The APL Orchard

Wzl

12:00 AM

does it really omit combinators though?

Adám

It really does.

Marshall

Other aspects of K that I don't like are that it puts too much functionality in one glyph because of the ASCII restriction and that it separates integers and floats. Given that doubles contain 53-bit ints as a subset I think the added complication of separate higher-precision ints is very much not worth it for just 11 bits.

Adám

One of my main gripes with K is that you can't (easily) create your own infix functions.

Marshall

You could call K trains combinators, but that's the only kind, so it's fair to say it omits nearly all of them.

Wzl

can't you write them though? Then it's just a syntax difference

Marshall

12:08 AM

No. No user-defined adverbs/operators.

Adám

You can write functions that take functions as arguments, but you have to use name[f,g,…]

Wzl

@Adám exactly: just a syntax difference

Adám

@Marshall How do you distinguish between user-defined functions and operators for K, where you must use square brackets to call a user-defined thing with multiple arguments?

Marshall

Oh, true. I guess that does qualify as "just a syntax difference", although I view the syntax of tacit programming as very important.

Wzl

mumbles something about forth and postscript

Marshall

12:11 AM

I have used Factor, and wrote a few stack-based array languages early on. I discovered that I do not like to use these.

Wzl

I mean the syntax, that is what I like about them

Adám

IMO it is hopeless to have to write over:{x[y[z];y[d]]};over[=,_,a,b] instead of over←{(⍵⍵ ⍺)⍺⍺(⍵⍵ ⍵)} ⋄ a=over⌊b

There, fixed it with a header:

And why not put the left arg on the left while we're at it:

over:{[x,f,g,y]f[g[x];g[y]]};over[a,=,_,b]

Wzl

... thinking of something APL is bad at :P

Adám

@Wzl Wouldn't it be Lisp? K uses Lisp's (never implemented) M-expressions.

Wzl

@Adám they were implemented but didn't catch on btw

I was mentioning postscript's very extensible syntax

Adám

12:27 AM

Oh.

Marshall

Also you can't implement a combinator like Over as a function that takes two functions and returns one of them, because no closures. over:{f:x;g:y;{f[g[x];g[y]]}} doesn't save f and g long enough.

Wzl

however Adám's version works the same as that (because you can curry it)

Adám

The first one, you mean?

over:{[f,g,x,y]f[g[x];g[y]]}

Marshall

Well I guess that's the way dops work in Dyalog too. In J and BQN you can do computation when you get the functions.

Adám

And I never understood how the system knows when to start computing.

Marshall

12:32 AM

No difference for combinators though.

Adám

(Oops, I had ,s instead of ;s)

Marshall

There are two sets of special names, 𝕨𝕩𝕤 for a function call and 𝕗𝕘𝕣 for a modifier call. The block runs once all sets that appear have values.

So the four possibilities are no special names (immediate block), function names only (function), modifier names only (immediate modifier), and both (deferred modifier).

Adám

Ah, thanks.

@dzaima Now decided: I don't like the modifier guessing.

Wzl

true there should be closures (in K) but I still don't like operators

Adám

I can confirm that this works in K:

 over:{[f;g;x;y]f[g[x];g[y]]}
 caselessEq:over[=;_]
 caselessEq["Abc";"aBd"]
110

So, equivalent to:

      over←{(⍵⍵ ⍺)⍺⍺(⍵⍵ ⍵)}
      caselessEq←=over⎕C
      'Abc'caselessEq'aBd'
1 1 0

I definitely prefer APL's syntax.

elliptic00

1:32 AM

can we compile APL code from command line?

Adám

@elliptic00 APL is an interpreted language.

elliptic00

then can we run APL code from command line?

Adám

Yes. How is your APL code stored?

elliptic00

I think you can run GNU APL from command line?

just put one a file?

put code in a file?

Adám

OK, but how does the file look?

Is the content of the file a function?

elliptic00

1:35 AM

sure..

Adám

Then try dyalog LOAD=filename

elliptic00

It means I can't run APL ouside RIDE?

Adám

Wait, which OS are you on?

elliptic00

macOS

Adám

From inside APL, can you run ]config DYALOG please?

elliptic00

1:40 AM

yep

Adám

What did it return?

elliptic00

/Applications/Dyalog-18.0.app/Contents/Resources/Dyalog

Adám

That'd be the directory where the executable is found.

elliptic00

Dyalog is the APL binary?

Adám

Yes.

elliptic00

1:42 AM

if I put APL code in a file, for example.

m <- 1 2 3

Adám

Another method is, from inside RIDE, go in the menu File>Connect… then select Type:Start and look at which interpreter RIDE is intending to start.

@elliptic00 Right, that's not a function, but you should be able to pipe that into dyalog -script

elliptic00

in a file, and Dyalog. myapl ?

myapl contain my APL code

Adám

No, try dyalog -script < myapl

elliptic00

oh..OK.. let me try it.. thanks..

Adám

Of course, not much will happen, as you're only assigning without causing any output.

elliptic00

1:45 AM

Yep.. how to use stdout like printf in C/C++?

Adám

⎕←stuff to print

elliptic00

nice..

Adám

For simple questions like that, remember that you can ask APLcart.

elliptic00

can I.. Nice.. I don't know that.. thanks your suggestion..

Dyalog APL could not initialise because the script file could not be opened.

dyalog -script < /tmp/apl.x

It seems there is some issue when I try to run my script

Adám

Sounds like a file permissions issue.

elliptic00

1:49 AM

  m ← 2 3 ⍴ ⍳ 10

  ⎕ ← m

this is my code inside apl.x

Adám

That's fine.

elliptic00

apl.x is under tmp.. should not be any permission issue?

Adám

I think you may need to give dyalog permission to access files outside your documents dir.

macOS has become very restrictive lately.

elliptic00

how I do that?

Adám

Can you try to just run dyalog and then from inside APL, do ⎕NGET'/tmp/apl.x'

Enter )off to close APL when done.

@elliptic00 System Preferences -> Security & Privacy -> Privacy -> Files and Folders and Full Disk Access.

elliptic00

1:59 AM

It ok, from do ⎕NGET'/tmp/apl.x'

I don't have this on my MacOS Files and Folders and Full Disk Access.

⎕NGET '/tmp/apl.x'
m ← 2 3 ⍴ ⍳ 10

⎕ ← m
UTF-8-NOBOM 10

Adám

Hm, ok, so APL can access the file.

elliptic00

It seems to me..

Adám

Then I'm not sure why it doesn't work. I suggest you contact support@dyalog.com

elliptic00

Ok..Thanks

Adám

Meanwhile, if you change instead use a file called apl.aplf and add a line at the top and at the bottom so it says:

apl x;m
m ← 2 3 ⍴ ⍳ 10

⎕ ← m
⎕OFF

Then try dyalog LOAD=/tmp/apl.aplf

Matthew Daniels

2:10 AM

how to output using APL?

Adám

Put ⎕← on the left of what you want to print.

elliptic00

ok, I can useing LOAD=/tmp/apl.aplf

Adám

Great, so it prints the matrix?

elliptic00

it is just command line like RIDE..

yep. it print the matrix

Adám

@elliptic00 What do you mean? Does it not quit when done?

elliptic00

2:13 AM

no, when I'

Matthew Daniels

so how to output Hello.?

Adám

@MatthewDaniels ⎕←'Hello.'

Matthew Daniels

I can't figure out the way.

oh.

what about input?

:)

Adám

⍞

elliptic00

when I use dyalog LOAD=/tmp/apl.aplf, print out the matrix, and I'm still inside the command line, and I can run APL code in REPL, but it is like a shell

Adám

2:14 AM

@elliptic00 I see. Did you make sure to add ⎕OFF at the end of the file?

elliptic00

OH.. NO. maybe this is why I'm still inside the prompt..

ok, )off, I quit

Adám

Indeed. ⎕OFF turns APL off under program control.

Matthew Daniels

I'm trying to write an APL program of Hello, <name>. and I'm trying to set up the input and adding the name in.

elliptic00

like exit in C/C++

Matthew Daniels

my variable is n.

Adám

2:16 AM

n←⍞
⎕←'Hello, ',n,'.'

elliptic00

It works..now.. thanks..

Matthew Daniels

thx. :D

Adám

@elliptic00 Sorry for all the trouble. All this should get much easier in 18.1.

Matthew Daniels

yeah, APL looks very hard.

considering the characters can't be typed in with a basic keyboard.

Adám

The code language is in fact very easy. Unfortunately, some of the tooling is still awkward, but we're working on improving that.

@MatthewDaniels apl.wiki/typing_glyphs

LdBeth

2:40 AM

It seems I'm the only one have developed APL keyboard for Japanese keyboard (JIS) layout.

Adám

@LdBeth Ooh, is it available to the public?

LdBeth

It's developed using Ukelele (a keyboard editor for Mac), available as a Mac input method bundle.

Adám

@LdBeth Right. Maybe it could be added to apl.wiki/Typing_glyphs#By_platform?

LdBeth

Good idea, I just need to write down a readme so people can figure it out easily. It is not based on the current most popular layout but dyalog.com/uploads/conference/dyalog06/presentations/…

dzaima

8:21 AM

@Adám i mean, yeah, there's not much to like about them, but the alternatives aren't pretty either. (and i absolutely hate immediate function blocks)

actually, now that i think about it, I'd take forcing headers for everything other than non-immediate functions over needing the extremely stupid 𝕤⋄ or 𝕩⋄ or 𝕊: for functions that don't use 𝕨 and 𝕩

Adám

@dzaima Well, my suggestion was to use _{…} and _{…}_ but that wouldn't work for immediate blocks.

dzaima

@Adám then you'd need to have spaces between them and names, leading to an ugly F _{…}_ G (and F _{…}_G would be 2 modifiers, leading to more confusion)

Adám

Maybe the left and right indicators should be two different chars.

Adám

8:56 AM

If the left symbol was _ and the right symbol was ¯ then F_{…}¯G is a 2-modifier and ``F_{…}_G` is 2 modifiers. Maybe the symbols are not ideal, but you get the idea.

@dzaima F◞{…}◟G vs F◞{…}◞G

Razetime

9:19 AM

interesting (and defunct) array language

Adám

@Razetime Ah yes, I was meaning to add an APL Wiki page for it. Do you want to do that?

@Razetime Wait, Glee is a LtR APL?

Razetime

9:40 AM

@Adám idk i found it on hello world and honestly have no idea about it

1 + 2 3 4 	3 3 4
1 + (2 3 4) 	3 4 5

looks like it's ltr

Adám

And stranding is after function application. Clearer example:

__ ~(13 10#asc)=> A;

Razetime

it has.. decimal indexing?

Adám

No, that's ⎕ASC i.e. ⎕UCS

Razetime

looks like a fully ascii version

Adám

Yes, it is ASCII-only

Razetime

9:45 AM

seems like square brackets can also work with filtering predicates

Adám

The above has __ be the end of a here-doc, and then ~(⎕UCS 13 10)

So A ← (⎕UCS 13 10)~⍨…

LtR indeed.

Razetime

"raw text"

this site keeps breaking

looks like it has lazy sequences

unimplemented sockets lib

ok, enough rabbit hole for now

xpqz

10:05 AM

What's today's CMC?

Razetime

implement J in Dyalog APL

xpqz

Before lunch?

Razetime

sure

Adám

CMC: Given the representation of a tradfn (your choice of vector, vector of vectors, or matrix) return a list of local names.

E.g.

 {r}←(F _Op)(a b);⎕AVU;abc
 ⍝ hello world
 ;more;names;here
 boo

should give F _Op a abc b here more names r ⎕AVU

Harder test case:

 {r}←(F _Op)(a b);⎕AVU;abc⍝;fool
 ⍝;hello;world
 ;more;names;here
 boo[;goo]

xpqz

10:27 AM

I didn't know that the ;local;names;can;span

;several;lines

LdBeth

10:38 AM

github.com/LdBeth/APL-JIS

This is the before mentioned APL layout for Japanese keyboard

hope someone would help put on aplwiki. I don't have account.

Razetime

10:52 AM

I'll add it

I'll be putting it here: aplwiki.com/wiki/Typing_glyphs#By_platform

@LdBeth added

LdBeth

thank you

rak1507

12:19 PM

meetup.com/ACCULondon/events/277980280 @RikedyP is there a way to attend this without making an account and uploading a profile picture?

RikedyP

12:34 PM

@rak1507 I'll ask - this was the only link I was given

Adám

1:41 PM

@user3821550 Hi. If you want to participate here, email access@apl.chat

xpqz

2:30 PM

@Adám No one else seems to be playing? Here's my tee-off:

{('{}(); ←'(~⍤∊⍨⊆⊢) 0⌷d),⊃↓';'(≠⊆⊢)⍤1⊢(';'=⍵[;0])⌿d←(~∨\'⍝'=⍵)/⍤1 1⊢⍵}

      ]repr ex2
4 31⍴'{r}←(F _Op)(a b);⎕AVU;abc⍝;fool⍝;hello;world

      cmc ex2
 r  F  _Op  a  b  ⎕AVU  abc  more  names  here

Sorry:

4 31⍴'{r}←(F _Op)(a b);⎕AVU;abc⍝;fool⍝;hello;world                  ;more;names;here               boo[;goo]

RGS

2:43 PM

Announcement: Dyalog webinar on “Drawing the Mandelbrot Set” in 17 minutes at dyalog.tv/Webinar/?v=ozaRMHeYWYM

rak1507

typo: well-know

RGS

wooopsie

@rak1507 should be fixed by now

rak1507

yep

xpqz

Mandelbrot as a tradfn? This I must see.

rak1507

sounds blasphemous, I can't wait

RikedyP

2:56 PM

A recentish code_report video includes an example which I think pretty well shows how an idiomatic APL approach to a problem compares to an alternative yet still functional approach
https://www.youtube.com/watch?v=QtvvQ7MdwKY

⊃(⊣,' ',⊢)/4↑' '(≠⊂⊢)sentence
Vs
{⍵⌿⍨4>+\' '=⍵}sentence

In my opinion, this example is simple and applicable enough to be interesting, but other nested array techniques (like the ones from the webinar on boolean scans and reductions) are a bit esoteric and probably less convincing. I think it's like striking a balance between what is good idiomatic APL for a given pr…

Adám

@user3821550 Welcome. You should now have access.

rak1507

@RikedyP classic example of under

not sure under could ever actually work here

it's basically 'take under splitintowords'

haha, martin replied the same thing!

RGS

5:01 PM

@rak1507 how well did you take it?

rak1507

the talk itself was great, so good job, as for the for loop I'd probably use something with ⍣ but it doesn't matter too much

I think the one thing that would be a good change would be outputting in PPM format to avoid a dependency on that PNG thing

RGS

@rak1507 I'm definitely open to suggestions on alternatives to the :For

@rak1507 Yes, Adám suggested pretty much the same thing, but I didn't have time to change to that.

rak1507

Fair enough

Sherlock9

9:33 PM

I was reading back to the discussion last week about how Dyalog APL is broken and can't really be fixed without breaking backwards compatibility. How do you guys think dfns, tradfns, and modified assignments should work, and what else needs fixing (or would go into a clean-slate APL-like)?

rak1507

preferably tradfns would be erased from the universe, potentially dfns should have an alternative return syntax then you could have multiple guards, and multiple things executed without having to add _← in front of stuff, modified assignment not sure

dzaima

also dfns should always return the last statement

rak1507

yeah

can't think of many ways to adapt things without turning it into haskell - array edition

dzaima

modified assignment should at the very least allow parenthesizing the function (so a (+←) 1), or go BQN and be a separate character

Sherlock9

I would absolutely be down for Haskell - Array Edition, for the record

rak1507

9:37 PM

the problem with that is the type system can either be far too general to really be useful or be far too complicated to be useful

like, what's the type of an inner product?

+.× :: ?? -> ?? -> ??

user

I like the way Haskell's functions are defined. If APL could do that too, it'd be one step closer to math notation

@rak1507 I guess that's where dependent types or overloading come in?

rak1507

APL's composition abilities are better than Haskell's

Wzl

@user I like pattern matching but not pattern matching in definitions

rak1507

@user yep, which makes the signatures so long

oh yeah, pattern matching, that is a cool thing but maybe not that useful in APL

Wzl

(for moral reasons :P)

user

9:39 PM

@rak1507 idk, I sometimes find myself needing it

rak1507

it would certainly be interesting if you could do haskell style function definitions with pattern matching and APL primitives

Wzl

I like inline pattern matching like case (I think it's called case in haskell)

user

I guess inner product could be f.g :: f⍨ a g b if you could somehow treat the functions as also function types

Oh wait, that doesn't work

rak1507

rev ⍬ = ⍬
rev (x,xs) = (∇xs),x

Wzl

haskell doesn't have a type for operators so that would have to change (operators removed or at the very least first class functions, please)

rak1507

9:42 PM

yeah, first class functions

that would be useful

then you could basically emulate control flow with higher order functions too

Wzl

@rak1507 you don't need recursion as often in APL though

rak1507

so you could bin :For :While etc

@Wzl true, that was just an example of how a recursive definition of reverse might work in hask-PL

it's just
rev [] = []
rev (x:xs) = (rev xs) ++ [x]
but with APL primitives

Sherlock9

I'm trying to think of what I'd want to import from Haskell into APL, and now I'm realizing that it's more that I want to import APL into Haskell

rak1507

a bit of both I think

Sherlock9

The compositions, sure, but also array-based programming but with lazy functions and so we get those infinite lists

I have not a clue how it would work when implemented, though

rak1507

9:51 PM

large scale APL projects seem to be very OOP-ish so I think it'd be interesting if it was more functional

Sherlock9

I might try writing up a specification at least when I learn more Haskell and more APL

rak1507

would be an interesting experiment

user

How would you pass operators around as first class functions without hyperoperators or something?

Sherlock9

And I could try building an interpreter for it. That's another project I've been meaning to pick back up, just below learning APL and learning Haskell on the list of things to do

user

Or would you use Haskell syntax instead?

Adám

10:01 PM

@user BQN allows you to switch syntactic class of named entities on the fly.

rak1507

haskell syntax can be pretty terse, so it might be a good fit

Sherlock9

Am I off in my understanding that operators are functions that take other functions as parameters? Because Haskell does that just fine, if so

Adám

Operators can take arrays and/or functions as operands, but they cannot take operators as operands.

Sherlock9

What I'm not sure about is whether the best syntax to keep is Haskell's currying from the function on the left to the right (or from right to left?) until it's taken enough arguments, or to keep to APL's on one argument on the left, one argument on the right.

This requires a great deal more thought than I am currently able to give it

rak1507

well there are still infix functions in haskell

so you could still have APL like syntax, but the potential to use multiple arguments if non infix

Adám

10:13 PM

Like K.

rak1507

similar to k, but still allowing user defined infix functions and some other nice APL things like that

user

I suppose operators would be infix functions, and functions would be normal currying functions

I think it's similar in math?

rak1507

most existing APL primitives could stay infix maybe

user

Perhaps the syntax for expressing a type could be a (dynamically typed) APL expression

rak1507

that would be interesting

user

10:26 PM

Basically, it could have first class types, with the type of a function or operator depending on the arguments' types.

No idea if that's feasible or if it would be convenient, though

Transcript for