The APL Orchard

Bubbler

02:01

CMC: Tie or outgolf me in Evaluate Left-or-right using APL. I have 23-bytes solution in vanilla Dyalog APL.

rak1507

damn

I got it

Shame that < and > don't work , would be nice if there was some a w so a<w = a and a>w = w

Dyalog APL

<moon-child> easy, a←⊢

rak1507

02:16

Can't pass functions as arguments

Although I'm not sure what you mean

Bubbler

maybe it's HTML artifacts?

moon-child: How does the message (the one starting with "Shame") render to you?

rak1507

It's purely ascii so I don't know why there'd be a rendering issue

Bubbler

If <w = a and a> gets stripped, you just see aw = w

rak1507

Ohh

Improper filtering? weird

Bubbler

so the obvious answer is indeed a←⊢

rak1507

02:22

(does that work? lol)

Bubbler

@rak1507 Maybe add some explanation of how the code works and why?

rak1507

On it

Dyalog APL

<moon-child> oh, yeah, hmmm. Idk why the < and > got filtered out

<moon-child> oh, I see

Bubbler

(actually the same approach could be shorter in Jelly, as it has all the necessary built-ins to mimic it)

rak1507

argh, alignment is hard

Bubbler

02:32

Write explanation in TIO and copy that

They don't perfectly align when the monospace font doesn't support the chars anyway

rak1507

yeah

oh well it'll have to do, it's still obvious enough which bits correspond to what

fun challenge, I initially started writing something that would parse it recursively lol

Bubbler

I have a custom stylesheet that renders all the textarea and code in DejaVu Sans Mono

rak1507

Oh clever

Bubbler

It only works for me, but it works very well for me, so I just write all explanations to align in my view and hope the others understand

rak1507

I think I've fixed the alignment

Bubbler

02:38

@rak1507 The explanation is a little off. "length of the left argument" → "last index + 1", and the train is a dyadic train, so you should say x (f g h) y → (x f y) g (x h y) instead

rak1507

True

(⍎'(⊣⊢)'['(<>'⍳⍞])/'<>'
         '(<>'⍳⍞          take the input ⍞ and get the indices of '(<>'
                          this fills with the last index + 1 if not found, saving 1 char
  '(⊣⊢)'[ stuff ]         index into '(⊣⊢)', replacing <> with ⊣⊢, the left and right tack functions
 ⍎                        evaluate this as a train
                          APL trains of the form (f g h) when applied to left and right arguments x and y produce (x f y) g (x h y)
                          This translates directly to this language

decent?

So much easier to explain things using symbols rather than words :p

Bubbler

Looks good

rak1507

02:54

back to doing my physics homework... :(

Razetime

@rak1507 abusing APL trains is big brain

rak1507

@Razetime it's nice that it works out to be like that

Razetime

(Now that I finally got Electorate, I can upvote your answer yaya)

rak1507

What's Electorate?

Razetime

Gold badge

rak1507

02:57

What does it allow you to do?

Razetime

nothing at all

I just like having more than 1 gold badge

only tag badges and rep give you privileges

rak1507

Why does having it mean you can upvote the answer?

(sorry, I'm probably being a bit thick here)

Razetime

electorate requires you to vote on 600 posts

and maintain a 1:4 answer to question vote ratio

I upvote too many answers

rak1507

Ahh

I don't really pay attention to badges or rep that much really

I cannot believe my answer to the rod cutting problem got 8 upvotes that's just obscene, people really don't take into account how hard the answer was to make when they vote, it's a shame

Razetime

when questions get more attention, more non code golf voters come in, who vote for really short answers

rak1507

03:03

True

Razetime

so don't beat yourself up about it

rak1507

I'm not beating myself up, it's just annoying to me that people view it and the JS answer equally, when the JS one is clearly superior in every way, other than code size

Razetime

not very clear to new people

rak1507

Maybe

Bubbler

03:22

in The Nineteenth Byte, Oct 21 '20 at 0:09, by Bubbler

How to get 10 upvotes on an answer: See a new code golf challenge popping up, solve it quickly in APL, and post as first answer with some fancy MathJax.

rak1507

lol

Kamila Szewczyk

10:15

@Adám i'm having a weird problem

 inter←{
     op←⍺⍺
     single←{
         s←⎕AI[2]
         _←(op)⍵
         s-⍨⎕AI[2]
     }
     times←{
         ⍺←1
         time←single¨⍺↑⍵
         20<+/time:⍺
         (⍺+1)∇ ⍵
     }
     times ⍵
 }

when i run it as 5 (!inter) 1 2 3 4 5, it loops infinitely

when i step it in a debugger, recursion somewhat breaks, and it starts recursing in the operator, and i get a value error, sometimes it returns correctly

dzaima

@KamilaSzewczyk that sort of makes sense - you break out of the infinite tail recursion only when the execution of all single calls takes longer than 20ms, and the debugger probably makes things slower

(oh, not the value error part)

xpqz

12:58

If we think of tacit as a function composition DSL with its own grammar, this grammar is context-free and entirely parseable, is that true?

Or maybe even regular; my school days are too far behind me.

Adám

@xpqz a b c d could be (fn op) fn fn or fn(fn fn fn)

xpqz

So that ambiguity stems only from the fact that some operators can also be functions.

This dfns.dyalog.com/n_tacit.htm suggests at least to me that it's parseable? What am I missing?

Marshall

@xpqz A restriction of APL expression grammar is context free if and only if the type of every variable is known statically. So if for example you only allow variables to be functions then it would become context free.

@xpqz You can't distinguish cases like {X f Y} and {X f A} without knowing types.

dzaima

@xpqz I give you the expression f←a b c d. Parse it. (hint: you can't)

Adám

@xpqz If you only have primitives, then APL is parseable in general. No, I was talking about user-created names.

xpqz

13:10

Quite. Hypothetically, I could write a 'compiler' that takes a single-line dfn and makes it into a train if the type of each variable is known.

dzaima

@xpqz as I have done

xpqz

@dzaima any chance of a copy-pasteable version?

dzaima

@xpqz not without me typing it :p (this was pre-paste)

xpqz

I might try it at some stage in python for the learns.

Marshall

@xpqz On further reflection, this doesn't really say anything about parsing tacit code. It's translating dfns to tacit functions, but not all tacit functions can be produced this way. Looking over the seven rules, you'll note that each of them ends in either a tack ⊣⊢ or another dfn to be expanded. The result is that this procedure can only produce trains that end in another train or a tack!

rak1507

13:16

@xpqz If you did it in APL you could potentially utilise 60⌶

dzaima

@xpqz here - turns out my project for it still had a commented line for that

@dzaima (the no-op monadic operator ¹ there is my way of signifying that its operand is a guaranteed-monadic function (⍤⊢ in the picture but that doesn't really make sense))

Adám

@xpqz If you stick to a strict naming scheme, then APL is parseable. BQN makes that naming scheme a core part of the language.

Marshall

@Adám (Document)

dzaima

dfnification fails to complete though:
https://dzaima.github.io/paste/#03Zg7bsQgEIb7nAS0sZQ9QE6G0kbJal2kSx0pkmvHKVyk4SicJDPgB8ZmDBu8ZrPyYobHzPePsEFmqmwKdfpUZaVe3tnDozp9qHOrylc01fMb3o4HGAWXrBh034N5/uHQocovqadBRZsNx98dKwooNvGMTmVtu0ULHcPdcY0txrmstXtZ68bIEESM1SDoXLbcDoT@sanlfLz5wjIiLvtrYAuBCC9bCgB6bxVh8QrDgJoBMZUolG4uWhngtFkAWQi7Isk2Vyg2UuUJ1gfYGa2@Bhxy5Ic3bLoZM3rwRMdXfnOKUfSQMKwHtaa6tND2FA5sRvO5uYgtwrBFGLYYsDvmC8ht/klLOL/QAoSm95OLIHbNqauILjq7cUetl8TSvi0RUfy6O2AZidiFNMoxIdz1NCyjBRF0r0ciliAT5oQpxfLQjw8QPI7PVbe2I7V3c@L0o5zpuy9NGvhaJjidjJVczF@bIYmYpIHOQoIdwPMG@gfKLhSVRJD1wLpKvAszIf@m7OM2sQ3xVrTJU…

xpqz

@dzaima what does the superscript 1 mean?

dzaima

13:23

5 mins ago, by dzaima

@dzaima (the no-op monadic operator ¹ there is my way of signifying that its operand is a guaranteed-monadic function (⍤⊢ in the picture but that doesn't really make sense))

yay dfnifier actually completes \o/ (i forgot i need to specify arity for the input, otherwise it's impossible)
https://dzaima.github.io/paste/#03ZhNboMwEIX3PYmtFCk9QE5mdVu1UVh013WlSqwpXbDoxkfxSTpjG7AdbIYUEpMqBcY/8743cgyBMVU2hTp@qbJSrx9sf1DHT3VqVfmGoXp5x9PTDkbBR1YMuh8hPP1y6FDlt9TT4EKHDcc/2fIHVhRwWC07JpbtdlJDZje5rN30GKEAnAMJbDEistYystaNF0jJ9upiCbVJOZTB5I6k0UA97p0mAFiCgP0XwbsigECYQIHe7cMkPnQg0wFIdsQcKDsXo6zA2szQHJhM4GSbPx4b@HJH7ESygayviYk82YL2T0eboBWd1g8fpxUWFwakkJ08PbczNYSHtmc6vxnNx0J7nTYiaEYEzYjojVgXF3g5t9C3RByJwdKoHaH9CG0mbkSQrGhsfYlOhI2bcNS8Y/fVuD9bUUfRtSf6xScIq0/MXX@DSSMRLsNgrY0vwGQv1Tse…

xpqz

My apologoies; my eyes missed the fact to typed that two messages ago... :/

rak1507

That's really cool

Adám

@dzaima What language is it written in?

dzaima

@Adám java ofc

the transformation rules:
https://dzaima.github.io/paste/#0hVU9i9swGN7zK14yxRAITrodHQzl4JYSel3KcRRfYvXcO9tBdu5ITZZObWn2jqV0uNnQwR3zU/xL@urTkqxLQ0ikV88rPXreD11WNM0/XF1fXcPbN9HF64vzd/AS8uQRLvVKPQLA72wGN3GZrqCicV6SgmZxlRY50O19UuJ6PZ6QELrDnwDIHCZkwcfjKXA7msgiQMt4Px2A2QfHYIC7b7/AhLNl/nH3ZsDh7ibc3F2YusOTsCl4FGZLz96Rh3cPtVhHYfflBzeaYPBzhgFjLXHJ1F3j35aQM3hI6A5uaHGX5LZoEyZRL8Fndng4lYO5q4RGg4R7kXpzZgN3Y5erVEJGC9eUYMwQHFuwdz62QRQyDK7jQIu5WTL/aJ4t2WI0l9FRECHKqshRiQrix3gHhBYZrIkQpBac92LnY@uoWWyqNEs/JRzLmSFEJ1mg8MhQUORhxUjxKP4mL9zsZSblwW6qY4u@//XgVkGUa6@51oI6X2HG/dlo1Jcfluarc29hitKs2UWzmN5BwtKlusVViEtI0u…

Adám

CMC: Implement ^ in APL.

xpqz

13:28

Tres cool.

Imagine having that hanging on a ]command ...

dzaima

@Adám no thanks

rak1507

lol

M4X_

13:54

Hi guys, I want to create an APL interpreter. Is there a resource that clearly describes how to do this properly?

xpqz

@dzaima in your transformation rules, what do mP and pP signify?

dzaima

@M4X_ What's the best approach hugely varies between different goals. Performance? statically or dynamically parsed?

@xpqz let me check. i have no clue lol

A_mP disallows the thing to contain ⍺ or ⍵ at any depth. A_pP requires it contain ⍺ or ⍵ at any depth

so {"A1pP f2 A2mP", "A2 f2⍨ A1"} means that you reverse args only if as a result you'd bring arguments to the right side

M4X_

@dzaima I am looking for a learning experience really. I was thinking of doing it in Nim.

So I will probably be more interested in the easier solution

dzaima

@M4X_ this is a thing I like to link about how APL expressions are interpreted. There's also this, but it isn't completed yet.

@dzaima (first link is really about J, but the same thing applies to APL)

M4X_

Thank you very much

dzaima

14:09

@dzaima (and how that looks in dzaima/APL)

the important thing to note is that APL traditionally can't be statically parsed. So a big decision you must make is whether you parse at runtime (which results in overhead), or at compile-time (which restricts things quite a bit)

dzaima

14:36

@dzaima (that could be notated more sanely, but the condition system i have sucks, and i couldn't be bothered to make it better. If ignoring ⍺, that's equivalent to {"A1 f2 A2", "A2 f2⍨ A1", "A1∋⍵", "A2∌⍵"}, but adding the "or ⍺" part is impossible)

xpqz

@dzaima Thanks!

dzaima

@dzaima (not that that really matters, the current rules do almost nothing about dyadic functions and work only on a tiny subset of cases)

Marshall

@M4X_ This question comes up a fair amount, so I made an APL Wiki page with some links. There's not much out there. If anyone here knows more resources, particularly for interpreters, please add them!

xpqz

14:55

@Marshall the last two links in mlochbaum.github.io/BQN/implementation/vm.html 404s for me.

Marshall

@xpqz Fixed. I also got to switch the link from bcases.bqn (now simple.bqn) to bytecode.bqn, which has tests specifically for VM testing.

The VM document is somewhat outdated overall, as there are some new VM-provided functions that are needed and LEB128 isn't used any more.

Know your compiler ABCs: April, BQN, Co-dfns, dzaima/BQN.

Adám

15:26

@Marshall Really nice. Can you add dfns.parse?

Marshall

16:05

@Adám Dunno why you're asking me, but I added it.

Kamila Szewczyk

16:40

I left my APL session running for a while

and it segfaulted for no reason

RikedyP

17:01

@KamilaSzewczyk Ooh that looks like chat.stackexchange.com/transcript/message/56956584#56956584

Kamila Szewczyk

yeah

RikedyP

noted, thank you

@KamilaSzewczyk actually can you go to Help → About and copy that and paste it to me somehow or send to rpark at the domain what is dyalog dot com please?

Kamila Szewczyk

there isn't much inside it

Interpreter:
  Version: 18.0.39712
  Platform: Linux-64
  Edition: Unicode/64
  Date: Dec 11 2020 at 01:07:55

RikedyP

Doesn't include RIDE version?

IDE: something

Kamila Szewczyk

it does, but i'm not sure how can it help

IDE:
  Version: 4.3.3463
  Platform: Linux x86_64
  Date: 2020-07-07 10:44:21 +0100
  Git commit: 0cd6e9faf39a5d6a5f5caa94d1ff51743d5cdd75

RikedyP

17:07

@KamilaSzewczyk Nor am I, but the more info the merrier

Kamila Szewczyk

i can mail you my aplcore

if it's of use to you

RikedyP

ooh yes please

Kamila Szewczyk

i should write a script to automatically compress my aplcore and mail dyalog when the interpreter crashes

RikedyP

loool

I'm afraid supports email would fill up if you did that :P

Kamila Szewczyk

did you receive the e-mail?

RikedyP

17:09

or rather, I would like @Adám to confirm whether doing that is actually a good idea, but he is sabbathing

Vince gets the support emails

support at

Kamila Szewczyk

okay, forwarded there too

RikedyP

although end of day has passed today so I can't do much

but yeah thank you for reporting / sending :)

Kamila Szewczyk

17:32

no problem

Kamila Szewczyk

20:17

if i wanted to copy over my current workspace to the child isolates, how would I do that

so that if i have a variable a, it can be used within an isolate too

the documentation mentions 'DRC'⎕CY'something', but i'd preferably not save that anywhere

> The listenproperty is used to enable callbacks

I set it, just like the documentation shows it, but i still get an error about callbacks not being enabled

also:

      {⍵+##.a}IÏ 2 3
36:VALUE ERROR: Undefined name: localaddrs
Run[69] :If ~ok←(⊂addr)∊localaddrs[;2] ⍝ remote
                        ∧

Kamila Szewczyk

20:42

hm, okay, i got it to work with usual stuff like the root namespace, but how can i pass around dfn-local variables? {a←3⋄{##.a}IÏ⍵}1 2 3 doesn't work, a←3⋄{##.a}IÏ 1 2 3 works

section 3.1 seems to explain that, but is there a way to pass around a custom namespace?

Inside an isolate, the special symbol ##(aka “Parent”) refers to the root (#) of the parent process - and not, as one might expect, the space inside which the isolate was created (as would be the case for a normal namespace).

i tried passing around a namespace, but that didn't work

      {op←3 ⋄ {self←⊃⍵⋄self.op}IÏ ⎕this,¨⍵}1 2 3
FUTURE ERROR: 6: VALUE ERROR: {self←⊃⍵ ⋄ self.op}: iEvaluate[13] (,⍕(⍕rc),': ',(
0⊃res),{(⍵∨.≠' ')/': ',⍵}1⊃res,'' '')iSpace.qsignal rc
      {op←3 ⋄ {self←⊃⍵ ⋄ self.op}IÏ ⎕THIS,¨⍵}1 2 3
      ∧
      {op←3 ⋄ {self←⊃⍵⋄self.op}¨ ⎕this,¨⍵}1 2 3
3 3 3

Transcript for