The APL Orchard

4:30 AM

@user @Bubbler I think you'll find that it depends largely on the programmer. CS-type are likely to always use ⎕IO←0 while mathematicians sometimes will, and business people will usually stick to ⎕IO←1. Ancient code will use ⎕ML←0 and new will use ⎕ML←1 while medium-aged code, and code that has been ported from or written by people from other APLs will use ⎕ML←3. I don't think I've ever seen ⎕ML←2 used.

LdBeth

4:58 AM

I like the idea of choosing index origin based on problem context. but ⎕io implemented as a system variable is sometimes annoying.

Adám

Iverson actually had it as a subscript on functions themselves.

Bubbler

Also user-written functions often have ⎕IO←(0 or 1) on its first line as a local setting

LdBeth

the iota in A Programming Language is more like "range".

non 0/1 "index origin" can also be supplied.

Adám

5:13 AM

@LdBeth Indeed. I never noticed. Makes sense that it is called interval vector then, and hence ⍳. So now that we use ⍳ it doesn't actually fit the meaning, and those languages that have adopted Iota are doing so in "error".

1234

6:55 AM

how to input multiple lines in RIDE?

Adám

@1234 What exactly are you trying to do? Define a multi-line function or prompt for multiple lines of input, or something else?

1234

yep, try to define a multiple line function

f ← {⍺+⍵}

{ ---> line one

⍺+⍵. --> line two

} --- line three

Adám

It is probably nicest to write )ed f then edit and press Escape when done.

However, if you want to try an experimental feature, you can set the environment variable (or command line parameter) DYALOG_LINEEDITOR_MODE=1 in which case it may "just work".

Alternatively, you can begin the first line with ∇ and end the last line with ∇:

∇f←{
⍺+⍵
}∇

1234

Cool, It is hard to believe RIDE still can not use multiple line input yet,

Adám

RIDE is a "dumb" frontend. It has very little clue about what's going on. It is the interpreter that handles function definition. If the interpreter accepts multi-line input, then RIDE does too.

1234

7:03 AM

so the interpreter has hard time to accept multiple line input?

Adám

No, it just has to be enabled.

1234

Ok, I will try it.. thanks

Wong Jia Hau

@user I think you should take a look at this lecture by Simon Peyton Jones youtube.com/watch?v=brE_dyedGm0&t=3454s

Basically the benefit of having a type system is not linear, but parabolic, for example though Coq has the higher guarantee than Haskell, becoming productive in Coq is significantly much harder than Haskell, Jones also joked about you need to have a Phd to be able to write Coq

So, arguably, having a type system in APL might not have a net positive due to the rank-polymorphism

Aaron also has a nice talk regarding this topic youtube.com/watch?v=z8MVKianh54&t=4s

In short, he said the type annotation would be longer than the actual code itself

Adám

@WongJiaHau How is that important?

Wong Jia Hau

@Adám sorry which one do you mean?

Adám

7:14 AM

Why does it matter that the type annotation would be longer than the code?

Wong Jia Hau

@Adám then it would be bad in terms of productivity

"long" might be a bad description, "arcane" should be a better description

Adám

Except if all the type annotations could cut down on bugs, no?

Wong Jia Hau

for glue code yes

so I was suggesting to have gradual type system (a la typescript) for BQN

basically we can give the 'any' type to all the primitives since they are not trivial to type, but user should be allowed to define simple types like functions, record/structure, adt etc

Bubbler

Having records and ADTs would be definitely nice

Wong Jia Hau

@Bubbler yea especially when interfacing with external systems

Bubbler

7:22 AM

Not necessarily external systems. I once tried golfing an actual ADT manipulation task in pure APL, it was crazy

Magic numbers for tags and tag-dependent destructuring assignments all over the place

Rudra Joshi

Can someone explain me what is entropy

Adám

@RudraJoshi Are you sure you're in the right room? (Also, did you try Wikipedia?)

Bubbler

This isn't a general programming chatroom (or general "anything" chatroom)

Wong Jia Hau

@Bubbler I would imagine you have the first column as the tag, and the rest of the columns as the properties, is this right?

Adám

@Bubbler General APL, as opposed to specific (implementation/dialect of) APL :-)

Wong Jia Hau

7:27 AM

@Adám so some APL has ADT?

Adám

What is an ADT?

Wong Jia Hau

Algebraic Data Type, basically sum and product type

Adám

I don't know enough CS to answer. Sorry.

Wong Jia Hau

7:48 AM

When do we usually use a hook?

rak1507

in APL? never :P

Adám

@WongJiaHau Any time you want to somehow combine the original value with a processed value. Monadic example: IsPalindrome←⊢≡⌽ or IsPalindrome←≡∘⌽⍨

Dyadic example: IsReverseOf←≡∘⌽

@rak1507 Not true. X f∘g Y is a hook.

rak1507

to me a hook means the (f g) form

Adám

That's a mistake. A hook is the compositional form, not its notation.

rak1507

Ah, ok

Adám

7:52 AM

See apl.wiki/Hook

Wong Jia Hau

8:08 AM

how to understand this? ≡∘⌽⍨

Adám

@WongJiaHau It is monadic, so ≡∘⌽⍨ → {≡∘⌽⍨⍵} → {⍵≡∘⌽⍵} → {⍵≡⌽⍵}

Wong Jia Hau

8:30 AM

≡∘(⌽⍨) is this correct?

Adám

@WongJiaHau Not sure what you're asking. Are you asking if this is how ≡∘⌽⍨ binds? Then no, it isn't.

Wong Jia Hau

oh then what is the correct binding?

Adám

(≡∘⌽)⍨

Operators have long left scope and short right scope, mirrored from what functions have.

@Mast Hi there. Interested in APL?

Wong Jia Hau

oh no wonder

is (f∘g) x the same as f g x?

suppose f and g are functions and x is an array

Adám

Yes.

Wong Jia Hau

8:38 AM

Ah so when use dyadically, f∘g forms a hook?

Adám

Yes.

Wong Jia Hau

Thanks

Adám

Maybe I should add the monadic hook (f∘g⍨) to APLcart too?

Wong Jia Hau

@Adám yea that would be great

Adám

8:57 AM

@WongJiaHau Added.

Wong Jia Hau

@Adám thanks

can you also help me review the examples I added here? aplwiki.com/wiki/Beside#Examples

Adám

@WongJiaHau Done.

Wong Jia Hau

thanks, oh no there are two types of hook ?!

i just realized monadic hook and dyadic hook is different @.@

Adám

Well, "hook" is an artificial name we can decide the meaning of. Some clever people decided that X f g Y and Y f g Y should both be called "hook".

Similarly with "beside": Some clever people decided that X f g Y and f g Y should be called "beside".

Similarly with "atop": Some clever people decided that f X g Y and f g Y should be called "atop".

Similarly with "over": Some clever people decided that (g X)f(g Y) and f g Y should both be called "over".

Wong Jia Hau

9:16 AM

so basically all kinds of trains has two variants: monadic and dyadic. is this correct?

Adám

There are infinitely many possible compositions of two functions, both for dyadic application and for monadic application. We can call them what we want (though a consensus would be nice) and we can give any subset of them whichever names we want (though a consensus would be nice), and when we do that, we can let arbitrary pairs of monadic and dyadic compositions share a glyph (though a consensus would be nice).

@WongJiaHau All kinds of derived functions, yes. Trains are but a subset of derived functions.

Wong Jia Hau

@Adám ah, so I just have to treat a train as a function

Adám

A train is a function.

Wong Jia Hau

ok need to wrap my head around that, don't know why I think of them as different stuff all along

Adám

Language Features of Dyalog version 18.0 in Depth - Part 1 – Dyalog Webinar with Adám Brudzewsky

►

@WongJiaHau Trains are special in being compositions of three functions, while all other APL compositions are of two functions. (Though you could also look at ⍨ as forming a type of composition, albeit with a single function.)

Wong Jia Hau

9:25 AM

@Adám thanks

Adám

@WongJiaHau Have you wrapped your head around the two atop forms being equivalent? (f g) vs f⍤g

rak1507

do you know how you can display two histograms on the same chart?

Adám

@rak1507 You mean with SharpPlot?

rak1507

yes

when I just try and add another I get 'multiple expressions mapping to the same series of chart1'

Wong Jia Hau

@Adám I'm haven't read about the operator yet

rak1507

9:28 AM

oh, if I just add a new chart it overlays, nvm

Wong Jia Hau

can't seem to find the wiki page about ⍤

Adám

@WongJiaHau Well, you know about monadic f∘g – they are the same monadically.

@WongJiaHau Someone has to write those articles… Meanwhile, the documentation exists.

Wong Jia Hau

o i see, so I guess the 2 dots above the jot implies dyadic?

Adám

No, the similar symbols just mean they are related.

Wong Jia Hau

What's the name for ⍤?

Adám

9:31 AM

The symbol or the primitive?

Wong Jia Hau

what's the difference between small mouth and big mouth

Adám

Have a look at apl.wiki/Function_composition

Wong Jia Hau

@Adám primitive

Adám

Atop.

Wong Jia Hau

ok so small mouth is called atop, big mouth is called over?

and the eye-less is called beside?

Adám

9:33 AM

Yes, but their English names are not very important.

Wong Jia Hau

Why?

Adám

I think of f∘g as "use g to pre-process the right argument of f". I think of f⍥g as "use g to pre-process all arguments of f". I think of f⍤g as "use f to post-process the result of g".

@WongJiaHau When you converse in English, do you think it is important that w is called "double-yoo"?

Wong Jia Hau

If I need to refer to it specifically, for example when spelling a word verbally, then yes

in that way, f⍥g if use monadically, it is the same as f∘g right?

Adám

Yes, but I recommend using ⍤ for the monadic case whenever possible.

@WongJiaHau Anyway, once you see that (f g) is just a choice of interpreting two consecutive functions in isolation as f⍤g then you might ask if it'd be possible to define one (or two) operators so that (f g h) might be written with operators in the "spaces" rather than through a series of three consecutive functions.

And yes, it is possible:

     (⊢   -   (+/   ÷   ≢)) 3 1 4 1 5
0.2 ¯1.8 1.2 ¯1.8 2.2
      ⊢_fk-fk_(+/_fk÷fk_≢)  3 1 4 1 5
0.2 ¯1.8 1.2 ¯1.8 2.2

Here, _fk and fk_ are dyadic operators (which you can find in the dfns workspace or on APLcart).

Wong Jia Hau

9:50 AM

Is ^/2</x y z the same as the math notation x < y < z?

Adám

Uh, yes, sure, if x and y and z are scalars.

I'd prefer to write (x<y)∧(y<z)

CMC: Rewrite monadic ⊢-+⌿÷≢ as a derived function by reordering the symbols and adding ∘s and ⍨s and parentheses.

Wong Jia Hau

I see. What is CMC.

rak1507

I assume -∘(⌹∘(=⍨)⍨)⍨ is cheating?

Adám

@WongJiaHau Chat Mini Challenge.

@rak1507 Yes, as that doesn't abide by the spec of "reordering the symbols and adding ∘s and ⍨s and parentheses."

Wong Jia Hau

\○/

rak1507

9:57 AM

:(

Adám

I've got 17.

Wong Jia Hau

⊢-+⌿÷≢ this means negative mean right? just want to confirmed before i start

rak1507

-∘(÷∘≢⍨∘(+⌿)⍨)⍨ then

Adám

@WongJiaHau No, it means subtract the mean from the argument.

@rak1507 Same as mine.

Wong Jia Hau

oh i have to treat - dyadically?

rak1507

10:00 AM

@Adám that's 15 not 17 :P

Adám

You're missing ∘⊢

rak1507

huh?

what does that do

Adám

@WongJiaHau Yes. You may want to review some of the tutorials on trains to get how longer trains work.

@rak1507 Nothing.

rak1507

oh, you have to use all the symbols in there?

Adám

> reordering the symbols and adding ∘s and ⍨s and parentheses.

Wong Jia Hau

10:03 AM

@Adám the challenge means to rewrite the expression as trainless right?

Adám

Yes.

xpqz

@WongJiaHau I wrote a bit on trains, too: xpqz.github.io/learnapl/tacit.html

dzaima

@Adám don't you need another ⍨?

Adám

@rak1507 Really, the challenge is to rewrite an arbitrary (monadic) sequence of functions to a derived function using ∘ and ⊢ and parens.

rak1507

it's not too hard

x - (+⌿x)÷(≢x)
x - f x
-∘f⍨ x
(+⌿x)÷(≢x)
(+⌿x)÷∘≢x
x÷∘≢⍨+⌿x
÷∘≢∘(+⌿)⍨ x
-∘(÷∘≢∘(+⌿)⍨)⍨

Adám

10:05 AM

@dzaima How so?

Wong Jia Hau

@xpqz so (a b c d) means (a (b c d)) right assuming a b c d are functions

rak1507

yes

dzaima

@Adám -∘(÷∘≢⍨∘(+/)⍨)⍨∘⊢⍨

rak1507

-∘(÷∘≢⍨∘(+/)⍨)⍨∘⊢ works fine

Wong Jia Hau

so trains are also right-associative is this correct?

Adám

10:06 AM

Yes.

dzaima

@rak1507 it works fine, but it does (⊢-(+⌿÷≢)∘⊢) and not (⊢-+⌿÷≢)

Wong Jia Hau

ok, so function call and trains are right associative, while operator call is left-associative

Adám

@rak1507 Hm, so we zigzag, while the right and middle gets ⍨ and parenthesise the middle?

f g h i j
g∘(i∘j⍨∘h⍨)⍨∘f⍨

@benWindsorCode Welcome to the APL Orchard. I see that you've welcomed feedback on your answer:

3

A: "-rot" transform

APL (Dyalog Classic), 50 bytes {⎕←1↑⍵⋄({(3↓⍵),⍨(+/2↑1⌽3↑⍵),1↑⍵}⍣{⎕←1↑⍺⋄(⍴⍺)<3})⍵} Try it online! Explanation: (+/2↑1⌽3↑⍵),1↑⍵ take the first three elements, rotate them, sum the first two and stick the third (which was originally the first element) on the end again so (a b c) -> (b c a) -> (b+...

rak1507

10:35 AM

would be a lot shorter porting jelly probably

Wong Jia Hau

10:46 AM

ok so the formula to translate a (f g h) fork using on ⍨ and ∘ is to turn it into (g∘h⍨∘f⍨)?

rak1507

yep

Wong Jia Hau

With explicit parenthes: (((g∘h)⍨∘f)⍨)

ah my brain

Adám

@WongJiaHau Actually, g⍨∘f⍨∘h would be more correct (and slightly easier to read, imo).

rak1507

      (f g h) 3
¯4
      g∘h⍨∘f⍨ 3
¯4
      g⍨∘h⍨∘f⍨ 3
4
      g⍨∘f⍨∘h 3
10

Adám

Huh.

Wong Jia Hau

10:51 AM

@rak1507 wait so nobody is correct now?

rak1507

@WongJiaHau g∘h⍨∘f⍨ is right

Adám

No, g∘h⍨∘f⍨ gives the right result.

rak1507

(f g h) x
(f x) g (h x)
(f x) g∘h x
x g∘h⍨ f x
g∘h⍨∘f⍨ x

Adám

 (f g h) x
 (f x) g (h x)
 (h x) g⍨ (f x)
 (h x) g⍨∘f x
 x g⍨∘f⍨ h x
 x g⍨∘f⍨∘h x
 g⍨∘f⍨∘h⍨ x

Where's my mistake here?

rak1507

that works

you must have just missed the ⍨ at the end

Adám

10:55 AM

@rak1507 Oh, right copy-paste error. It is indeed in my APL session ⍨

Now, as to why g⍨∘f⍨∘h⍨ is more correct than g∘h⍨∘f⍨…

Wong Jia Hau

why?

Adám

      f←{⎕←'f'}
      g←{⎕←'g'}
      h←{⎕←'h'}
      (f g h)'a'
h
f
g
      g⍨∘f⍨∘h⍨'a'
h
f
g
      g∘h⍨∘f⍨'a'
f
h
g

Or in other words, the right tine is applied before the left tine.

rak1507

ah, application order

I really hope no one uses side effects in trains but inevitably someone will...

Adám

I do. It is really useful for avoiding lingering names.

Wong Jia Hau

what does 'lingering name' means?

rak1507

11:00 AM

:(

Adám

I might want some operation that analyses existing names, so I can't use temporary variables.

Wong Jia Hau

I see

is there some settings that I can use in TryAPL such that I can see the evaluation of an expression step by step?

Adám

@WongJiaHau Not really, but just Entering a tacit function (train/derived) will show you a tree structure: ⎕←⊢-+⌿÷≢

TryAPL

@Adám
┌─┼───┐
⊢ - ┌─┼─┐
    ⌿ ÷ ≢
  ┌─┘
  +

Adám

⎕←-∘(÷∘≢⍨∘(+⌿)⍨)⍨∘⊢

Wong Jia Hau

11:04 AM

but seeing this isn't enough for me to really understand g⍨∘f⍨∘h⍨

TryAPL

@Adám
         ∘
        ┌┴┐
        ⍨ ⊢
      ┌─┘
      ∘
     ┌┴┐
     - ⍨
     ┌─┘
     ∘
   ┌─┴─┐
   ⍨   ⌿
 ┌─┘ ┌─┘
 ∘   +
┌┴┐
÷ ≢

Adám

@WongJiaHau Not even ⎕←{g}⍨∘{f}⍨∘{h}⍨?

TryAPL

@Adám
        ⍨
      ┌─┘
      ∘
     ┌┴┐
     ⍨ {h}
   ┌─┘
   ∘
  ┌┴┐
  ⍨ {f}
┌─┘
{g}

Adám

Read this from the top…

Elias Mårtenson

In Dyalog, is ∘ an operator?

rak1507

11:06 AM

yes

Adám

@EliasMårtenson ^ except in ∘.

Wong Jia Hau

the part that I don't understand is how in the end g will receive the result from f and h

Elias Mårtenson

@Adám Yes, of course.

rak1507

g⍨∘f⍨∘h⍨ x
x g⍨∘f⍨∘h x
x g⍨∘f⍨ h x
(h x) g⍨∘f x
(h x) g⍨ (f x)
(f x) g (h x)

Adám

@WongJiaHau The overall function has ⍨ at the outermost layer: g⍨∘f⍨∘h⍨x is x g⍨∘f⍨∘h x

Elias Mårtenson

11:08 AM

@Adám So is ∘ also a function (like a "null function")? Or is it considered special syntax?

In the case of ∘.FN

Adám

Special syntax.

Elias Mårtenson

Thanks.

Adám

In Sharp APL, it ∘ was a regular array just like ⍬

Wong Jia Hau

@EliasMårtenson see aplwiki.com/wiki/Outer_Product

Adám

Or even better: apl.wiki/Jot

Elias Mårtenson

11:09 AM

@WongJiaHau I'm familiar with how these operators work. I wanted to know how it's implemented on the Dyalog side.

Adám

If I was implementing an interpreter for this (though of course I would choose to add this wart) I'd start by replacing all ∘.s

Elias Mårtenson

Kap implements ∘ as a function (called NullFunction), which raises an error when called. The product operator has a special case where it checks if the left function is NullFunction and delegates to outer product if it is (outer product is actually called ⌺ in Kap, with the NullFunction hack mainly being there for the purposes of APL compatibilty.

@Adám Replace, how? You mean having a dedicated outer product operator?

rak1507

@Adám +∘.1

Adám

@EliasMårtenson But then you can use ∘ for composition, arguably a much more important/powerful usage, especially since you have ⌺ for outer product and it can also be written using ⍤.

Elias Mårtenson

@Adám I already have ∘ for composition as well.

Adám

11:14 AM

@rak1507 Ugh, yeah, OK, token formation first, then replace ∘.

@EliasMårtenson Then what syntactic role does ∘ play then?

Elias Mårtenson

@Adám No syntactic role. It's both an operator and a function.

I was asking about Dyalog whether it implemented ∘ the same way, or if the parser itself has special knowledge of it.

A hybrid… SOS.

SOS?

Save Our Souls.

Hah, well. There is already /

dzaima

11:15 AM

@EliasMårtenson dzaima/APL just has a special parser rule for ∘., which is implemented as a regular monadic operator. ∘. doesn't have any ∘ or . anywhere

Elias Mårtenson

But yes, I would actually like to get rid of the function role for ∘. It would make things a lot more clean.

I only added it at some point to be APL compatible :-)

Adám

@EliasMårtenson KAP is already so not-compatible, I think you could just get rid of ∘.

Elias Mårtenson

@Adám Thank you for the encouragement. I think I will.

Adám

But I'm not sure I agree on using ⌺. Seems very heavy for such a basic thing.

rak1507

wdym by 'heavy'

Wong Jia Hau

11:17 AM

x g⍨∘f⍨ h x
(h x) g⍨∘f x
@rak1507 oh I know what I'm missing

Elias Mårtenson

@Adám "heavy" as in its shape?

Adám

Yes. Lots going on in that symbol.

Wong Jia Hau

I just found out that ⍨ means swap when use dyadically

Adám

@EliasMårtenson Remind me, did you split the other hybrids?

@WongJiaHau ⍨

dzaima

@EliasMårtenson you can keep the syntax but get rid of any role funkiness by just handling it as a special case in the parser

Elias Mårtenson

11:18 AM

Agree. The multiplication table looks ugly ×⌺⍨⍳10

Wong Jia Hau

but I still don't get it, the parsed tree clearly doesn't show that ⍨ is being used dyadically

Elias Mårtenson

@Adám You mean / ?

No I didn't.

Adám

@EliasMårtenson Well, either way, you could overload ⍀ and/or \ for outer product.

I'd highly recommend splitting them as dzaima has done.

There's no need for the trailing axis editions if you have Rank, so the dashed symbols can be a pair, and the non-dashed the other pair.

Elias Mårtenson

@Adám I could, but it doesn't look like a product :-) I picked ⌺ partly because it's a rectangle.

Adám

I think 1 2 3 ×\ 1 2 3 4 looks fine.

dzaima

11:21 AM

dzaima/APL has ⌾ as a proper monadic operator for ∘. fwiw

Elias Mårtenson

It does.

@dzaima Oh... Pretty.

dzaima

I guess i thought it was . + ∘, but it's ∘ + ○

Adám

⊙ exists.

dzaima

@Adám so it does

Elias Mårtenson

@Adám by the way, in the recent arraycast, you mentioned a palindromic palindrome checker, where each symbol was vertically symmetric. While driving to the office, listening to that episode, I was trying to figure out what that looked like, but failed. Clearly the symmetric symbols are ⌽ and ≡, but how to make it palindromic?

Adám

11:25 AM

@EliasMårtenson It is in the show notes ("0:54:50 Palindromic Palindrome Checker in APL"): ⌷≡⌽⌽≡⌷

rak1507

⌷≡⌽⌽≡⌷ iirc

ah, beat me to it

Elias Mårtenson

@dzaima Such pretty symbols. If I didn't have such a severe case of decision anxiety, I'd use all of them.

@Adám Wait what? OK, give me a moment to parse that one.

Adám

@EliasMårtenson BQN uses ⌜ for outer product. Another candidate was ⟔

I think these are good because they show how the rank is extended based on the args.

Elias Mårtenson

Ah, BQN's choice isn't bad at all.

Thanks for ⌷≡⌽⌽≡⌷. It revealed a bug in KAP.

Adám

@EliasMårtenson Wouldn't ⊡ be better?

(Just don't confuse it with 🝕 which stinks.)

rak1507

11:29 AM

@EliasMårtenson monadic ⌷ is kinda pointless

Adám

What does monadic ⌷ do in KAP?

Elias Mårtenson

@Adám It definitely represents the action better than ⌺

@Adám Nothing. That's the bug.

Not implemented.

Adám

Well, what would you have it do?

Elias Mårtenson

The same as Dyalog.

rak1507

⌷ in dyalog is just a no op

Adám

11:31 AM

No, it returns the elements of the default property of objects.

rak1507

so I wouldn't bother, unless you're interfacing with things and you actually do need the 'materialise' thing but you probably don't

Elias Mårtenson

It's identity isn't it?

Adám

Only for arrays that are not collection-type objects.

Elias Mårtenson

12:48 PM

What is monadic ≡ useful for? I've never needed it myself.

Adám

Mostly analysing arrays. E.g. asserting that an array is simple with 1≥|≡

Or checking that you've got a vector of simple vectors (1=≢⍤⍴)∧(2=≡)∧(1∧.=≢∘⍴¨)

Elias Mårtenson

Thought so. So the main use is just to check if it's =0 or >0. In KAP, a much more efficient way to find that would be {1∊{⍬≢⍴⍵}¨⍵}

Mast

@Adám Not really, I was just taking a look at the room.

Elias Mårtenson

Perhaps I should make ≡ return only 0 or 1. 0 for simple vectors, an d1 for compound.

Mast

I don't need a golfing language to write incomprehensible programs.

;-)

Adám

1:01 PM

@Mast APL is not a golfing language, though.

@EliasMårtenson Well, in the above, I check for 0, 1, and 2…

Dyalog's version can also be used to check for raggedness with 0>≡

Elias Mårtenson

@Adám I see.

Mast

@Adám I'm sorry.

Using mathematical notation to program must be a joy as a mathematician. For simpler people like me, it looks quite complicated.

Elias Mårtenson

@Mast APL can be difficult to learn. But it may surprise you that it's not actually the notation that makes it difficult.

Adám

@Mast I'd argue that APL is essentially a simpler language than many popular ones.

Elias Mårtenson

The main problem is that you have to think about solving problems in a different way.

Mast

1:06 PM

And about language in a different way, if you're unfamiliar with the notation.

Elias Mårtenson

Indeed.

Adám

@Mast APL is actually quite similar in structure to English (and probably to Dutch too then).

Elias Mårtenson

You can learn the syntax of the language in a few hours. That's really easy. The difficulty comes when you put the pieces together, since you do it in a way that is quite different from most other languages.

Mast

@Adám Considering Dutch is more complicated than most programming languages I know, I'm not sure that's a good thing.

Elias Mårtenson

But, it's also very satisfying when you do.

@Mast Well, I find Dutch remarkably readable since I already speak Swedish and some German. However, coming to APL from, say, Java is like coming to Dutch if all you know is Chinese.

I'm sure @Adám is going to criticise me for making it seem like APL is the super-difficult thing.

Mast

1:09 PM

Besides, I have too many hobbies already and no use-case for APL. I'm actually quite terrible at programming, despite the many years of practice.

Elias Mårtenson

It's not true actually, like @Adám said, it's very simple.

@Mast then perhaps you should give it a go? I've heard it mentioned in many places that the people who find it easiest to learn are people with no pre-conceived notions of other languages.

Mast

That's probably true. If you don't know it's supposed to be hard, it usually isn't.

And if it looks more complicated than it is, it's perfectly learnable indeed.

Adám

@Mast If you've struggled with programming for years, APL might actually turn out a good fit for you.

However, @EliasMårtenson we might come across as pushy, so I'll be quiet now.

Mast

Don't worry about it :-)

Elias Mårtenson

@Adám have you ever visited a Haskell forum? :-)

There is this famous chat log from #haskell on IRC, where a troll comes in trying to make fun of the people there. The log ends with the guy being converted and starts to experiment with it. :-)

Mast

1:31 PM

I've tried Haskell. I didn't come so far that I understand it, but at least now I know the basics and why I don't understand it.

Kamila Szewczyk

@EliasMårtenson can you please link it

i really want to read it

Elias Mårtenson

@KamilaSzewczyk gist.github.com/quchen/5280339

Razetime

1:45 PM

this is an amazing chatlog

@Mast if you do consider trying out APL, feel free to ask for help and leanring resources here.

user

2:07 PM

@EliasMårtenson Where did this happen lol

Elias Mårtenson

@user It was on #haskell on Freenode at the time, I'd guess.

user

Ah

Junyu Zhan

I'm using Dyalog on a mac. The default way to type symbols is to use the ` prefix. I'm wondering how do most APLers type symbols?
(I have tested on Windows and the default way is to use Ctrl as a combine key.)

Razetime

@JunyuZhan I use the AltGr keyboard from @Adám on Windows: github.com/abrudz/Kbd

alt is a more convenient key than ` so i got used to it much easier

Elias Mårtenson

2:43 PM

@JunyuZhan Depends on tool. For GNU APL I use gnu-apl-mode which uses the super key. In Dyalog I use Ride, which loads a Linux keymap that also uses Super. However, that keymap is a bit flakey so having the ` prefix is helpful.

In KAP I use the Alt key, as part of the UI.

LdBeth

@JunyuZhan see aplwiki.com/wiki/Typing_glyphs , and if you are using Mac there's a freeware called Ukelele that enable's you create your own keyboard layout

I used that to create an APL layout for JIS keyboard, which I guess no one else would use that :P

Elias Mårtenson

3:14 PM

After implementing monadic ≡ the palindromic palindrome checker function now works: ⊣≡⌽⌽≡⊢

I find it prettier than the Dyalog version :-)

This works because in KAP, both monadic ⊣ and ⊢ are identity.

user

3:33 PM

Aren't they both identity in Dyalog too?

Elias Mårtenson

@user Yes, I just realised that.

They are in GNU APL (although one of them tags the result as non-printed (just like assignment)

For some reason I thought only ⊣ was identity in Dyalog.

user

Odd (GNU APL's behavior, that is)

Elias Mårtenson

@user It's quite useful. If you call a function for side effects that returns something large, you might not want to print it.

In Dyalog you need to do something like 0⊣somethingBig ...

user

But couldn't you just do _←big thing though?

Or that, it's just one extra character

Elias Mårtenson

But then you'll end up assigning it.

user

3:38 PM

Yours works, though (0⊣)

Elias Mårtenson

I think the idea in GNU APL is basically, "why not".

user

lol

Elias Mårtenson

I mean, GNU APL is very bound to its APL2 compatibility, which is fine. I do want to give them credit for some the good ideas they have, and the special behaviour of ⊣ is one of them.

At least in my opinion.

The idea of allowing double quotes is also a good idea. Where "a" is the same as ,'a'

That way strings with double quotes are consistent in a way that traditional single-quotes strings are not.

user

Yeah, it's very annoying when you have 'a' 'b' 'c' and it gets turned into 'abc'

dzaima

@EliasMårtenson dzaima/APL does that fwiw

Elias Mårtenson

3:48 PM

@dzaima Good. It's the correct thing to do.

Does it use double quotes for proper strings like GNU APL?

dzaima

@EliasMårtenson if that means escape parsing, then yes

Elias Mårtenson

I mean, is ⍴"a" equal to 1?

And what is ⍴'a' ?

dzaima

@EliasMårtenson yes, that's what my original message was replying to

Elias Mårtenson

@dzaima Got it, thanks.

dzaima

@EliasMårtenson ⍬

Elias Mårtenson

3:52 PM

The way it should be.

I simply didn't implement 'a'

:-)

dzaima

so you just don't have single character constants?

Elias Mårtenson

@dzaima I do, but it's prefixed with @

so @f @o @o is "foo"

Yes yes, it's the worst syntax ever.

Problem is that ' is taken for quotation like Lisp.

And ` is taken for line continuation.

Point is, you're not really supposed to work with single characters, since it makes Unicode sad.

So you don't see it very often in practice.

dzaima

@EliasMårtenson when you have constants is the one time when it makes sense to work with single characters though. Splitting on spaces or counting '('s is always safe regardless of unicode

Elias Mårtenson

@dzaima Yeah, fair enough.

I guess I was trying to justify the horrific syntax.

I mean, I really have nothing else :-)

benWindsorCode

9:16 PM

Looking through code of the mystika repository I found the two function definitons:
mean←+⌿÷≢
var←×⍨∘mean-⍨∘mean×⍨

the var definition is the one I am confused about, at first I thought it was a three train with the '-' in the middle, but printing it out using ]boxing -trains=tree, I dont see it behave like that

my question is, given a right hand side argument, say '1 2 3 4', how does this '1 2 3 4' make it to both sides of the '-' operation, in particular how does it get all the way through to the mean on the left hand side?

I think I'm just not unwrapping the tacit code correctly, but the tree didnt' shed light for me intuitively

dzaima

×⍨∘mean-⍨∘mean×⍨ → (×⍨∘mean)(-⍨∘mean)(×⍨) → (×⍨∘mean ⍵)(-⍨∘mean)(×⍨ ⍵) → (×⍨ mean ⍵) (-⍨∘mean) (×⍨ ⍵) → (×⍨ mean ⍵) -⍨ mean (×⍨ ⍵) → (mean ×⍨ ⍵) - (×⍨ mean ⍵)

rak1507

@benWindsorCode what a coincidence, I was writing some stats stuff earlier today

imo my version var←+.×⍨⍤-∘mean⍨÷≢ is nicer :P

benWindsorCode

@dzaima Thank you very much, that certainly helps!

dzaima

i'd personally write (mean×⍨) - ×⍨∘mean

benWindsorCode

@rak1507 interesting!

some nice range of functions here that I was browsing, is Aaron Hsu's work github.com/Co-dfns/mystika/blob/master/chron.dyalog

Transcript for