The APL Orchard

Razetime

4:02 AM

is it a good idea to have an account on the dyalog forums

rak1507

6:57 AM

@Marshall is the rest of this talk youtube.com/watch?v=WlUHw4hC4OY available anywhere?

rak1507

9:52 AM

Is ⊃∘⍒ optimised to be O(n)?

dzaima

@rak1507 you can test

Adám

@rak1507 No, but ⊃⍤⍒ is.

rak1507

@dzaima I did and I thought it was strange, so I wanted to check as I thought I remembered it being mentioned

@Adám thanks

Adám

(⊃⍒) is too, btw, but only as a train, not as a phrase.

rak1507

Ahh makes sense

How do programming languages implement these special cases? Are there checks before the code is run for certain phrases or what?

Adám

10:06 AM

Dyalog "idioms" are phases recognised by the tokeniser. Things like this atop are checked for at run time.

rak1507

Does that add much time to the overall speed? If it has to check for several different idioms

Adám

I think it is a very small check. Basically just "is LeftOperand=SomeValue". @Marshall can give more details.

dzaima

@Adám it's a small check, but potentially executed many times. Still, modern cpus should be able to execute it in parallel to what's afterwards, and the branch should be very predictable

Razetime

12:19 PM

Try it online!

I want to replace each character in '×÷-+' with each of 'xdmp'

how do I do that?

dzaima

@Razetime 1) there's no ⍵ in tradfns, 2) you're not giving an argument to @

Razetime

yeah the thing is incomplete

> I want to replace each character in '×÷-+' with each of 'xdmp'
mainly need to find out how to do this

and one more thing

is there a way to include operators in variable names

dzaima

@Razetime i've got a long {('×÷-+'⍳⍵)⊃¨'xdmp'∘,¨⍵}

@Razetime as in assign an operator to a variable?

Razetime

no use an operator as within a variable/function declaration name

like x÷←{⍵}

should define x÷ as a constant function

dzaima

@Razetime nope

Razetime

12:25 PM

figured as much

dzaima

in Extended you can get away with reassigning functions in dfns, but that's the most you can do

Razetime

that's funny, I'll remember that

dzaima

@dzaima (note: only for builtins that have covers, so not e.g. +)

Razetime

so stuff which is in the github readme spec

Marshall

1:01 PM

@rak1507 The rest of the talk wasn't given, but slides, scripts and some commentary are linked here.

rak1507

Great thanks

'We didn't reach this section in the talk. Consequently, no thinking occurred.' lol

Adám

1:32 PM

@Razetime ⎕*3 Studios

@user8773069 Hi John Doe. If you want to participate here, email me: adam@ with the same domain as www.dyalog.com

Razetime

2:01 PM

@Adám I made that name in 5th grade

still pretty proud of it

cannadayr

3:35 PM

not sure if this was posted, but Gilad Bracha did a webinar talking about a recent upcoming array language (video in link)
https://news.ycombinator.com/item?id=25146654

i skimmed thru it, it seems to use a K-like syntax (?) including apl type research by Justin Slepak (didnt link to a specific paper), and something he calls "stream polymorphism".
https://www.researchgate.net/scientific-contributions/Justin-Slepak-2048362989

it also does "scalars are tensors of rank 0" which (i think) makes it a floating array model ?

also, got some time last cpl nights to work on bqn -> erlang port. I got most of the stdlib implemented a few weeks ago.
I started using the process dictionary which has made dealing w/ mutability significantly easier (acts as a c-like escape hatch).

rn it returns a function, that when called gets the next opcode and leb128 arg.

lot of putting similar logging in js version and erlang version and walk thru whats happening.

using the process dictionary has been biggest help. its very discouraged but it makes sense to use for this purpose -> each process will have its own vm

Razetime

3:56 PM

@cannadayr this looks cool

cannadayr

@Razetime it seems like its going to be open-sourced and supported by the company he works at "shape security".

Razetime

yup watching the video

cannadayr

their use case is probably streaming analytics

Razetime

why exactly that?

cannadayr

guessing based on their acquisition press release
https://www.f5.com/products/security/shape-security

Razetime

4:00 PM

not bad

Adám

Announcement: BAA open session now.

Razetime

that was unexpected

ima join

@Adám I love language based passcodes

xpqz

4:29 PM

If I want a vector of (say) 25 scalars, which of 25/4 or 25⍴4 is preferrable, or more idiomatic?

Adám

5:10 PM

@xpqz I think they run the same code under the covers, and both are fine APL. I think I'd use ⍴ if I conceptually reshape, but / if I conceptually repeat stuff. I also have a bit of a preference for ⍴ because it doesn't want to be an operator…

      ⎕SE.Dyalog.Utils.repObj 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4
25⍴4

Razetime

I'd use / cause it simply means "create a list with ⍵ repeated ⍺ times" rather than reshape

@Adám this was nice!

user

Hi, I'm new to APL and I was wondering how one defines variables in dfns in Dyalog APL. I tried using jot in f←{g←0∘g+⍵}, but that gave me an error saying g was undefined. Is there any way to define variables inside dfns without newlines?

cannadayr

statement separator ⋄

I dont have a lot of dyalog experience but i think jot is the compose operator (when not used as an outer product)

user

Thanks! I guess I was getting confused because they look similar.

Razetime

@user ⋄ (diamond). The language bar shortcut is typing backtick twice

Dyalog APL

5:17 PM

@Razetime ``

Razetime

it's also used to call the apl bot lol

user

Oh

Adám

@user Welcome to the APL Orchard, then!

user

@Adám Thanks! APL seems like a cool language, although I've yet to get the hang of it.

Razetime

I suppose we can use the new TryAPL to post Dyalog 18 answers now

right?

Adám

5:21 PM

Ah, yes, of course!

Razetime

cool

'night

bit late here

Adám

5:34 PM

@user I can recommend APL Wiki's list of learning resources.

user

@Adám Cool, thanks again

rak1507

5:49 PM

Oops, I missed the BAA session

Adám

6:01 PM

@rak1507 Will be published ― eventually.

rak1507

What was the subject?

Adám

6:13 PM

@rak1507 It was an open session, but most of the time was consumed by me showing off the new TryAPL 3.0.

rak1507

Cool

rak1507

6:48 PM

Have you thought about moving the buttons at the top to the right hand side only? I sent a link to a discord group I'm in and someone suggested that, it might look better, idk

Adám

7:10 PM

@rak1507 Which buttons? The language bar?

rak1507

Yeah

Adám

We are primarily targeting desktop users, and they usually have plenty of horizontal space, but are vertically constrained. By spreading the language bar over the entire width, it can use as few lines as possible. Most of the time, it's bottom edge wouldn't be flush with the tabs' bottom edge anyway.

Of course, it could be made into a side pane, but that tends to look odd:

@NicholasG04 Hi. Interested in APL?

rak1507

Another thing is in my opinion this paragraph 'TryAPL runs on Dyalog, which you can download for free, or try it now by entering an expression (use the language bar above to type the special APL symbols), or clicking one of these expressions, followed by Enter, to see it in action:' should definitely mention using ` to type characters

user

8:33 PM

⋄ f←{⍵<2:{⍺+⍵}⋄{⍺×⍵}}

@user ``

⋄ f 1

@user VALUE ERROR

⋄ f←{⍵<2:{⍺+⍵}⋄{⍺×⍵}} ⋄ f 1

Dyalog APL

@user SYNTAX ERROR

Adám

8:34 PM

@user The bot doesn't remember from command to command.

user

Yep, I just realized that

Adám

{…} is an actual function, not a code block. So the ⍺+⍵ and ⍺×⍵ don't need braces.

user

Actually, I wanted to return a different from within the function depending on the first argument

Adám

You will, however need to supply a left argument if you have ⍺ in the code.

@user Sure and {⍵<2:⍺+⍵ ⋄ ⍺×⍵} will do that.

user

Oh, I see.

Adám

8:37 PM

⋄ f←{⍵<2:⍺+⍵ ⋄ ⍺×⍵} ⋄ 3 f 1 ⋄ 3 f 2

Dyalog APL

@Adám
4
6

user

@Adám What about if I want to give it three arguments (something like a (f n) b)?

Adám

Ah, like to have n choose between addition and multiplication?

You can use a (n f) b as follows: ⋄ f←{⍺⍺<2:⍺+⍵ ⋄ ⍺×⍵} ⋄ 10(1 f)20 ⋄ 10(2 f)20

Dyalog APL

@Adám
30
200

Adám

Do you know about operators?

user

8:50 PM

I know of them, but haven't really used them. Does the double alpha allow you use it with postfix syntax?

⋄ f ← {⍺⍺ + 1} ⋄ 4 f

Dyalog APL

@user Illegal code

Adám

The entire parentheses (n f) defines a new derived function.

Operators are higher-order functions that take one or two operands (not arguments) and derive one out of a family of related functions.

user

Oh, I think I see

Adám

So e.g. +/ and ×/ are instances of the reduction family of functions. / is the Reduce operator.

user

Okay, and f is now another operator?

Adám

8:53 PM

Yes, this f is an operator, because it's source contains ⍺⍺ which means "left operand".

As you can see, an operand can be either a function or an array.

user

This actually makes what I want to do even easier, with a separate symbol for ⍺⍺. Thanks for helping me!

Adám

No problem. Btw, I'd be interested in what you are trying to do, as there may be a neat mathematical way to do it.

user

9:07 PM

@Adám I'm trying to answer this question by porting another answer of mine to APL (Dyalog). I've got f←{⍺⍺<2:⍵+⍺⋄((⍺⍺-1)f)/(×⍺⍺-2),⍺⍴⍵} right now, and it looks like it's working correctly.

Adám

@user That's very nice. One trick: You can make the operator anonymous by using ∇∇ to have it refer to itself.

user

Cool, thanks.

Adám

Did you check the test cases?

user

Yeah, unfortunately, it's not passing the 2 4 4 case :(

rak1507

9:48 PM

a b c←⎕⋄⍎⍕(a,{'(',⍵,'/⍴⍨)'}⍣(c-1)⊢'+'),b this is horrible (but it works)

user

@rak1507 Thanks! I actually got it to work with f←{⍺⍺<2:⍵+⍺⋄((⍺⍺-1)∇∇)/(⍵⍴⍺),×⍺⍺-2}

rak1507

?

ah nice

user

Sorry, I hit the enter key by accident

rak1507

I was thinking of doing something like that but you were already doing it, so I went with the string route

user

How does yours work? It looks really interesting.

rak1507

9:54 PM

It makes a string and then evals it, something like 2 (((+/⍴⍨)/⍴⍨)/⍴⍨) 4

user

Oh, cool

Adám

@user You don't need parens around (⍺⍺-1)∇∇

@rak1507 No need for parens: ⋄ 2 +/⍤⍴⍨/⍤⍴⍨/⍤⍴⍨ 4

Dyalog APL

@Adám 65536

user

@Adám Nice, it's down to 31 bytes now

rak1507

@Adám Ah yeah, that's nicer

Adám

10:09 PM

@rak1507 I was wondering i something like ⋄ ×/2/⍣(4-1)⊢4 ⋄ ×/3/⍣(4-1)⊢3 could work.

Dyalog APL

@Adám
65536
7.625597485E12

rak1507

That's really clever, looks like it does!

Adám

Doesn't quite degrade to n=1 but I'm sure you can work something out.

Adám

10:25 PM

@user Congrats on your first APL answer. May there be many more. One thing: You should avoid using Dyalog Classic, and instead use Dyalog Unicode (you can still count characters instead of bytes by referring to SBCS).

user

@Adám Does Dyalog Classic take up more bytes?

Adám

No, but it is deprecated, is missing features, including some glyphs, and doesn't really support scripting like what TIO uses. That's why it echoes everything to stdout.

user

I see. I'm changing to Unicode right now. Thank you for all your help!

Adám

That's what I'm here for.

Transcript for