The APL Orchard

code_report

03:49

Anyone have a shorter (or more idiomatic) way of determining whether a number is a power of two in APL? My solutions: github.com/codereport/LeetCode/blob/master/…

Razetime

      f←(⊢=⌊)2∘⍟⋄f ⍳8
1 1 0 1 0 0 0 1

      f←0=1|2∘⍟⋄f ⍳8
1 1 0 1 0 0 0 1

that is 7 bytes

Bubbler

04:24

For pure golfing purposes, ⊢∊2*⍳ works at 5 bytes, assuming ⎕IO←0

Since APL doesn't have bitwise ops (which are to be simulated with bit arrays, which involves 2∘⊥⍣¯1 anyway), seeing if the bit count is 1 is probably the most idiomatic way

Given the input range, I think you can do (33⍴2)⊤ instead of 2⊥⍣¯1 to avoid positiveness check

kimmolinna

@Adám I can change those. I didn't use ] because those are not user commands in Dyalog. I have three commands - vegalite, markdown and multiline. Vegalite and markdown indicates how to render the following output.

Bubbler

@code_report Also your first solution 1=(+/2∘⊥⍣¯1)∧0∘< is logically weird (though technically correct). I think you meant 0∘<∧1=(+/2∘⊥⍣¯1) (and (+/...) can be golfed to 1⊥)

A solution using (33⍴2)⊤ would be 1=1⊥(33⍴2)⊤⊢

Bubbler

04:49

Another one for creativity: (<\≡⊢)(33⍴2)⊤⊢

Adám

08:42

@Bubbler @code_report Fails on negative numbers.

@Razetime These fail on 0.

@kimmolinna 1) they could become (]html started out exactly like that, but was ported to the normal product). 2) they are certainly not keywords like :return.

kimmolinna

@Adám I could live with the present .apla format if Dyalog editor would open it in a proper format.

Adám

@kimmolinna You mean the stand-alone Editor (on Windows)?

kimmolinna

@Adám I already changed those. Now I have ]vegalite, ]markdown and ]multiline

@Adám Yep.

Adám

@kimmolinna Nice.

@kimmolinna Right, I understand. I think there's a better chance we'll start representing plain texts as normal text files. I'll take it up with the involved, and let you know when I have news.

Razetime

09:00

@Adám ah make sense

Adám

@code_report More idiomatic is, imo, {1=+⌿⍵⊤⍨33⍴2} because it handles a numeric array of any shape.

(This is equivalent to Bubbler's 1=1⊥(33⍴2)⊤⊢ but maybe easier to read.)

Also, Bubbler's is fastest, even on individual numbers.

      n←¯256+?1e3⍴512

      ]runtime -c (0∘<∧1=1⊥2∘⊥⍣¯1)¨n (1=1⊥(33⍴2)⊤⊢)¨n
  (0∘<∧1=1⊥2∘⊥⍣¯1)¨n → 1.6E¯3 |   0% ⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕
  (1=1⊥(33⍴2)⊤⊢)¨n   → 5.1E¯4 | -68% ⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕

      ]runtime -c (1=1⊥(33⍴2)⊤⊢)¨n (1=1⊥(33⍴2)⊤⊢)n
  (1=1⊥(33⍴2)⊤⊢)¨n → 5.1E¯4 |   0% ⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕
  (1=1⊥(33⍴2)⊤⊢)n  → 5.1E¯6 | -99%

Adám

09:55

Hm, I really want ⍛ (reverse-compose or before), so things using the pattern (f/g⊢) can be written as f/⍛g

dzaima

wouldn't the full version there be (f/⍤⊣g⊢)?

Adám

I had monadic application in mind.

dzaima

oh, so f/⍛g⍨

Adám

Yes, typo.

dzaima

also the very similar-looking f⍛/⍨ is awesome

Adám

10:00

Oh yes, that's "filter by function".

Probably many such constructs we don't think of because ⍛ is missing.

f⍛⊂⍨ partition by.

dzaima

/ is the primary one that comes to mind in BQN (where filter by function is of course F⊸/)

Adám

Right.

With sane indexing as ⊇, f⍛⊇⍨ is another filter, for functions that give indices.

dzaima

yeah, was about to mention that

Adám

But even without that, ⊂⍤f⍛⌷⍨ works.

dzaima

fwiw, mlochbaum/BQN contains 1862 ⊸s and 902 ⟜s. But many of those are with array operands, which would both translate to ∘ in APL

chrispsn

12:28

@Adám i've been pondering a (f)! dict creation overload to enable this kind of thing in k (defined as {y!x y})

so partition by length would be =(#')!

Adám

I don't understand how that can parse right.

How wouldn't it end up applying monadic !?

chrispsn

@Adám which bit?

(f)! would be noun verb which is ok

Adám

So parenthesised verb is a noun?

chrispsn

like 3+

@Adám yep

Adám

Ugh.

dzaima

12:32

@Adám that's a core k thing

Adám

I don't have to like it.

chrispsn

(also i meant 'group by' in my message earlier)

dzaima

13:52

did some grep.app-ing of APL chars, and apparently this is the most common APL expression on GitHub

Adám

14:04

@dzaima What in the world is ⌷←⍳→⍴∆∇⊃‾⍎⍕⌈?

Razetime

something in a library common to all those repositories

Adám

Yeah, seems to be a font or encoding test .

((V⍳V)=⍳⍴V)/V←,V is a classic idiomatic expression, though.

But today, people would use ∪V if they have it available.

I guess it is good that many systems check to make sure they support APL chars.

Elias Mårtenson

15:26

Where can I find the documentation for Dyalog ⎕S and ⎕R?

Never mind. Found it.

Adám

@EliasMårtenson ]help ⎕R

@hyper-neutrino Bot died?

@EliasMårtenson also https://aplcart.info?q=⎕R operator.

hyper-neutrino

⎕←42

TryAPL

@hyper-neutrino 42

hyper-neutrino

since it appears most of the time the bot goes down, the program itself actually crashes, i can probably just surround this in a while loop

Adám

⍨

hyper-neutrino

15:29

something about websocket unexpectedly closing

alright hopefully this shouldn't happen again, unless some other strange error causes it to stop working but doesn't crash the program

Elias Mårtenson

I demand `⎕IO←1.2

Or ○1

Adám

If that was possible, how would one use fractional indices for things like laminate and mix?

Elias Mårtenson

@Adám Thank you. It was an answer like that that I would have hoped for. I need something to think about that is not so serious :-)

rak1507

I wish 1 2 3 ⍪ 4 5 6 was 1 2 3 ,[⎕IO-.5] 4 5 6, is there any sensible reason why this couldn't be the case?

Elias Mårtenson

Also, complex IO

rak1507

15:40

⍉⍤,⍥⍪ is really ugly

Adám

@rak1507 Do you mean + rather than ← there?

rak1507

@Adám oops, meant -

Adám

@rak1507 Well, you wouldn't want to lose ⍪ but BQN has ≍ for exactly this.

dzaima

@rak1507 hmm, isn't that question kind of similar to the question of whether 1,2 should error?

rak1507

@dzaima similar

dzaima

15:42

@Adám you'd still have ⍪, it'd just do something different for rank 1, like , does for rank 0

rak1507

^

xpqz

Anyone else fancy reading my APL-book? Caveat: it contains the solution to one Dyalog comp part 1 problem.

rak1507

@xpqz if I can prove to you that I've solved that problem can I read it?

Adám

@dzaima That'd make it really hard to write rank-agnostic code.

rak1507

(and then if I submit your solution they tell me off)

xpqz

15:42

I take your word for it...

@rak1507 can you msg me an email address?

dzaima

@Adám I have a hard time imagining it'd be more of a problem than many other things. (and, as you may know, i don't particularly like suggesting writing rank-agnostic code)

Elias Mårtenson

@Adám I think the laminate question is easy to answer though. If ⎕IO←1.2, then you would do A ,[0.6] B

rak1507

@xpqz how? I don't really want to post my email publicly here, maybe I can email you?

Elias Mårtenson

(that's assuming that the stride length is always 1. Clearly we need a ⎕SL←N setting that allows you to configure this)

Adám

@xpqz @rak1507 I can connect you by email.

rak1507

15:45

alright, thanks

xpqz

@Adám Sure works for me

Adám

@dzaima It'd be awful for things like PDI etc.

No, the real problem is that , is last-axis. In a fresh APL, , should be first-axis, and then ⍪ could do this coupling.

I.e. , and ⍪ could be BQN's ∾ and ≍ (everything dyadic here).

dzaima

@EliasMårtenson so A ,[1.2] B would be a non-laminate, and A ,[¯1+1+1.2] B would be laminate?

(not that that can't work, it's just a fun consequence (of course it won't practically be usable but that's not the goal here))

Elias Mårtenson

@dzaima well, laminate would be if the index is ⎕SL÷2 less than some index.

dzaima

i assume ⎕SL is like ⎕CT?

Elias Mårtenson

15:49

So with ⎕IO←1.5 and ⎕SL←10, then the indexes would be 1.5, 11.5, 21.5, ... And to laminate you'd have a value half-way in between, such as 16.5.

dzaima

so a me who always wants precise comparisons won't be able to use ⎕IO←1.2 without rounding errors messing with things :'(

Elias Mårtenson

@dzaima You're right. Clearly we need decimal arithmetic now.

dzaima

⎕IO←÷3 :P

Elias Mårtenson

Now... what about ⎕IO←3J1

Adám

⎕IO←○¯1j1 for complex irrational numbers.

Anyhow, I want ⎕IO←' ' – deal with it! ;-)

dzaima

15:55

⎕IO←⎕NS⍬

⎕IO←3⍴⎕NS⍬

Adám

Can we all agree on ⎕EX'⎕IO'?

Elias Mårtenson

@Adám You do realise I will now spend far too much time (i.e. more than zero) thinking about what that would look like?

dzaima

@EliasMårtenson with BQN-like char arith, it's not even too impossible

Elias Mårtenson

@dzaima Tell me more about BQN char arith

dzaima

if you had ⎕IO←⎕UCS 0, you'd even have the nice side-effect of negative indices being literally impossible

Adám

15:58

@EliasMårtenson mlochbaum.github.io/BQN/doc/types.html#characters

dzaima

@EliasMårtenson "BQN" ≡ "APL"+1‿1‿2

Elias Mårtenson

Oh nice

Adám

@EliasMårtenson It'd be awesome: ⍳95 for printable ASCII.

dzaima

but you're limited to arrays of length 1114112 :/

Elias Mårtenson

@Adám You know, it's not too far from how COBOL handles numbers :-)

Adám

15:59

@dzaima That should be plenty for most purposes.

Elias Mårtenson

Oh wait... there is a ring of complex numbers defined for + and ×, no? That suggests that ⎕IO for a complex number is OK as long you you limit yourself to those two operations.

dzaima

you generally need just + for indexes, no?

Elias Mårtenson

Well yeah

I mean, I'm not that great at math. I only know enough to pretend to understand basic classical mechanics and beginner QM :-)

In other words, I know enough to be practical, but number theory and that stuff is way beyond me.

dzaima

@EliasMårtenson i wouldn't even pretend to understand those

rak1507

indexing should clearly be based on points on an elliptic curve

Elias Mårtenson

16:04

@dzaima You could. Classical mechanics is mostly calculus, and basic QM is matrices, complex numbers and a few derivatives.

@rak1507 I think you're on to something there. An array lookup is basically the same operation as performing an EC encryption operation.

Let' call it Crypto-APL, or CRAPL for short.

Throw in a few blockchains as well and there is money to be made.

rak1507

lol

Adám

I had a potential customer ask me (in all seriousness): What do you think of blockchains?

Elias Mårtenson

Hang on a minute, surely someone must have marketed a "native blockchain programming language"

@Adám Hah, I've had colleagues ask that :-)

What did you answer?

Adám

I think blockchains are great for some specific problems, but certainly not for everything.

rak1507

@Adám in an APL context?!

@EliasMårtenson solidity is probably the closest thing

although I did see one the other day... hang on

Adám

16:13

@rak1507 Well, in the context of him wanting to create a startup and make money fast.

rak1507

startup and APL don't sound like they'd go together

Adám

OK, can we please not do another APL-bashing session now? I have other things I want to do.

Elias Mårtenson

@rak1507 I did a DDG search for "blockchain programming language", and while I expected some absolutely garbage articles... This was the number one hit. Warning, there is a rish of losing quite a few braincells after reading this... In fact, I suspected it was written by an AI, but even AI are better at generating actual words: mobcoder.com/blog/blockchain-programming-languages

rak1507

Fine :P

1. C++
oh no

Elias Mårtenson

@Adám I mean, was this recent? A couple of years ago you could make money by just starting a company with blockchain in the name.

Adám

16:16

@EliasMårtenson Yes, this spring.

Elias Mårtenson

Now, I'm no so sure.

@Adám Do people still think blockains is a "world changing" technology?

Adám

¯\_(⍨)_/¯

rak1507

doesn't have to be world changing to get dumb investors to throw money at it

Elias Mårtenson

Fair enough.

Sometime I wish I wasn't such an honest person. I'd be rich.

hyper-neutrino

@EliasMårtenson did they just like, list 10 popular languages and make up some BS connection to blockchain...

Adám

16:19

I could probably plan some pretty spectacular crime if I put my head into it. I could probably be a successful (‽) terrorist too.

rak1507

Not one for the CV I think...

Elias Mårtenson

@hyper-neutrino Pretty much, yes. Combine that with terrible grasp of English and absolutely no knowledge of the actual subject matter, and that's what you get :-)

The article is almost a work of art

RGS

“The only problem we encounter.”

I see.

rak1507

17:43

@Adám getting NONCE ERROR: __session_input__ is in use when I try and use :Section

oh it's when I try and re-enter a section

Adám

@rak1507 So, you solved it?

rak1507

no

I don't know how to run a section again

Adám

Just press Enter anywhere inside it.

@rak1507 What exactly did you do to provoke the error?

rak1507

hang on, will try again in a sec, for now a niladic tradfn will do

rak1507

17:59

oh, got it, it's when you hit enter from inside a dfn

      :Section
       f←{1}
       :EndSection
      :Section
       f←{1}
       :EndSection
NONCE ERROR: __session_input__ is in use

hit enter when the cursor is in front of the 1 in the dfn

Adám

18:15

@rak1507 I can repro in 18.0, but has been fixed in 18.1. That's OK, since it was very experimental in 18.0.

rak1507

20:06

@Adám Alright great, what's the behaviour in 18.1? Do you get to edit the dfn, or does it re-enter the section?

Transcript for