The APL Orchard

Dyalog APL

04:03

<klg> ⋄1 '̇−̣̇' 2

(no output)

<klg> ⋄1 '−̣̇' 2

┌→──────────┐
│   ┌→──┐   │
│ 1 │−̣̇│ 2 │
│   └───┘   │
└∊──────────┘

Adám

klg: Welcome to the party.

Dyalog APL

<klg> sorry for the noise, it wasn't fully intelligible what the problem is from this side ;)

Adám

It is all but clear from this side either.

But try putting a 248 and a 775 in a string for the bot.

Dyalog APL

<klg> ⋄⍴'ø̇'

┌→┐
│2│
└~┘

<klg> ⋄'ø̇'

┌→─┐
│ø̇│
└──┘

Adám

OK, so it is definitely when scraping the SE side that things go funky.

⋄'ø̇'

Dyalog APL

04:11

@Adám
┌→──────┐
│&#248;̇│
└───────┘

Moonchild

@Adám ah, devious! The logs from my end show the bot gets <code>⎕UCS⎕UCS'ø̇'</code>; note the squiggle atop the ;. Raku does normalization before the html entity decoder is able to get at the text, so it's unable to decode that. I can probably fix that by decoding the entities in the python wrapper

Adám

Python wrapper? How many languages are involved here?

Moonchild

there was a pre-made SE api for python, and it seemed easier to just use that than to reimplement the API in raku

Adám

HTML→Python→Raku→APL→Raku→Python→HTML :-D

Moonchild

⋄'ø̇'

Dyalog APL

04:15

@Moonchild
┌→─┐
│ø̇│
└──┘

Adám

Aha, you fixed it?

Moonchild

@Adám haha, pretty much

@Adám this is because it's trying to get term with, can probably be fixed with some stty magic

Adám

I wouldn't be surprised if that Python code used some JavaScript…

Moonchild

oh, no, not term width, it's just hard capping at 80 chars. Is there a way to make it not do that?

Adám

@Moonchild I suspect that's APL's fault. Try injecting ⎕pw←32767

Moonchild

04:19

⋄ '¡¢£¤¥¦§¨©ª«¬®¯°±²³´µ¶·¸¹º»¼½¾¿ÀÁÂÃÄÅÆÇÈÉÊËÌÍÎÏÐÑÒÓÔÕÖ×ØÙÚÛÜÝÞßàáâãäåæçèéêëìíîïðñòóôõö÷øùúûüýþÿ'

Dyalog APL

@Moonchild
┌→──────────────────────────────────────────────────────────────────────────────────────────────┐
│¡¢£¤¥¦§¨©ª«¬®¯°±²³´µ¶·¸¹º»¼½¾¿ÀÁÂÃÄÅÆÇÈÉÊËÌÍÎÏÐÑÒÓÔÕÖ×ØÙÚÛÜÝÞßàáâãäåæçèéêëìíîïðñòóôõö÷øùúûüýþÿ│
└───────────────────────────────────────────────────────────────────────────────────────────────┘

Moonchild

whee!

Dyalog APL

<moon-child> are there any other oddities I missed? Tables not lining up I think apl can be blamed for, as it doesn't seem to recognize combining characters

Adám

It isn't that it doesn't recognise them, but rather that it stands no chance. How should it present

↓ab↓
↑a´↑

as a matrix if that accent was combining?

Dyalog APL

<moon-child> it could deal in units of extended grapheme cluster

<moon-child> as raku .comb, .chars, .substr, etc. do

<klg> I'm not convinced it's a good idea but it's imaginable that boxed output could turn combining characters into spacing ones (exactly like you just did by hand using a´), e.g. GNU APL uses ␀ for ⎕UCS 0 when boxing

Adám

04:32

If it did, then what should ≢'a','´b' give if that accent was combining?

klg: Do all combining characters have a spacing equivalent?

And what about zero-width characters and double-width characters?

Dyalog APL

<klg> not all have a single precombined character equivalent, but those that don't can always be encoded as space+combining

Adám

Ah, but that wouldn't print right on platforms that don't combine. Then it'd have to know about the frontend, which it really cannot.

The Windows IDE, e.g. outputs in a grid, rather than as running text.

Dyalog APL

<klg> well, you have to assume some level of Unicode support

<klg> but yeah, zero-width and fullwidth ones are problematic, and there's no generally good solution for all

Adám

I guess it could try to patch things up if connected to a RIDE client.

Razetime

@ngn how do I get the result of ⎕← in ngn/apl without an alert box showing up?

Dyalog APL

04:38

<moon-child> now I'm curious how (if at all) those are handled

<moon-child> ⋄,'字'

┌→┐
│字│
└─┘

Adám

From the Windows IDE:

Dyalog APL

<klg> that's not that bad, actually

<moon-child> (i.imgur.com/QwWs0r8.png)

Adám

It is hard to read multiple double-width characters in a row, because they are forced to fit into the narrow grid:

However, the IDE let's you choose font, and if you select one that had wide chars:

Dyalog APL

<klg> but can you mix them? ⋄'全角とﾊﾝｶｸ'

<moon-child> klg: inline evaluated code isn't supported from irc side, since it wouldn't be clear where it ended

Adám

Can you not allow backticks?

Another issue is directionality. RtL characters can make for very confusing output.

Bubbler

04:54

(for anyone participating in AoC: day18 starts in a few minutes)

Adám

Like if you have a matrix with RtL characters in the top row and LtR in the bottom row. Then the first character of the first row would be above the last char of the bottom row!

Not to mention that if you write 'string1'⊣'string2' but the strings consist of RtL characters, then the entire expression is reversed, including that the ⊣ will be rendered as ⊢

Dyalog APL

<moon-child> if the ⊣ is rendered as ⊢, then it'll still point at the right argument :)

<klg> not if you change the direction in the middle of a string :>

Adám

"Proper" Unicode rendering in RIDE:

Dyalog APL

05:10

<moon-child> (i.imgur.com/N6fVgw7.png)

Adám

The exact same in the Windows IDE:

https://i.imgur.com/16w2vMw.png

https://i.imgur.com/2c7L1a4.png

Dyalog APL

<moon-child> wow...does it still happen if the numbers are characters?

Adám

imgur.com/a/wGKhQnp

Sometimes, "proper" rendering is bad.

.daer ot drah si EDI swodniW eht ni txet werbeH gninnur a taht si edisnwod eht ,esruoc fO

Dyalog APL

<klg> ⋄'سلام'

┌→───┐
│سلام│
└────┘

Dyalog APL

05:47

<moon-child> at the end of the presentation earlier, somebody in the chat mentioned a ‘lean project’. Anyone know what they were talking about? Possibly leanprover.github.io or leanin.org/circles-network/lean-in-apl?

Razetime

06:09

⋄⍳5

Dyalog APL

@Razetime
┌→────────┐
│1 2 3 4 5│
└~────────┘

Razetime

I came up with a userscript which executes ngn/apl code

anyone wanna test it out?

Adám

Sure,

@Razetime Where do I find it, and how is it used?

Razetime

gist.github.com/razetime/053c80f62b8b820b69721b18cdfc27eb

should work on greasemonkey

same as the bot

prefix the code with ⋄

Dyalog APL

@Razetime (no output)

Razetime

06:15

and the output should appear in red below it

I should disable usage of ⎕ cause it makes alert popups

Adám

@Razetime I got a popup when I entered the sandbox room.

Razetime

great

that means it's working

Adám

Gave me that one popup, and answered me once, but now it is silent.

Razetime

hmm

Adám

It seems that reloading the page makes it work once.

Razetime

06:37

⋄'apl'

Dyalog APL

@Razetime
┌→──┐
│apl│
└───┘

Adám

06:50

@user4069 Welcome. Interested in APL?

Razetime

08:21

Day 18 answers

Warning: Regex

ngn

09:37

regex too

@Razetime only 3 ≠⊆⊢s today :)

Razetime

10:00

btw do you have experience with userscripts?

@ngn needs more I suppose lol

ngn

@Razetime me? very little

Razetime

I was trying to use n9n/apl in SE chat: gist.github.com/razetime/053c80f62b8b820b69721b18cdfc27eb

ngn

10:16

@Razetime cool :) i've just installed greasemonkey and created a new user script. test: ⋄ 1+2 3

Dyalog APL

@ngn
┌→──┐
│3 4│
└~──┘

Adám

Maybe you should temporarily use a different prefix character?

RikedyP

@DyalogAPL Ah, nice to see the good old APL logo reappearing after all these years

Elias Mårtenson

A question: How does Dyalog handle bitwise logical operations?

ngn

@EliasMårtenson only through bool arrays

Elias Mårtenson

10:33

@ngn I see. So you have to use ⊥ and ⊤ to split it into an array and recombine it again?

dzaima

@EliasMårtenson or 11⎕DR (or equivalent)

Elias Mårtenson

GNU APL has bitwise operations like so: gnu.org/software/apl/apl.html#Section-3_002e16

But I don't like its syntax, it requires very particular parsing rules.

dzaima

@EliasMårtenson and doesn't allow for monadic ∨∧⍱⍲

Razetime

@ngn is it working?

ngn

@Razetime i don't see any red text on the page, but there are no errors in the js console either

⋄ test,1 2

Dyalog APL

10:41

@ngn VALUE ERROR

Razetime

hmm is it not loading

it worked in the sandbox

yep it isn't loading

Adám

@EliasMårtenson It'll help if you define monadic ⊥ and ⊤ as 2∘⊥ and 2∘⊥⍣¯1

dzaima

@Adám given that dyalog stores signed ints, those will rarely be optimal as they only work with unsigned ones

Elias Mårtenson

@Adám I was thinking of having a bitwise operator, let's call it B. So bitwise AND would be ∧B or ×B.

Xor would be ≠B or +B

Adám

@EliasMårtenson That's a much better solution than GNU APL's strange syntax.

ngn

10:56

@EliasMårtenson so, B←under (11∘⎕dr)

Adám

@ngn What if the arguments have different internal representation?

dzaima

@ngn that's assuming 11⎕DR works the same on both args

@dzaima (which is an incredibly stupid thing that ⎕DR allows that to not be the case)

ngn

right, i didn't think of that

Elias Mårtenson

I have not heard of ⎕DR. I'm reading the docs now to understand it.

Adám

@EliasMårtenson Did you look at aplcart.info?q=bitwise ?

dzaima

11:03

@Adám should i bother testing how many orders of magnitude those are slower than proper binary operations?

(ignoring that you'd never ever use those inline, at which point ⎕NAing in a library is much better)

Adám

@dzaima Sure, go ahead.

dzaima

@dzaima i thought a 2-arg ⍺ was supposed to solve that, but it doesn't seem to work for converting to bits‽

Adám

@dzaima How would two separate calls to ⎕DR know about each other?

dzaima

@Adám they don't need to know about each other, but there should be a way to tell it what type the args are supposed to be used as

@dzaima (at which point it's not really "data representation" but it's at least usable)

like, if my 16-bit integer array happens to be all <128, ⎕DR will just quietly create a half as long array which is just stupid

@Adám {⍉2⊥(-≠/0>b)⍪≠/2⊥⍣¯1⊢b←⍉⍺,[0.5]⍵} on signed 8-bit args gives an array of 32-bit ints :|

Adám

@dzaima You need ⎕DR twice. ⊃0 type ⎕DR data will upgrade data to type, and then you can get the bits with 11⎕DR

@dzaima Uh oh. Can you fix?

dzaima

11:15

@Adám oh wow, that makes it keep the type through monadic ⎕DR calls o.O

@Adám seems 2⊥ on non-boolean args always makes 32-bit ints

Adám

@dzaima Ah, it may not squeeze immediately.

dzaima

11:30

only 50x:
https://dzaima.github.io/paste/#0xVRNS8QwEL33VwwLgrp2bbtdrRe96MGT4FU8JLvpWty2S5Kqi3oTQaUigkdBFC9e9qIIHvWf5Jc4ia24K36AikMhTTKTlzfzJgDa1PHZ4pLaP3EsMwUicOJ60zbJ8GfOdYyp/NZrTIE62tMB88voZgX1IoSWIfTTEPomZJVniYxiBnYTKkRUqajgmG2r/MBTh9fq4GLSjPnVY9/FRZU/TKw4tcaqyu92aaadRek8amt3Z5aOqfxmKPLwkuKV3h2AaCo/RxIii5mAKkQCSLfL060oJpJ1ergFIRFSjyQBssE4aTOgUUJ4D9JuQeNPzdLFwMx86aj2T8GruQuPfR92zIozonP@baBx@H7q3yFP13xEriNyte7PBAb5Xz7D40eqMHz8WlDy8TzX/W0@w7kv2y4cbB3fGWg2I1cRk06HcZAMhSlTiMk6Q@0KSXoQJSDXGAQ2jSRwkrTZm/787OyvuzKs0rBi/Z6cX@TqvMrV@WfJaH7mUu7rpezGD@v@UQa0FQJ7ui/xdG27jIcpj0nSZFg@GaVAmdxkLClKSlqtCF…

Adám

That's not so bad.

dzaima

@dzaima oh right, i can test the performance of xor. so 40x

Elias Mårtenson

@dzaima 40x slower than what?

dzaima

@Adám i'm sure you wouldn't say that if + was 40x slower than what it could be

@EliasMårtenson APLCarts binary xor of 8-bit ints than an actual binary xor

Adám

How often do you actually need bitwise operations on integers in APL code? And how often is that operation a significant part (computation-wise) of the overall work being done?

Elias Mårtenson

11:37

@Adám It's a good point, and I'd argue that bitwise operations are a fundamental feature of the CPU on which the code runs, so it shouldn't at least be terribly slow.

dzaima

@Adám depends on what's being made. Many projects won't need them, but those that do, will definitely care about their performance

Adám

In all my APL years, I can't recall ever needing bitwise anything of anything but already Boolean arrays.

dzaima

@Adám i've definitely used bitwise stuff many times in APL (most recently, working through bitwise multiplication)

(20x slower for 16-bit, 40x for 32-bit:)
https://dzaima.github.io/paste/#07VTLSgMxFN3nK0JB6MOJk2SSTDe60R9wK11MxEUXunA6K3UnhVYiIrgURHHjphul4LL9k3yJd9IpTu1DxVGLGAYyCTfnnnNvcqy5xlR6utlC2I1I2/YFpfb8anMbRzEsOFNSeVECvxvUd8OaRykEl9ieneJxKKKSZyA6D6JfQfQ7IDoP0jhMDlrN/T3s7eJSlBxZ02G2e287N2tuNneDHoVNa55Xd3wiGtY8neikBMHxOLjspeH@uq5Y8/DmZPc2ZTkFEJey/N84IMPHBU2ctO1LTInYGvQYPoa1v5KW7lc@jKr4a5V2akKnhoMaLxBFqJld7@kbFdeg1zBrYKt1AV1Pu5qCOlmSKJAVLEWTRgodrYDQjJbH64XSmlPzrL3DfpYZMZ8oTpWvOEPIgv1wttB@ynCDaqzKaWXCg8A9YJNNOhBn8x0oh6MX4ug8zr8JzTGhkLA/ZELUqXEmVB89i09U8aetJSAhkBXLZi3cPXDhrEUUTWt21bO2DftZbsRDovy6VCpgLw#APL

Adám

@EliasMårtenson FYI.

dzaima

@dzaima (but i'm biased, i've been working a lot around binary-level stuff)

@Adám so that means that the only general (i.e. doesn't care about the argument's representation because it could be anything) useful ⎕DR construct is A⎕DR 0 B⎕DR X? ._.

Elias Mårtenson

11:54

@Adám Thanks. Those are indeed the operations I'm implementing. It's trivial so why not do it. However, I'm not entirely settled on the syntax yet. The bitwise operator is nice, I think, but while bitwise rotation is naturally nshifts⌽B, bitwise shifting isn't as natural.

Then I have to pick a name for B :-)

Adám

@EliasMårtenson That's probably why J uses a numeric code. You could do a hybrid approach, and allow both functions and arrays, so n B would be a shift and n⌽B would be a rotation.

@EliasMårtenson ₿ or ฿ ;-) More seriously, maybe &? How about ₂?

Dyalog APL

<klg> A+ had ∵ for a bitwise operator

dzaima

@dzaima (though it is called "data representation", not "data converting" i guess)

Adám

12:11

@dzaima Maybe there's a way to do a single call, that I don't know about. I can ask. I never use that though. I only use ⎕DR to analyse what my argument already is.

Elias Mårtenson

12:24

@Adám Then I'd rather use ⍉ or ⊖ for shift.

klg, really?

So the idea of a bitwise operator isn't new then.

I fell that validates my idea.

@Adám ₂ is really clever.

Adám

@EliasMårtenson Hold on, then your operator won't be a true operator, but rather a mode-switch à la NARS2000's multi-set operator or J's :s and .`s

Elias Mårtenson

@Adám It will be implemented as a true operator (i.e. from a programmer's point of view, it's parsed as an operator, its evaluation rules are the same, and in the implementation it's implemented an APLOperator instance. However, it will not call the operated-on function directly, but rather derive it using a special "bitwiseFunctionOf" derivation function.

So yes, that would allow a programmer to implement their own bitwise-derivation function for their own functions to allow the bitwise operator to work on them.

I'd have to decide on a syntax to allow that though.

Adám

@EliasMårtenson Sure, but there's no consistent system where ⌽ is bitwise rotation and ⊖ or ⍉ is bitwise shift.

Elias Mårtenson

@Adám True. Calling "bitwiseFunctionOf" on ⍉ would simply return the shift function.

dinner time

Adám

Maybe n↓ can be shift.

If you drop 1, we pad on the right. If you drop ¯1, we pad on the left.

dzaima

12:31

@EliasMårtenson how about '<<'B/'>>'B?

Adám

Please no!

dzaima

fwiw, BQN

Elias Mårtenson

@Adám That's not terrible.

Dyalog APL

12:51

<klg> Elias: well, in A+ it's not a true operator but like reduction it only works with limited set of functions, namely ∨∧~<≤=≥>≠ so no rotations

Elias Mårtenson

klg, yes, but that's pretty close to what I'm doing.

dzaima

13:24

yay per-char timing cause why not

Adám

Cool. One thing I miss from APL*PLUS is primitive-by-primitive tracing.

dzaima

@Adám that's something i might do at some point too

Elias Mårtenson

13:49

I implemented bitwise operations based on klg syntax

github.com/lokedhs/array/commit/…

I.e. using ∵

Although ₂ is still neat

By the way, can anyone explain to me what ⎕AV is actually used for?

I'm trying to understand the argument for not extending the characters available in GNU APL, which is that ⎕AV cannot be larger than 256 elements.

Wezl

@Wezl the colons look better this way

dzaima

@EliasMårtenson originally it was for the way APL code is actually encoded, allowing for a single byte per character

Adám

@EliasMårtenson ⎕AV is the character set from when implementations used custom mappings between bytes and characters.

It is only included in Unicode Dyalog for backwards compatibility. It isn't used for anything, and it is considered deprecated.

dzaima

oh wow, GNU APL

Elias Mårtenson

@Adám Thanks, that means I don't have to pay attention to it.

I mean, GNU APL has ⎕UCS to handle mapping between chars and numbers, so I don't understand the desire to keep ⎕AV.

Adám

14:01

They also try to stay standards compliant and compatible with APL2.

Elias Mårtenson

@Adám True, they are very committed to APL2 compatibility.

It's an honourable goal, actually. It's just one that I personally have no interest in.

Adám

@EliasMårtenson No, actually, less than I thought (and than they originally claimed).

Elias Mårtenson

@Adám Really? All I have to go on is people saying that they took their APL2 code and ran it unchanged on GNU, but that's obviously very YMMV.

Adám

@EliasMårtenson Then their code never contained /¨

In APL2 and APLX (the true modern APL2 clone), replicate is an operator. In Dyalog and GNU APL it is a function.

Elias Mårtenson

@Adám Didn't they change the behaviour of that recently in order to be APL2 compatible?

In KAP it's also a function. I'm not sure why you'd want to make it an operator?

Adám

14:06

@EliasMårtenson No, they recently (when I made them aware of their mistake) stopped claiming compatibility.

Elias Mårtenson

Ah, got it.

Who was it that suggested I use ` at the end of a line to indicate continuation? I implemented that.

Adám

Their documentation used to say that IBM's documentation didn't clarify the syntactic rules for replicate. But that was wrong. The APL2 reference was 100% clear and unambiguous ― about APL2 not complying with the standards.

Elias Mårtenson

@Adám I think it's the ISO spec that is unclear.

Adám

@EliasMårtenson I did, based on it being the only line continuation character used "out there" (PowerShell) that wasn't already taken by APL.

Elias Mårtenson

@Adám OK, then it's you that I have to thank :-)

Adám

14:09

@EliasMårtenson No, the spec is clear about it having to be a function. It isn't clear about when and how to disambiguate the role.

Elias Mårtenson

I have to admit I considered ⤶

@Adám Yeah, and I think his original wording in the manual was referring to the ISO spec.

Adám

@EliasMårtenson That's possible, but when I saw their info doc, the claim was about APL2, which was reasonable because they claimed compatibility with APL2, not the standard.

Elias Mårtenson

Yes. But I recall reading that paragraph all the way when I first used GNU APL (which wasn't that long after its initial release). It's been there for a long time.

dzaima

@EliasMårtenson reference

Elias Mårtenson

@dzaima Thanks. :-)

ngn

14:45

@Adám $RIDE_JS - afaict it's untouched except for formatting. eclipse plugin - iirc the developer's name is oleksandr but i can't find the project on gh.

Konrad 'Unrooted' Klawikowski

hey @Adám, I'm that beginner from yesterday, can we continue my APL lesson this evening? I finish my lessons at 8pm (GMT +1 timezone)

is there maybe an APL cheat sheet or something? it can be useful

Razetime

15:01

@Konrad'Unrooted'Klawikowski abrudz.github.io/lb

the keyboard gives hints if you hover over it

Konrad 'Unrooted' Klawikowski

Thanks!

ngn

@Razetime replacing alert with your own function might work. the problem with ⎕ and ⍞ is that any function implementing i/o must be synchronous, so that the interpreter can continue it work immediately after the function returns. otherwise it becomes a callback hell, a.k.a. "continuation passing style".

Razetime

oh make sense

ngn

btw, the version on npm is really old and i don't have the cryptographic key to remove it

Razetime

oh

whats a better place to get it from

ngn

15:14

@Razetime gitlab for now. i'm tempted to move all of my projects to sr.ht despite the risk of commercialization.

Razetime

for unobvious reasons, that doesn't seem to load propely

Wezl

@ngn I've been looking at forges, what advantages does sr.ht have?

ngn

@Razetime you could do git clone https://gitlab.com/n9n/apl from the cmd line without visiting the website

Wezl

I keep having it recommended by friends that haven't tried it

ngn

@Wezl simple fast web ui

Razetime

15:18

ok

ngn

@Wezl also, it's maintained by a single person, not a big corp

rak1507

github.com/rak1507/Advent-Of-Code-2020-APL/blob/main/… whew, today was a tough one

looking at ngn's solution I think I overcomplicated it...

Razetime

@ngn yourjavascript.com/122102728101/apl.js hosted it here

ngn

@Razetime this should work too: gitlab.com/n9n/apl/-/raw/master/apl.js

Razetime

so I just need to do apl("string") right?

or does it change

it just shows "[Object object]" everywhere now

ngn

15:26

@Razetime apl('string') returns a js object representing the result. if you also want to format that, apl.fmt(apl('string')) returns a list of strings to display

Razetime

oh nice

now it's not showing anything..

@ngn I'm not sure why, but this doesn't seem to load up in the userscript

@ngn The resource from “gitlab.com/n9n/apl/-/raw/master/apl.js” was blocked due to MIME type (“text/plain”) mismatch (X-Content-Type-Options: nosniff).

ngn

@Razetime oh..

Razetime

15:41

@ngn even this one isnt working for some reason

ngn

⋄1+2 3 ⍝ test

Dyalog APL

@ngn
┌→──┐
│3 4│
└~──┘

Razetime

⋄⍳7

Dyalog APL

@Razetime
┌→────────────┐
│1 2 3 4 5 6 7│
└~────────────┘

Razetime

ok so there's some weird SE chat thing where it just deletes the code block for no reason

⋄3+⍳9

Dyalog APL

15:50

@Razetime
┌→───────────────────┐
│4 5 6 7 8 9 10 11 12│
└~───────────────────┘

Elias Mårtenson

I wish this chat was using Matrix instead :-)

Razetime

@ngn so here's the current version

it evaluates whatever was previously done on page

but doesn't interpret new ones

Martin Janiczek

16:22

Hi folks, could anybody talk me through massaging ⎕JSON object into some sort of a tree? I'm studying Aaron Hsu's papers and talks, but I think before I can use those I need to get my data to the representation he works with...

I know ⎕XML gives the depth vector as one of the results of the parsing, but ⎕JSON doesn't...

I'm thinking some recursive fn that runs ⎕NL ¯2 and holds the current depth in its left argument might be a way to go, but I'm probably thinking way too functionally about this.

Oh wait I can variant Format to 'M' and get the depth vector? That would be sweet

yup that's so much more usable!

ngn

16:48

@MartinJaniczek cf. json parsing in k

(ctrl-enter to eval)

ngn

17:01

@EliasMårtenson is there free hosting?

Elias Mårtenson

17:52

@ngn Well, yes. You can always use matrix.org

Or set up your own server, it's federated.

I run my own, but I'd say most people just use matrix.org

Also, (and finally, since I don't want to be seen as marketing :-) ), matrix.org automatically bridges to all Freenode channels, with a much better bridge than stackexchange.

So you can just connect to Matrix.org and go to the channel called #freenode_#apl:matrix.org

Hmm, OK, #apl is not bridged for some reason.

In any case, there is an APL room (not bridged to IRC)

Martin Janiczek

are StackExchange rooms IRC behind the scenes?

dzaima

@MartinJaniczek i'm pretty sure SE chat is custom-made by SE

Elias Mårtenson

@MartinJaniczek No. There is a bridge, but it's pretty bad. It works by having a single IRC user relaying messages between the IRC channel and the stackexchange chat.

This is different from the bridge used by Matrix, where the Matrix server is a real IRC server. People on Matrix-bridged rooms often don't know who is on IRC and who is not, unless they look it up.

phantomics

Are the numeric designations for types in ⎕DR used in Dyalog part of the APL standard, or is there significant difference between APL implementations? The types available to me in CL are different and it's important to have more types (like unsigned ints) for some stuff I work on

Elias Mårtenson

@phantomics It's not part of the standard

phantomics

18:01

Ok cool

Elias Mårtenson

⎕DR isn't part of the standard at all.

phantomics

Also, ⎕DR will convert a vector of 8 16-bit numbers to a vector of 4 32-bit numbers if 32-bit is the specified output type, right?

dzaima

@EliasMårtenson if those are prevalent, how does it handle feature discrepancies?

Elias Mårtenson

I do wonder if SE chat is going to remain. It does say that there are only 77 users online across all rooms at the moment. That doesn't bode well for the service. I'm not suggesting everybody move away from SE chat, but I'm hoping that if SE chat is decommissioned, that people move to Matrix.

@dzaima Well, as best it can. For example: When an image is uploaded, the IRC channel receives a link to the picture. Formatting it's supported in IRC of course, so IRC users see the markdown markup. The big one is editing, and IRC users simply don't see the edits.

phantomics

No kidding, I hadn't used SE chat before this place but figured there'd be more people using it

ngn

18:05

personally, i don't really care which chat i use, as long as it's public and there are many like-minded people in it

Elias Mårtenson

No, wait. I just tested. The IRC users do see the edits. It's displayed as an updated message prefixed with a *

rak1507

I haven't used SE chat that much, it seems ok, but I don't like not being able to paste images directly in chat

dzaima

@rak1507 you can post direct links to images though, they should onebox just fine

Elias Mårtenson

@rak1507 Matrix is so much better it's not even funny. But that's just me, and like I said, I am not going to run around telling everybody to migrate just because I personally am not a fan of SE chat.

rak1507

@dzaima having to link to an image is a lot less convenient than being able to send screenshots with a few clicks imo

dzaima

18:07

@rak1507 well, there's also the upload button

phantomics

Could probably hack some interesting APL interactivity features into Matrix chat

Elias Mårtenson

@phantomics Yeah, the API is pretty neat.

dzaima

@EliasMårtenson since there are so many different options, what client would you suggest me to look at? i'm intrigued by a "much better" claim

also, what markdown/extra message features does it have? can't find anything

Elias Mårtenson

@dzaima You can use the standard Element one. element.io

rak1507

@dzaima means I have to save the image

Elias Mårtenson

18:10

This is the main login: app.element.io/#/welcome

You can test join the channel #APL:matrix.org

ngn

recaptcha >:(

dzaima

aight now to make a userscript to make it look like SE chat

Elias Mårtenson

@ngn Yeah, if you want to join matrix.org, they have that. Sadly.

No such thing on my Matrix server (which is the one I connect from). But then again, my server doesn't accept open registrtations.

dzaima

@EliasMårtenson is there a proper place to test things?

Elias Mårtenson

Yes. I just created #test:dhsdevelopments.com and invited you.

ab5tract

19:14

/me imagines all the chatting has migrated to matrix.org

dzaima

@ab5tract well, the message count in the matrix APL room did increase significantly

ab5tract

:)

I've got an account somewhere.. but it feels a bit like an unnecessary brain split at the moment. I guess if the interactivity really improves then switching makes sense but having two chats is usually an issue for smaller communities.

dzaima

@dzaima (as in, 30 → 44 messages if my message counting js is correct)

@ab5tract right, noone's really recommending moving to it now. But just in case SE axes us, there's a good alternative (plus i have an archive of this chat (limited, only rough time of messages, and only HTML (no raw markdown), but it's good enough))

ab5tract

@dzaima It's a good point. If the statistics on the RHS are to be believed, the SE chats are not exactly high volume, ie. may not last too long. And for the archive point, definitely it's a better situation considering all the cultivation work and other learning material that appears here.

Wezl

19:34

@ngn and the terms of service are unacceptable

phantomics

19:50

@Wezl What did I miss in the TOS?

Wezl

nvm I reread

Martin Janiczek

22:48

Is [...]← indexing assignment syntax supported for ⍺, like {⍺[...]←...}? It gives me a syntax error (where non-dfn didn't); unsure if I should investigate further or if that expression is just not supported...

To prevent XY problem, here's my full problem:

mincounts[nearest_node[idxs]]←values[idxs←⍸'namespace' in name] works

lift←{⍺←(⍴d)⍴⍬ ⋄ ⍺[nearest_node[idxs]]←values[idxs←⍸⍵ in name]} and

mincounts lift 'min_count' doesn't

ngn

@MartinJaniczek most likely deliberately not supported. assignment to ⍺ is special. it has effect only when no left arg was passed.

Transcript for