The APL Orchard

Bubbler

00:18

@Adám For aesthetics: It would be nice for Phase 2 problems to have an indication in the sidebar like Phase 1 ones (say a checkmark for "successfully submitted"). Having an empty space with no use is somewhat weird.

rak1507

^ agreed

Adám

10:48

Bye bye, APL2?

Razetime

11:05

looks like it

Marshall

12:49

@Adám Seems to really be just a change to a new vendor, who will presumably offer roughly the same pace of development as IBM. I requested to merge on the APL Wiki.

Adám

@Marshall I'm not sure. Let's wait a bit and see.

Marshall

@Adám If we're waiting, the default action would be to not make a page for Log-On APL2 yet, as it makes some pages harder to navigate. APL Wiki is not a crystal ball.

Adám

@Marshall True. Feel free to merge if you think that's the right thing to do.

Marshall

@Adám Okay, I'm merging.

Adám

13:12

@Marshall You'll roll back the changes on all the relevant pages, right?

Marshall

@Adám Yes, got stuck a little on the timeline of APL dialects. I'd written the page to include only languages that an APLer could start using immediately, so not things like J or Ivy.

I'm thinking now that APL Wiki probably should make that the definition of a dialect, rather than saying now that it doesn't define the term.

The only sort of borderline case is [[A+]], and we can clarify that it is an APL dialect.

Adám

@Marshall I added some less APLy entries because I saw ELI there. I agree that we should define the term.

Maybe anything that includes a superset of APL\360 core language, using the same symbols. "Core" excludes system commands and function, I-beams, bracket indexing and output lists.

> APL\360-ish languages only.

Heh, GMTA.

Marshall

@Adám That would rule out A+, and probably Rationalized APL, since they remove last-axis forms.

I think we'd prefer to just say that the syntax, primitives, and glyphs are very similar.

Razetime

13:46

How do I use ⌸ to group points based on X value?

Adám

@Razetime X as in left arg?

Razetime

right arg is a list of points

Adám

@Razetime As a vector of 2-element vectors?

Razetime

yes

might need to be a reduce..?

Adám

@Razetime ⋄ (⊃¨⊂⍤⊢⌸⊢)(3 5)(2 9)(3 1)(2 6)(2 0)(1 1)

Dyalog APL

13:59

@Adám
┌→────────────────────────────────────────────────┐
│ ┌→────────────┐ ┌→──────────────────┐ ┌→──────┐ │
│ │ ┌→──┐ ┌→──┐ │ │ ┌→──┐ ┌→──┐ ┌→──┐ │ │ ┌→──┐ │ │
│ │ │3 5│ │3 1│ │ │ │2 9│ │2 6│ │2 0│ │ │ │1 1│ │ │
│ │ └~──┘ └~──┘ │ │ └~──┘ └~──┘ └~──┘ │ │ └~──┘ │ │
│ └∊────────────┘ └∊──────────────────┘ └∊──────┘ │
└∊────────────────────────────────────────────────┘

Razetime

cool

Adám

14:25

CMQ: Any things you would have loved to have been told when you first started with APL, but not listed at abrudz.github.io/tips?

rak1507

the existence of the APL orchard

Razetime

@Adám testcase problem on problems.tryapl.org:

Adám

> problems

:-)

Razetime

Adám

@Razetime link?

Razetime

14:35

problems.tryapl.org/psets/…

rak1507

14:50

Is it just me or are this years phase 2 problems way harder than previous years on average? There are basically no 'trivial' ones (subjective ofc)

Razetime

      '"."'⎕R''⊢'this "is" a test'
this "is" a test

what am I doing wrong here

Adám

Missing + or *?

Razetime

ah yes

Adám

16:44

@Razetime Fixed. Thanks! It now allows any ordering, since the problem spec doesn't strictly say.

Razetime

great

Konrad 'Unrooted' Klawikowski

16:58

hey

rak1507

hi

Konrad 'Unrooted' Klawikowski

@Adám sorry for not being in there for a while, can we continue today where we left?

decided to f*ck school anyway so

rak1507

lol

Adám

@Konrad'Unrooted'Klawikowski Sure.

Konrad 'Unrooted' Klawikowski

if the offer of that personalized approach to APL is still avaible of course

Missed you man <3

Adám

16:59

I just have to look up where we were up to.

Konrad 'Unrooted' Klawikowski

I think it was making my own functions or somewhere around that

rak1507

do you know about this? dyalogaplcompetition.com/?goto=welcome @Konrad'Unrooted'Klawikowski

Adám

@rak1507 It hasn't been formally launched yet. We're still polishing some rough edges.

rak1507

@Adám true, even so he might like to give some of the problems a go

Konrad 'Unrooted' Klawikowski

@rak1507 I can give it a shot or take a look

we all know that @KamilaSzewczyk will try

rak1507

17:02

@Konrad'Unrooted'Klawikowski might be a bit tough for now but something to bear in mind

Adám

@Konrad'Unrooted'Klawikowski Yes, indeed. If I'm not mistaken, the last thing you did was to define {2 = +/ {⍵=⌊⍵}(⍵÷⍳⍵)} for determining if the argument is a prime.

@Konrad'Unrooted'Klawikowski Did we touch indexing into an array?

Konrad 'Unrooted' Klawikowski

nope

@Adám yes, I think we left there

Adám

OK, so the point is that as opposed to what many languages do, APL allows an array inside the square brackets: ⋄ 'APPLE'[1 3 4]

Dyalog APL

@Adám
┌→──┐
│APL│
└───┘

Konrad 'Unrooted' Klawikowski

⋄ 'SHE BELIEVED'[2 3 7 8 9 12]

Dyalog APL

17:10

@Konrad 'Unrooted' Klawikowski
┌→─────┐
│HELIED│
└──────┘

Konrad 'Unrooted' Klawikowski

makes sense

Adám

⍴ (Greek Rho) is "reshape", so we can reshape a vector (list) into a matrix: ⋄ 2 3⍴⍳6

Dyalog APL

@Adám
┌→────┐
↓1 2 3│
│4 5 6│
└~────┘

Adám

Now note that ⍴'APPLE'[2 3⍴⍳4] gives 2 3

The shape of the result of indexing into a vector is the same as the shape of the indices.

Konrad 'Unrooted' Klawikowski

@Adám so this example makes 3 colums and 2 rows of the output of iota 6? something like that?

Adám

17:12

Yes, it creates a 2D array.

This is ^^ a very important rule, that, if forgotten, can lead to some questions later.

Konrad 'Unrooted' Klawikowski

how many dimensions can arrays in APL have?

Adám

Dyalog APL supports up to 15 dimensions.

If you need more than that, you may be doing something questionable.

rak1507

lol

Konrad 'Unrooted' Klawikowski

just outta curiosity, how does a 15D array look like?

Adám

⋄ (15⍴1)⍴1

Dyalog APL

17:14

@Adám
┌┌┌┌┌┌┌┌┌┌┌┌┌┌→┐
↓↓↓↓↓↓↓↓↓↓↓↓↓↓1│
└└└└└└└└└└└└└└~┘

Adám

It is quite rare to need more than 4 dimensions, though.

Konrad 'Unrooted' Klawikowski

so it's strongly connected to matrix stuff?

Adám

Yes, it is. A matrix is just a 2-dimensional array.

Konrad 'Unrooted' Klawikowski

so every matrix adds additional dimension?

Adám

APLer's call the number of dimensions "rank", as in "tensor-rank".

So we'll speak of a list as a rank-1 array, a matrix as a rank-2 array.

Konrad 'Unrooted' Klawikowski

17:16

rank-1 is a vector right?

Adám

A single number or character is a rank-0 array, a.k.a. a scalar.

@Konrad'Unrooted'Klawikowski Yes, that's right.

Some people call rank-3 arrays cuboids, but at that point mentioning the rank is probably a good idea.

Konrad 'Unrooted' Klawikowski

got it

Adám

Remember IsDivisibleBy ?

Konrad 'Unrooted' Klawikowski

I think I do

Adám

I think we defined it as IsDivisibleBy←{0=⍵|⍺}

So now we can do ⋄ nums←(⍳5) ⋄ IsDivisibleBy←{0=⍵|⍺} ⋄ nums IsDivisibleBy 2

Dyalog APL

17:20

@Adám
┌→────────┐
│0 1 0 1 0│
└~────────┘

Adám

But what if we don't want to use Booleans to point at those numbers, but rather we want to extract them?

Konrad 'Unrooted' Klawikowski

⋄ nums←(⍳5) ⋄ IsDivisibleBy←{⍵|⍺} ⋄ nums IsDivisibleBy 2

Dyalog APL

@Konrad 'Unrooted' Klawikowski
┌→────────┐
│1 0 1 0 1│
└~────────┘

Konrad 'Unrooted' Klawikowski

wait

Adám

That's the division remainder. You want to check that the remainder is 0.

Konrad 'Unrooted' Klawikowski

17:22

I thought it would tell how much is 5/2 4/2 etc

Adám

It tells you how much is left over when doing an integer division.

Konrad 'Unrooted' Klawikowski

got it

Adám

Here's a new function (on Shift+I for "Indices where") which takes a Boolean array and returns the indices where there are 1s: ⋄ ⍸0 1 0 1 0

Dyalog APL

@Adám
┌→──┐
│2 4│
└~──┘

Adám

With that in mind, can you define ThatAreDivisibleBy? It should take a vector on the left and a scalar (single number) on the right, and return those numbers from the left that are divisible by the single number.

Dyalog APL

17:26

 @Adám
┌→──┐
│2 4│
└~──┘

@Adám
┌→────────┐
│0 1 0 1 0│
└~────────┘

@Konrad 'Unrooted' Klawikowski
┌→────────┐
│1 0 1 0 1│
└~────────┘

Adám

@DyalogAPL Stop! What are you doing‽

Konrad 'Unrooted' Klawikowski

⋄ nums←(⍳5) ⋄ IsDivisibleBy←{0=⍵|⍺} ⋄ ThatAreDivisibleBy ← {⍸(nums IsDivisibleBy 2)} ⋄ ThatAreDivisibleBy

Dyalog APL

@Konrad'Unrooted'Klawikowski {⍸(nums·IsDivisibleBy·2)}

Konrad 'Unrooted' Klawikowski

yeah, as I thought

Adám

This would return the indices of the numbers that are divisible by 2. We want the actual numebrs.

Konrad 'Unrooted' Klawikowski

17:29

here comes that rho thing right?

Adám

No, you just need to use these indices to index into the original list.

Also, remember to use ⍺ and ⍵ for left and right arguments.

Remember you can play around on tryapl.org

Konrad 'Unrooted' Klawikowski

yeah, doing that

btw, is RIDE in nixpkgs repo?

Adám

I don't think so.

Konrad 'Unrooted' Klawikowski

is there a repo for it on Github or somewhere? I can give it a try to add it to nixpkgs

Adám

github.com/dyalog/ride

Konrad 'Unrooted' Klawikowski

17:36

gonna play with it at the weekend or maybe even tommorow and try adding it to nixpkgs

is that okay with you?

Adám

Of course. Free world :-)

rak1507

well...

Adám

@rak1507 ?

rak1507

is RIDE free software?

Adám

Yes, FOSS. MIT.

rak1507

17:40

Oh right

Adám

@Konrad'Unrooted'Klawikowski How is it going?

Konrad 'Unrooted' Klawikowski

{ stdenv, pkgs, wrapGAppsHook }:

let
electron_t = pkgs.electron_8;
in
stdenv.mkDerivation rec {
src = ./.;
version = "1.0";
pname = "ride";
name = "${pname}-${version}";
buildInputs = with pkgs; [
electron_t makeWrapper wrapGAppsHook
];

good

Adám

I meant with the ThatAreDivisibleBy function.

Konrad 'Unrooted' Klawikowski

⋄ (⍳5) {⍵÷⍺} 2

Dyalog APL

@Konrad 'Unrooted' Klawikowski
┌→───────────────────────┐
│2 1 0.6666666667 0.5 0.4│
└~───────────────────────┘

Konrad 'Unrooted' Klawikowski

17:49

lol

Konrad 'Unrooted' Klawikowski

18:05

yeah I have no idea

tho I think about nixpkg solution

wait

Adám

@Konrad'Unrooted'Klawikowski OK, let me know another time when you want to continue. My kids need dinner anyway…

Konrad 'Unrooted' Klawikowski

greet your kids and tell them that they have awesome father

Adám

Done. :-)

Konrad 'Unrooted' Klawikowski

<3

rak1507

'hi kids, I'm awesome!' lol

Kamila Szewczyk

18:27

i've been working on the promised APL dialect

i posted a few binaries on the esolangs server, but i got some stuff to work already

v0.1b4 binary: b.catgirlsare.sexy/q96vt_dK.

it has lambdas ("dfns"), I/O, basic monads/dyads, guards:

f:α+α×β
a={~$⎕α+β}[24;24]
g:α>0:β⋄α
b=g[0;9]
c=g[3;9]
R=a+b+c

i decided to blend together a few paradigms, and so i took a few K features, a few APL features, and a few Forth features to make this thing

it's still very young and i started working on this particular implementation yesterday afternoon, and there were a few bugs that i fixed

i brainlessly used unordered_map for the symbol table, which resulted in the evaluation order being undefined, so that for example a=2 and b=a would sometimes error, and sometimes yield a=b=2

this particular example should print 0 and return 216

rak1507

what is β?

Kamila Szewczyk

second argument

rak1507

fair enough

what is $ and ⎕

Kamila Szewczyk

$ - alternative ⎕ - output device

rak1507

cool

the link you posted seems to not work

Kamila Szewczyk

18:33

the APL evaluation order stays

i'm going to implement some sort of a looping construct, but i'm not sure what do i want to decide on

rak1507

as long as it's not :For i :In thing

Kamila Szewczyk

i'd probably either borrow the power operator, or support just recursion, or allow explicit loops over dfn's with a condition

but because everything is immutable, explicit loops make no sense

rak1507

oh immutable cool

I think ⍣ makes sense

Kamila Szewczyk

i've been oscillating around supporting it

i know that recursion works for sure, and it can be used as a looping construct because we already have guards

rak1507

if you do implement ⍣ can you think about calling the condition first

so like

function⍣{thing that returns false} value doesn't run

Kamila Szewczyk

18:36

yep, my language has quoting

rak1507

I think that would be good but other people might have other opinions

Kamila Szewczyk

function won't be evaluated unless it needs to be

Adám

@rak1507 No, more like "Hey Kinderlach, einer was heist Konrad Klawikowski hat mir gesagt, as ich soll enk sagen, as enk haben an awesome father!"

rak1507

yea I was more meaning a way of doing conditions using ⍣ because at the moment if you have f⍣g⊢x it runs f even if g x is false at first

@Adám Is that german?

Adám

@KamilaSzewczyk So how was it again, only lists of lists, or proper multi-dimensional arrays?

Kamila Szewczyk

18:38

currently, it has only scalars... :p

but lists are going to be added soon-ish

Adám

@rak1507 Kind-of. Yiddish.

rak1507

Aha

Was wondering why google translate didn't work

Adám

@KamilaSzewczyk Only, right? No matrices?

Kamila Szewczyk

nope

@rak1507 lambda+guard kind of solves this

parameters which aren't shadowed, aren't overwritten

rak1507

true, but even so, I think the power operator should check g x first

idk

Kamila Szewczyk

18:42

so that for example this code returns 20:

f:α+α×β
g:f[2]
R=g[4;4]

rak1507

no brackets required for your lambdas?

Kamila Szewczyk

there's no lambda involved here.

i'm speaking of shadowing now

rak1507

what's f:α+α×β

Kamila Szewczyk

replace beta with omega

and : with ←

and add braces around it

Adám

@KamilaSzewczyk Not that it matters much here, but now, according to official APL Wiki terminology, it wouldn't be called an "APL dialect" but rather a separate language. RAD calls itself an "APL derivative". Do you have a name for your language yet?

rak1507

18:43

oh right so like the 'whole thing' runs inside braces?

Kamila Szewczyk

yes

check the more sophisticated example above

rak1507

or is f:α+α×β equal to f←{⍺+⍺×⍵}

Kamila Szewczyk

yes

rak1507

got it

Marshall

@Adám It's not official, it's just APL Wiki convention. No one outside APL Wiki has to follow its guidelines.

Adám

18:44

@KamilaSzewczyk Are the braces optional then, or what?

Kamila Szewczyk

braces define a lambda

f:α+α×β <=> f={α+α×β}

i use the first one, because it's shorter

Adám

@Marshall Of course, though it is "official APL Wiki terminology", no?

Marshall

@Adám I don't think I'd call anything on APL Wiki "official". There's no officiator.

Adám

I see. I was more thinking of official as meaning explicitly declared by itself.

rak1507

are square brackets going to be the only way of calling a function?

Kamila Szewczyk

18:48

no

i plan on adding more stuff, but i need to work my way up gradually :p

rak1507

cool

Marshall

@Adám I don't think it's important whether it matches any definition of "official". It's just kind of misleading because it sounds like some People In Charge have decreed that's how things work, but it's just some stuff I wrote.

rak1507

have you thought about supporting more than two arguments? f[x;y;z]?

@KamilaSzewczyk (btw, I'm not trying to say 'add all this stuff', just interested in what your plans/ideas are)

Kamila Szewczyk

currently 4 arguments are supported

but i'll extend it to infinity later

rak1507

first 4 greek letters?

Adám

18:50

@Marshall Right. Sorry @KamilaSzewczyk, I didn't mean to be harsh about your use of the word "dialect".

rak1507

first... infinity ... greek letters?

Adám

@Marshall RAD and Kamila's language are actually interesting cases in maybe looking even more like traditional APL than A+, but not supporting arbitrary rank. "APL dialects"?

Kamila Szewczyk

first 4 greek letters for referring first 4 arguments

then i'd use a separate letter for argument vector

Adám

What would we call something that say worked exactly like Dyalog APL or APL2, but if you tried to do anything that would result in an array of rank≥2, it'd error?

@KamilaSzewczyk Heh, problem with ε vs ∊?

Marshall

@KamilaSzewczyk Okay, I'm still confused about how this example works. What does it mean to call f[2] if f takes two arguments? Evidently g's arguments get passed to f, but what if you have a definition like g:f[2]+f[3]?

rak1507

18:52

epsilon isn't in the first 4

Adám

@rak1507 No, it is a reason to stop at 4. It is in the first 5.

rak1507

@Adám oh right yeah, I see what you were saying, I thought you were saying there might be an issue with that

Marshall

@KamilaSzewczyk Huh, I like that strategy.

Kamila Szewczyk

@Marshall the 2nd argument is taken from the parent

and the 1st argument is shadowed by the call

@Adám i'm not sure if i'll have all APL operators :p

Marshall

@Adám I hadn't looked into RAD at all, and had assumed it was more different from APL. Are there examples of RAD code out there? The question is whether typical expressions would be intelligible to a "generic" APL progammer.

Adám

18:57

@KamilaSzewczyk is f[a;b] the same as a f b or b f a?

Marshall

@KamilaSzewczyk In that case f would get arguments f[2,4] and return 10, right? Did I misunderstand?

Kamila Szewczyk

@Adám a f b

@Marshall i typoed, it should be 10 :/

Adám

@KamilaSzewczyk Isn't that a bit odd that f[b] is f b but a f b is f[a;b]? I.e. the right argument is not always the first (and only) argument.

Marshall

@KamilaSzewczyk Not sure I like it, but that makes sense at least.

Seems a little precarious in that if you add an argument to a function but forget to change one of it's callers, it can be passed an argument you didn't expect.

Adám

@Marshall Yes

rak1507

19:06

I don't think I've heard of RAD before, what is it?

Adám

@rak1507 apl.wiki/RAD

Kamila Szewczyk

@Adám this APL thing doesn't apply here

there's no right side and no left side in function application

this comparison is pointless

rak1507

Ah cool

Kamila Szewczyk

@Marshall i think it's nice

i wanted to put some practical use to APL, because i really want to make a bytebeat/graphics/music-oriented languge

Adám

?

Marshall

19:09

@KamilaSzewczyk Oh, I'd be interested in that. I've done some live (in the sense of live feedback, not actually listenable live) electronic stuff in J and it works all right but there are clearly some deficiencies.

Dyalog APL

<Kiscica> The alternative breaks math - you want 2 - 1 to be 1.

Marshall

@Marshall Started translating some of it to BQN but missing fast float to int conversion functions and FFT means it's slow.

@KamilaSzewczyk Did you know SuperCollider has a little J influence?

Kamila Szewczyk

@Adám i want to have, for example, a way to assume that a single argument is every argument of a call

i may also create some sort of train-like construction

at the moment i don't have everything clearly specified and i'm adding stuff as i need it :p

Adám

:-)

dzaima

@Adám chatexchange has given be duplicate messages once too once

Adám

19:19

@user1543385 Hi Lukas Hermann. If you want to participate here, email [email protected]

Marshall

22:22

@dzaima I added some specialized code for specific lengths in my filtering code, and I'm surprised to find that it's hardly any faster under OpenJDK, and actually slower with the Graal native build. Any insight into this? Is arithmetic really that slow?

Some test code:

g‿f ← •Import¨ "oscillator.bqn"‿"filter.bqn"
•←≢500 f.Lp2 g.Saw 1e6⥊200

dzaima

22:35

the old filter file appears to create 4M Num instances, whereas the new one creates 10M

so that's 2.5x more GCing

also the spam of VARO/VARMs (yeah, my compiler still uses those for non-local variables..) probably takes most of the time anyway

Marshall

@dzaima Oh, the numbers are all references. Yeah, that's tough. Switching to LOCs at least would be useful though.

dzaima

yeah, there's not really any way around that in java

@dzaima (reason being that blocks are compiled the moment its "token" is created, so the outer block just doesn't exist at all yet)

@dzaima (of course generating special code for scalar number arguments in scalar functions would be fun and all, but eh)

Marshall

22:57

@dzaima The point of compiling blocks early is so that you know the type, right? I never got why you would do it that way. Can't you just keep a stack of block information (opening brace location, which special variables or ; or : have been seen) and annotate the closing brace (also link the opening brace to it) when you find it?

dzaima

@Marshall I did it that way because that's the first thing that came to mind, and it had no minuses at the time. Of course I could switch to something different now, but it'd probably take >5 minutes so i haven't done it yet as there hasn't really been any reason to me to

Marshall

@dzaima Fair enough. Some day I'll spend >5 minutes myself and actually support headers...

dzaima

23:17

@dzaima actually like just moving this line to Somewhere Else™ would really be enough.. well i'll do that Tomorrow™

Transcript for