They would just have to give it a chance and be a bit playful about it

rak1507

^H^H^H^H? what

Martin Janiczek

^H removes a character in terminal

Dyalog APL

22:11

<moon-child> ^H is backspace

rak1507

Ah right

Dyalog APL

<moon-child> (also: ^W remove word ^U remove line)

rak1507

Convenient!

Martin Janiczek

@MartinJaniczek Also I have a branch where I refactor the prototype into something manageable (namespaces and so on), but it's hard to strike balance

Adám

@rak1507 APLcart to the rescue ;-)

rak1507

22:13

lol

Martin Janiczek

@MartinJaniczek I don't know what "audience" to cater for, I surely can't explain and extract and name every little idiom. The code would IMHO become harder to use with all the indirection to named functions. (I can't believe I'm saying that.)

Adám

My father always said that two reasons (not necessarily the only ones) APL never became widespread was that 1) you'd only ever need a tenth as many programmers, and 2) IT department managers don't like having only a tenths as many people working under them.

2

@MartinJaniczek Wow, you've become a real APLer. Yes, the APL code is often easier to understand/read than needless abstraction into English names.

user image

(My father's handwritten logo.)

How about a T-shirt saying "+/ or \sum_{i=1}^N"?

Martin Janiczek

@Adám The sentence I wrote above would be utter falsehood in any other language. "We need abstraction!" "Structured programming!" Well, maybe not, or just a sprinkle. APL certainly rewards more "inline" style, yeah.

Adám

22:28

@MartinJaniczek "We need abstraction!" is one of the those things Aaron Hsu loves to shoot down.

Martin Janiczek

@Adám I saw the Obesity talk, seemed to touch on that :)

But I'm guessing Aaron's style is not the usual APL style either? I mean, it's even more dense than usual

Adám

His style changes a lot.

rak1507

user image

Adám

A lot of people feel tempted to define _Filter←{⍵/⍨⍺⍺¨⍵}

rak1507

what about something like this where B is × and C is ⌈

(my amazing latex based graphics design)

Martin Janiczek

22:30

Re other languages, not to dunk on it but I even thought about learning a bit of Futhark, but after seeing the definition of matrix multiplication I had to close the tab :(

Knowing it's 3 characters in APL, that is.

rak1507

It's 1 character in numpy ;)

Adám

@rak1507 Sure, but what is ∨.∧ then?

And ∧.=

rak1507

someone cleverer than me would probably know

Adám

Or ⌈.⌊

rak1507

I don't even know what that does

Adám

22:34

Same as +.× but with ⌈ and ⌊ ;-)

Martin Janiczek

Huh, I'm getting university flashbacks with minmax :)

Adám

On two vectors, it'd be the maximum pair-wise minimum.

rak1507

Right, I'm not sure when that would ever be useful

But I guess it could be

Martin Janiczek

Game theory

Chess engines and so on

rak1507

I'll take your word for it

Adám

22:40

user image

rak1507

23:02

That's good

Adám

Now try writing +∘÷/ in TMN!

Martin Janiczek

23:19

Wow, is that continued fraction? Very cool

Adám

Yes, and also -/ is alternating sum and ÷/ is alternating product.

⋄ +∘÷/25⍴1

Dyalog APL

@Adám 1.618033989

Dyalog APL

23:35

<moon-child> shouldn't exists be ∨/, and ∧/ would be for all?

Adám

moon-child: Of course. Typo.

Martin Janiczek

⋄ +/÷2*⍳100

Dyalog APL

@MartinJaniczek 1

Adám

@MartinJaniczek Maybe more interesting: ⋄ +\÷2*⍳10

Dyalog APL

@Adám
┌→───────────────────────────────────────────────────────────────────────────────────┐
│0.5 0.75 0.875 0.9375 0.96875 0.984375 0.9921875 0.99609375 0.998046875 0.9990234375│
└~───────────────────────────────────────────────────────────────────────────────────┘

Martin Janiczek

23:39

⋄ (⍟2)--/÷⍳10000

Dyalog APL

@MartinJaniczek 0.0000499975

Martin Janiczek

I'm enjoying this TMN alternative :)

Adám

⋄ 1∧+∘÷\15⍴1

Dyalog APL

@Adám
┌→───────────────────────────────────────────┐
│1 2 3 5 8 13 21 34 55 89 144 233 377 610 987│
└~───────────────────────────────────────────┘

Martin Janiczek

wait what?

Dyalog APL

23:41

<moon-child> ⋄(-∘1,'÷',⊢)¨2*⍳10 ⍝rational numbers!

Adám

⋄ (,÷∨)∘1○1

Dyalog APL

@Adám
┌→──────────────┐
│5419351 1725033│
└~──────────────┘

Adám

⋄ 5419351÷1725033

Dyalog APL

@Adám 3.141592654

Martin Janiczek

@Adám Curiously putting that into @dzaima's Android app gives me just iota. Is there some difference in behavior of and?

Adám

23:45

@MartinJaniczek I think dzaima/apl's \ is left-to-right.

dzaima

@MartinJaniczek ^ yep. dzaima/APLs slashes are all different from Dyalog

also there's no comparison tolerance, so floating point ∧ doesn't make sense (and probably should just be disallowed). Besides that, +∘÷/¨,\15⍴1

Martin Janiczek

I still can't get over how GCD and LCM can be used to get numerator and denominator of a float

⋄ (1∧2.5) (÷1∨2.5)

Dyalog APL

@Martin Janiczek
┌→──┐
│5 2│
└~──┘

Martin Janiczek

although yes I'm seeing the issues with eg. 3.3

Adám

What's the problem?

Martin Janiczek

23:55

The numbers get very big

instead of 33 10

Adám

⋄ (,÷∨)∘1⊢3.3

Dyalog APL

@Adám
┌→────┐
│33 10│
└~────┘

Martin Janiczek

A different algorithm would be needed :)

Hmm

Adám

@MartinJaniczek If you like all this, you may also like ⍣¯1

Martin Janiczek

I read about that!

Blew my mind just a little :)

Adám

23:59

Did you try the numeric solver?

dzaima

@MartinJaniczek your ⎕ct may be 0 (or otherwise a number too small)

Martin Janiczek

@Adám No no, can you elaborate?

Adám

@dzaima Works for me with ⎕CT←0

« first day (1447 days earlier) ← previous day next day → last day (1504 days later) »

Transcript for

Jan '2111

The APL Orchard

apl.chat ― Learn, teach, ask, code, golf, & discuss usage. See ...

apl array-manipulation j tips

join this room
about this room

00:00

06:00

12:00

18:00

all times are UTC

site design / logo © 2025 Stack Exchange Inc; legal

mobile