The APL Orchard

nathan rogers

01:45

Any K users have comments on project architecture and how to organize projects , please take a look chat.stackexchange.com/transcript/message/57314394#57314394

Adám

07:13

@user958099 Hi LdBeth. If you want to participate here, simply send an email to [email protected]

rak1507

08:59

codegolf.stackexchange.com/questions/220516/… I got a comment on my answer, am I right in saying it is in that path if dyalog is installed?

Adám

09:11

@rak1507 Yes, but not only that. On a normal system, ⎕SE.SALT.Load'HttpUtils' works too.

rak1507

yeah

xpqz

10:14

@Adám I offer the following tweak to your solution:

{⎕JSON⍠'Dialect' 'JSON5'⊢2⌽'}}{"":{',∊'^;.*' '^\[(.*)]' '^(.*?)\s*=\s*(.*)'⎕R'' '},"\1":{' '"\1":"\2",'⊢⍵}

(last regex to allow space around the first =)

Adám

@xpqz Not sure if that's valid INI though.

xpqz

Call it "Python ConfigParser compatiblity" :)

This claims it's valid, but I've not taken it to the actual spec: docs.python.org/3/library/configparser.html

rak1507

wow, that module is 1363 lines long

does a lot more than just reading it I guess

xpqz

Looking at what actually is allowed, the story is way more complex than it seems.

rak1507

yea lots of stuff

xpqz

10:23

I didn't realise you can also use colon as the key-val separator.

@Adám's trick of converting to json first is handy.

Adám

11:10

@user958099 Welcome. So you're interested in APL? Can we be of assistance?

LdBeth

12:16

Seems I'm on board now, thank you @Adám

Razetime

@user958099 welcome

rak1507

welcome

xpqz

hi there

LdBeth

12:38

Some day I wrote an Atkinson image dithering

` Atkinson←{
x y←⍴⍵
s←x×y
t←⍺
a←1 2,x+¯1 0 1,x
p←{(((8÷⍨(⊢-(255×t<⊢))⍺⌷⍵)∘+)@((⊢(/⍨)(s∘>))⍺+a))⍵}
t>x y⍴1{⍺>s:⍵ ⋄ (1+⍺)∇ ⍺ p ⍵},⍵
}`

user41805

(btw, code indentation is done by prepending 4 spaces to every line, or by hitting ctrl+k)

*code formatting, not indentation

LdBeth

The idea is treat the image as an 1D array and compute the offsets to do the update. It takes long on slightly larger images. I find the idea is close to Levenshtein distance, but "fold"/reduce get WS FULL on large images.

rak1507

not that it matters but s can be ×/⍴⍵

what is your max ws set to?

user41805

(also related, the dithering done in k, beyondloom.com/blog/dither.html)

Elias Mårtenson

Argh, the story is too complicated to explain, but suffice it to say that I have spent hours searching for a single bug in the double-arrow function definition syntax and it all ended up with a stupid error where I had hardcoded the wrong namespace.

In any case, my current design is such that functions declared with ∇ are global, and if they use ⇐ they are local.

LdBeth

13:17

ah, guess I need to learn how to make a newline without sending out the message

Adám

@LdBeth apl.wiki/APL_Orchard#Basic_formatting

LdBeth

Figured it out,
thank you

Marshall

@LdBeth Shouldn't a be defined in terms of y, not x? When a matrix is ravelled, the rows stay together, and y is the length of a row, so it will be the distance between a position and the same one in the next row, that is, between an element and the one above or below it.

LdBeth

@rak1507 I think my ws size is default value.
@Marshall: x is the width image, to make it easier(?) the image is raveled to 1D

omg, I think u are right, shape gives the height as the first element

Marshall

@LdBeth Yes, you could consider x to be the width but you'd have to transpose before ravelling.

LdBeth

13:29

This is what it gives from the version I posted, that I didn't even realized there's a bug in the program

Marshall

@LdBeth You're saying here that you switched to recursion because reduction was using too much memory, right? The ∇ here is a tail call, so I see why it doesn't cause a WS FULL. But ⊃p/(⌽⍳s),⊂,⍵ (untested) should do the same computation without saving intermediate results. What did your reduction look like?

There's also a big performance problem with p: it can't change ⍵ in place because it doesn't know that it will never be reused (Dyalog doesn't do lifetime analysis on variables). I don't know of a way to fix this while keeping a pure functional style, and there might not be one.

What I would do for better performance is make one mutable variable i←,⍵, and modify it with a loop like {i[(⊢(/⍨)(s∘>))⍵+a]+←8÷⍨(⊢-(255×t<⊢))⍵⌷i}¨⍳s.

Personally I find this clearer because the state gets a dedicated variable rather than being passed as an argument. It's not good that there's no indication i will be modified, but you can add a comment.

LdBeth

13:52

@Marshall I tried to convert a simpler conventional one propagation array dithering to a fold, so it is similar to ⊃p/(⌽⍳s),⊂,⍵

```
foo←150 dither img
WS FULL
dither[2] ({t<(13⌷⍵)+1 1 2 1 1 2 2 1 1 2 1 1×8÷⍨(⊢-(255×t<⊢))⍵[3 7 8 9 11 12 14
15 17 18 19 23]
∧
```
This is where I got ws full, from the session log

Marshall

@LdBeth You can't do multiline code in the same message as normal text; put the code in its own message and use ctrl-k.

@LdBeth Nothing in that statement looks like it should use much memory, so I'm not sure what the issue could be. You could try just changing the last line of your current implementation to see if it has the same problem, but as long as it has to copy the image at each step it'll be just as slow.

LdBeth

ah, got it. Since that was a failed attempt, I don't have the definition left in the workspace.

nathan rogers

14:43

@Marshall does J have function arrays?

Adám

@nathanrogers Kind of.

nathan rogers

@Adám thanks!

Marshall

@nathanrogers The J interpreter also has true function arrays, but that's generally considered a bug and I wouldn't recommend using it.

xpqz

14:59

If you store APL code on github, add a file .gitattributes to tell github about Dyalog's many different file extensions: github.com/xpqz/aplutil/blob/main/.gitattributes

LdBeth

15:14

I just rediscovered that I'm loading the image data transposed, so that's why I use first instead second of the shape ∏∏

Adám

@xpqz Ah, yes, that should be publicised more.

xpqz

@RGS I've found that in jupyterbook, if an APL code cell contains an underscore anywhere, the syntax highlighting is switched off in the resulting html -- have you noticed this?

Kamila Szewczyk

can we somehow use windowed reduce on matrices

i basically want stencil, but without the 0 padding

Adám

@KamilaSzewczyk You can, but it'll treat each row or column separately.

Kamila Szewczyk

say I have

      shuffled
 3  4 15  7
 6  5 10  9
12 13  1 14
 8  0 11  2

      1{{⊂⍵}⌺3 3⊢⍵}shuffled
┌───────┬────────┬────────┬───────┐
│0 0 0  │0 0  0  │0  0 0  │ 0 0 0 │
│0 3 4  │3 4 15  │4 15 7  │15 7 0 │
│0 6 5  │6 5 10  │5 10 9  │10 9 0 │
├───────┼────────┼────────┼───────┤
│0  3  4│ 3  4 15│ 4 15  7│15  7 0│
│0  6  5│ 6  5 10│ 5 10  9│10  9 0│
│0 12 13│12 13  1│13  1 14│ 1 14 0│
├───────┼────────┼────────┼───────┤
│0  6  5│ 6  5 10│ 5 10  9│10  9 0│
│0 12 13│12 13  1│13  1 14│ 1 14 0│
│0  8  0│ 8  0 11│ 0 11  2│11  2 0│
├───────┼────────┼────────┼───────┤
│0 12 13│12 13  1│13  1 14│ 1 14 0│

that's what i get from stencil

and i'm not interested in the outer parts of this matrix, with zero pads.

i can probably strip them off after the stencil is done but this doesn't seem to efficient to me

Marshall

15:25

@KamilaSzewczyk It wouldn't really make sense because a reduction only ever happens along one axis. Some of the functions in towards improvements to stencil could be helpful.

Adám

@KamilaSzewczyk aplcart.info?q=partition%20matrix

nathan rogers

anyone seen where I can get this guys config?

K-means clustering in J, looking like APL

►

LdBeth

@Marshall thank you for the pointer of using a mutable state, the performance has been improved by a lot

Marshall

@nathanrogers Probably not the same but J can look like APL or English is similar.

From the 2019 APL competition winner.

nathan rogers

That's where I found the video

Kamila Szewczyk

15:36

1{(,v)/⍨{,~b∨⊖⌽b←a∨⍉a←1⍪(⍵-1)⍵⍴0}≢v←{⊂⍵}⌺3 3⊢⍵}shuffled

that's my solution, probably it's horrible :P

rak1507

what does that do?

Kamila Szewczyk

n-wise reduce, but matrices.

(third time the charm)

rak1507

do you know about the left argument to f in f⌺

Kamila Szewczyk

w-

what does it change

Adám

      {⊂⍺↓⍵}⌺3 3⊢3 5⍴⎕A
┌──┬───┬───┬───┬──┐
│AB│ABC│BCD│CDE│DE│
│FG│FGH│GHI│HIJ│IJ│
├──┼───┼───┼───┼──┤
│AB│ABC│BCD│CDE│DE│
│FG│FGH│GHI│HIJ│IJ│
│KL│KLM│LMN│MNO│NO│
├──┼───┼───┼───┼──┤
│FG│FGH│GHI│HIJ│IJ│
│KL│KLM│LMN│MNO│NO│
└──┴───┴───┴───┴──┘

rak1507

15:45

seems like an odd decision to me that stencil doesn't enclose by default

I can't think of many cases where you wouldn't want to

especially considering the enclose of a scalar is itself

Adám

@rak1507 wrong

rak1507

3≡⊂3

Adám

(⊂2 3)≢⊂(⊂2 3)

rak1507

oh, well... uh... ⊆ then!

Adám

@rak1507 All these new operators mix/don't enclose: ⍤ ⌸ ⌺

J/Roger Hui influence.

rak1507

15:49

I can't see why that for ⌺ specifically is ever more useful than doing it

Adám

Roger is proposing allowing non-scalar right operands for ⍣ and now the question is whether to enclose or not/mix or not.

rak1507

I think enclose would be sensible so if ⍣scalar returned a vector result for example ⍣(1 2 3) would return a vector of vectors

Adám

Me too, but Roger disagrees.

Difference is of course ⊂⍤f⍣array vs ↑f⍣array

dzaima

@Adám if it didn't mix, it could be extendable for any nested array with sane sematics. With mixing, the scalar and vector case have very different behavior

Kamila Szewczyk

if i had a boolean matrix with two cells set, and i wanted to swap these two values in the resulting matrix, how to do it in a smart way

Adám

16:04

@KamilaSzewczyk ⌽@{boolean_matrix}data_matrix

@dzaima Ah you mean that the result would have exactly the same outer structure as the operand, only with every simple scalar replaced by an enclosure of the result of applying the function that many times?

Kamila Szewczyk

okay, i completely missed that one

dzaima

@Adám yeah - what BQN does

RGS

@xpqz I don't think I have, but let me test it out for you.

dzaima

@dzaima mixing means extending further than just a simple array is fundamentally impossible

Adám

@dzaima That's a good point. Do you want to email roger@ about it?

dzaima

16:09

@Adám out of the blue just saying "further extension of ⍣ is impossible" feels a bit weird :p

RGS

@xpqz Aha, I see it. Looking at the tool log, I see a "\booktest\notebooks.ipynb:: WARNING: Could not lex literal_block as "APL". Highlighting skipped." message. It should be a deficiency in the plugin used for the highlighting.

Do you happen to know what they use? If it's open source we could fix it – or at least open an issue about it.

Adám

@dzaima You can mention that we discussed extension of ⍣'s right operand. We've recently been discussing it internally.

dzaima

@Adám you can just mention it whenever it makes sense. (though even otherwise i'd prefer no merging. ⌸ in dzaima/APL already doesn't merge, and ⍤ (or rather ˘ in BQN as it's more compact, the ⊢∘ tripling its length) merging is awful every now and then, though it's usually good)

Adám

ok

dzaima

(after finding out how matrix bridge bots get fancy user profiles informing about it being a single bridge, and deciding what to do with usernames (currently every SE user has -se appended to their name, which is ugly), my matrix bridge is probably good to go)

Razetime

16:18

trying to make the JSON for the vscode snippet plugin: https://tio.run/##dVZbbxtFFH73r9i6QGzSJlAuKt4GSJwLKXUS7HCR0rAd787aU69nltlZJyYEFQmlidP0BqVp1VLoRaQqEg8gglRaJPcxD/yH/QX8g/LN2E4aLi/2nG/OnDNzLt/Z06RBIleyUB1uHH36NKDKiprcDaKhJauRS39AgpimD1l@Lj0ec1cxwSHVc@mC8JjPqITk7UlHIPJceiqul81WJZeeoCrCKsylZ4ik@nQ5lx6RxK1RBaHWEbSPIJc@wSpExZJ2zQhVZbwCIcqlSzQkkiihzbq5dF7U65RrCwR7Smo9a9lO6QfUaNOjkTvU5/fnBT8dwya1hz3P8Q9PUSOU4rKCV@X4T66WWIXbhThQLAyaALaL1GWhFC4J7FHWYB51/GR9ZWwxFBweGeAZsUAl0NXrpU9ivMoqCqFs/QPwXGs8EELaBcZZPa5rZDVPWYAb2gWy2MVWfygJqaz3Q3uYe1reMvKoWOD2NGy3f8Tr7VJIuON/NlyORBArapl02Ih2HMQRTq3dPUGjyJqt…

user41805

@dzaima i don't know if you know, but in the irc side matrix usernames have a [m] appended

Razetime

it keeps displaying weird characters like before -_-

dzaima

@user41805 yeah. That's less ugly, but [se] is a bit too long for my liking, and [s] is a bit strange

Adám

@dzaima ≣̧

xpqz

@RGS Not sure -- I only discovered it recently that it seemed to correspond to the _ character. Odd.

dzaima

16:22

@Adám lol (my client doesn't do font fallback because that means a couple orders of magnitude slower window resizing so i had to resort to viewing that on SE ¯\_(ツ)_/¯)

Adám

@dzaima How about [≡] then?

dzaima

@Adám that's nice (not that my client problems should influence anything, just thought to note it)

@RGS github.com/pygments/pygments/blob/master/pygments/lexers/apl.py

xpqz

Right.

That looks clean enough. I'll see if I can get it to run locally.

RGS

@xpqz If you look at the Name.Variable, if I am reading that right, _ can't be the first character in a name, and that is wrong.

xpqz

Trying the pygments demo, it works, although it highlights the _ strangely.

But yes, you're right.

That's the fix.

    # following IBM APL2 standard
    (r'[A-Za-zΔ∆⍙][A-Za-zΔ∆⍙_¯0-9]*', Name.Variable),

RGS

16:38

@xpqz It fixed it? From your original msg I can't understand if you meant any underscore whatsoever or only isolated underscores, but mastering.dyalog.com has underscores in code cells with highlighting – in the middle of the name of variables

xpqz

Odd that high minus should be allowed in a variable name.

RGS

Btw, maybe that should be fixed to include accented letters...? Dyalog APL has support for those; anyone has any idea how the other APLs support this?

@xpqz Yeah that also doesn't look right...

Adám

@xpqz Ah, true _←4 is invalid in APL2.

xpqz

I'll see if I can hack the pygments that jupyterbook uses.

@Adám oh :(

RGS

@xpqz You probably can; for one I have messed with it before.

Adám

16:40

APLX errors too. But the APL2 clone, GNU APL does allow _←4

xpqz

Is ¯ allowed in a variable name (as long as not the first symbol) in APL2?

Adám

@RGS APL2 allows accented chars, APLX and GNU APL not.

@xpqz Yes, and in APLX and GNU APL too.

dzaima

while at it, ⍫⍩ should be a operators (⍡ᑈᐵ also are in dzaima/APL, but probably not worth adding)

Adám

@dzaima Uh, ⍫ is not an operator in any APLs but dzaima, afaik.

In many APLs, it is of the same syntactic class as ∇

dzaima

@Adám isn't it in Extended? (and aren't there at least plans for it in others?)

Adám

16:42

@dzaima OK, but that isn't a real APL, is it?

dzaima

@Adám i'd say it's approximately the same level of realness as dzaima/APL. (and ⍩ is only in dzaima/APL, but i'm suggesting it as it can't really be anything else)

Adám

Dyalog has tentative plans, not thing definite.

@dzaima True, and ⍩ has been used in APL texts as a dyadic op before.

xpqz

@RGS yes, that fixes it hooray!

RGS

@xpqz Cool! Now what should we do about it?

It is kind of annoying that pygments is broken.

xpqz

I can PR it and see if they take it.

dzaima

16:47

@RGS PR. it's been done before

RGS

@dzaima I know you can PR.

What I mean is, the lexer there seems to follow standards that don't match Dyalog's.

And I was wondering how to proceed with regards to that. Adám just said that _←4 doesn't work in APL2 (which is now dead, correct?) and the lexer says it's following APL2's standards.

dzaima

@RGS I don't think extending functionality (to support the biggest impl even!) is ever bad for a highlighter.

Adám

@RGS APL2 isn't dead, just moved out of IBM.

RGS

@dzaima True.

xpqz

It seems to follow the "standard", but arguably, Dyalog's the most widely used, and actively developed APL.

RGS

16:50

Yeah. @xpqz want to do the honours, or would you rather have me PR that?

dzaima

@RGS also ⍮ isn't a function in APL2, but it's in there so

xpqz

Don't mind. Maybe it has more gravitas if it comes from someone at Dyalog.

All I did was:

    # DYALOG APL
    (r'[A-Za-zΔ∆⍙_][A-Za-zΔ∆⍙_0-9]*', Name.Variable),

Adám

At some point they'll have to move to Dyalog as de facto standard.

Unicomp changed from APL2 to Dyalog layout for their mechanical keyboards a few years ago.

RGS

@xpqz I don't think that's relevant, at all. I don't mind doing that as part of my job, but I also don't want to prevent you from making a (small) contribution to the pygments project.

@xpqz So you removed the reference to APL2 and included the ref to Dyalog?

xpqz

Actually, maybe I should leave the high minus in the second bit so that it doesn't now break APL2 highlights.

That's probably a better compromise.

RGS

16:53

Yeah. And what about the accented letters? IMO using them is a terrible idea, but they are allowed.

Adám

@xpqz Yeah, it doesn't actually hurt. This syntax colouring isn't supposed to error-check the code. For display purposes, mainly.

@RGS I'd add them too. Both Dyalog and APL2 allows them.

RGS

Thankfully, MDAPL lists the valid letters, just slightly above mastering.dyalog.com/…

dzaima

worse problem - # is currently a comment starter there, but it's current namespace in Dyalog

xpqz

There is also this:

filenames = ['*.apl']

Adám

@dzaima Ooh, # is a comment in GNU APL.

RGS

16:56

@xpqz filenames = ['*.apl?'] ?

C. Quilley

Hello - in j (or apl, if it's not a builtin and I can try translating) is there an idiom for the coordinates of a boolean in a matrix? i.e. if I have (0 1 0;1 0 0; 0 0 0) I want to get (0 1;1 0) as a result

Adám

@RGS Shouldn't that be *.apl*?

xpqz

Not relevant to jupyterbook, but Dyalog's penchant for a separate suffix for every APL ...thing could benefit.

RGS

@Adám * matches arbitrarily many characters, but we only want to match 1 maximum, no?

Adám

@C.Quilley I. in J and ⍸ in APL.

xpqz

16:57

*.apl.

RGS

@xpqz so what makes the first . be a literal dot and the last one a placeholder?

Adám

@RGS A mistake, I think. This is globbing, not regex.

xpqz

Hmm, it's not a regex actually, it's a glob

So @Adám has it, I think: .apl

Meh

*.apl*

C. Quilley

@Adám Not quite - that finds the index on each row; I wanted to get the row information as well (as well as any higher coordinates)

Adám

@xpqz I can't think of anything it'll match.

RGS

17:00

tldp.org/LDP/GNU-Linux-Tools-Summary/html/x11655.htm

dzaima

['*.apl','*.apl?'] (and also probably ,'*.dyalog')

RGS

Yup, *.apl? is what matches .aplf, .apln, .apla, etc

Adám

@C.Quilley Oh, in J that is the case, not in APL. You can implement it as as the copy # of the flattened matrix over the flattened list of all coordinates.

xpqz

Is there a definitive list of all suffixes used by Dyalog?

Might be better to spell them out exactly

Adám

@xpqz afonci

RGS

17:03

@xpqz help.dyalog.com/18.0/#UserGuide/…

C. Quilley

@Adám Oh, thanks! I didn't realise that it was better implemented as ⍸

RGS

Oh wow, the docs page is missing a.

xpqz

What's the a?

Adám

@RGS No.

> Additionally, Link uses .apla files

@xpqz array

RGS

Yeah, you are right. If only I knew how to read.

Adám

17:06

Dunno if the syntax colourer should even colour .apla

xpqz

It's disallowing leading _ in function names, too.

RGS

@xpqz Are you looking at "name.function"?

Because that starts with a leading ⎕, which I guess is being used for system functions.

My guess is that actual functions are being highlighted the same way as other variables..?

xpqz

Yes, right

RGS

That has the interesting side-effect that quads get highlighted differently :P

Oh nvm, what I just said is wrong.

It is getting a special highlight because of the "Name.Variable.Global" rule, which looks weird as hell.

xpqz

Right, doing a full rebuild of my jupyterbook to see if there are any remaining odd artefacts

rak1507

17:20

@dzaima I love the comment there, 'I will pretend for a moment that I carefully reviewed your code change, double-checked the primitive functions against the latest APL dialect, and came to the conclusion that your changes are correct and valid. Thanks for the contribution!'

RGS

@rak1507 A little pearl from the world of open source.

rak1507

what on earth are ⌸⍯↗

oops copied too many, but the last two, what are they from?

(I have to say, I've always wanted a ↗ which does a diagonal take forming a square matrix)

nathan rogers

@Marshall will BQN support symbols?
@dzaima will BQN have true lexical scope? Closures? Partial application?
What about function arrays?

rak1507

bqn has function arrays

dzaima

@rak1507 I don't think ⍯ is used anywhere. ↗ is "throw error" in ngn/apl iirc

rak1507

17:23

ah

dzaima

@nathanrogers There's not much reason for symbols imo. It already has lexical scope & closures, and the 2-argument limit means ⊸ and ⟜ are enough for partial application

nathan rogers

I find symbolic programming to be equally as necessary as array notation, and I miss either when I'm without one or the other or both

Razetime

getting lb style autocomplete on vscode is pain

dzaima

@nathanrogers (and functions are first-class objects in BQN, just like k (of course also with closures/lexical scope))

@nathanrogers what do you mean by "symbolic programming"?

nathan rogers

Symbolic programming

In computer programming, symbolic programming is a programming paradigm in which the program can manipulate its own formulas and program components as if they were plain data.Through symbolic programming, complex processes can be developed that build other more intricate processes by combining smaller units of logic or functionality. Thus, such programs can effectively modify themselves and appear to "learn", which makes them better suited for applications such as artificial intelligence, expert systems, natural language processing, and computer games. Languages that support symbolic programming...

Mathematica/Lisp/Scheme/K

dzaima

17:27

@nathanrogers k isn't a symbolic programming language

xpqz

@RGS that seems to have fixed my issues. Only thing left is that boxes render a bit sub-optimally.

nathan rogers

i'm literally doing it right now

rak1507

@dzaima you can get the AST of something in K (ngn/k at least)

dzaima

@rak1507 you can also get the bytecode of a class in Java, so pretty much every statically-parseable language would have symbolic programming

rak1507

sure

nathan rogers

17:29

right but just like array notation

providing a literal notation for symbols is a useful tool

dzaima

@nathanrogers you mean k's `name? that's literally just a scalar string, nothing else

nathan rogers

until (.`name)

it means I can define a function once and use it in a variety of contexts

Razetime

github.com/razetime/bqn-vscode BQN snippets and base language configuration and file association complete. (Now to figure out how to expand things automagically)

RGS

@xpqz Isn't that fixed with some CSS?

@xpqz Feel free to PR and I'll go and comment and say that PR is helpful for me as well.

dzaima

@nathanrogers still wouldn't call that symbolic programming. It's just ⍎'name' with better syntax (and also ability to assign sanely i guess, but that's been proposed for Dyalog too). Worse yet, that uses dynamic scoping :P

user41805

17:33

@dzaima so the moral is, we need a more lispy k?

(is there a k that supports macros?)

nathan rogers

@dzaima it doesn't use dynamic scoping, it simply late binds a name to its value, where the name is what it is in whatever context

dzaima

@nathanrogers that was mostly a joke, but it shows that the symbol has no relation to any actual variable until .

nathan rogers

this is an inversion of dyanmic scoping

@dzaima right, just like a macro written in lisp doesn't have any relationship to any meaning until the macro is called

ngn

@dzaima +1. "homoiconic" and "has a symbol type" - yes. "symbolic programming" in the sense of mathematica etc - maybe not.

dzaima

oh wait, .`a only resolves the global a variable, i forgot. well, ignore my mentioning of dynamic scoping, it uses just global scope

nathan rogers

17:36

you could just as well say that lisp macros are nothing but ⍎'aplcode template'

ngn

everything is 01 and +× :)

xpqz

@RGS github.com/pygments/pygments/pull/1747

See if they take it

@RGS I took your CSS, but it's not 100% for me (box -s=max)

RGS

@xpqz You mean MDAPL's? The one I currently have also needs some tweaking.

xpqz

@RGS yeah

The acronym MDAPL always sounds like a designer drug to me.

RGS

@xpqz Do you spell it out in your head? In my head I say "mad apple" and I just find that very amusing.

xpqz

17:48

I spell it out

Marshall

@nathanrogers I don't plan to support symbols, and BQN isn't designed to do any sort of meta-programming. Functions describe computations and can easily be written, passed around, and applied. For example see the document on implementing control structures.

xpqz

@RGS ^^ that's what I get using the MDMAPL CSS

and playing a bit with the numbers.

RGS

I am terrible with CSS but it feels like the line-height is too short? It also feels like the font being used isn't monospaced – that would explain the misalignment in the bottom right corner of both boxes.

xpqz

Yeah, CSS is an annyoing dark art

RGS

@xpqz They just gave you homework 😂 Don't worry, I can do that for you if you want, just won't do it right now.

xpqz

17:57

I remade the mappings, but tests will have to wait. Weekend starts here.

RGS

Enjoy your weekend :)

Dyalog APL

<Wezl> Is the bot still broken?

Marshall

Wezl: I see your comment; do you see mine?

Dyalog APL

<dzaima[m]> yep. Still only works one way - never reads from the SE room

user41805

i believe the bot left on the irc side

dzaima

18:06

@user41805 it doesn't ⋄'evaluate' either, so it's definitely a problem of not reading from SE

Dyalog APL

<Wezl> I suppose I shouldn't feel entitled to it

Kamila Szewczyk

18:30

jsoftware.com/papers/amuse-bouches.htm - what is this x⍳x thing here? why does it (supposedly) help?

Dyalog APL

<loke[m]> shapr: Well, I wrote the GNU APL mode for Emacs.

Marshall

@KamilaSzewczyk It allows you to perform computations on numbers rather than arbitrary data. Roger's generally moved to advocating self-classify which does the same but has the advantage of fixing up some tolerant comparison issues; in APL it's written ∪⍳⊢.

Dyalog APL

<Wezl> you privileged matrix folks and your logs

<loke[m]> I run my own matrix server

<Wezl> respect +← 1

<Wezl> user41805: no, it's still here, but not doing anything

dzaima

19:04

ono ChatExchange crashed and broke my matrix bridge D: seems that problem will just exist until i rewrite the whole SE interface

Dyalog APL

<shapr> loke[m]: yay!

<shapr> loke[m]: does it have an input mode like swedish-postfix?

<shapr> Jag hoppas det finnas, eller

rak1507

is the bot only picking up a few things?

dzaima

@dzaima oh, in fact it broke exactly the same way as the IRC bot has - I'm writing this from my Matrix client, it appears in SE, but I don't see new messages from SE

Marshall

@rak1507 I think loke's replying to this.

Dyalog APL

<loke[m]> Yes. You can enable it with M-\ and choose APL-Z

rak1507

19:06

@Marshall ah

Dyalog APL

<shapr> jaså, tack!

<loke[m]> I made a video many years ago. youtu.be/yP4A5CKITnM

<shapr> ser ut bra!

dzaima

@dzaima restarting is always an option, but since it doesn't throw any python errors (i guess i could parse stderr, but i already use it for logging) that has to be done manually. (but a manual restart currently requires restarting the whole thing, which clears the message map which is only in memory, meaning replies to older messages won't work correctly)

Dyalog APL

<shapr> har du Patreon också?

<shapr> finns det #apl.se ? :-D

rak1507

jsoftware.com/papers/KEIQA.htm#Maple this anecdote is the perfect example of APLers superiority complexes

the idea that someone couldn't figure out how to find how many numbers in a list are over 100 is ridiculous

Kamila Szewczyk

19:48

> How do you find the number of elements of a vector that are greater than a hundred

not "can you find me how many elements in this vector are 100"

rak1507

same thing

Kamila Szewczyk

"can you find me the longest run of consecutive integers in this vector" vs how do you do that (as in: explain to the computer) are two different questions with different levels of complexity

rak1507

well yeah obviously the question would be how do you do that with code

but I don't believe that people were not able to do that like the anecdote claims

ngn

@rak1507 who knows. those "young people staffing the booth" might not have been programmers at all or in those "few minutes" they might have been busy with something else.

rak1507

Maybe so, even so the point of that anecdote is clearly 'apl is better than everything else' which is just not true

Marshall

20:05

It's also a big jump from "APL was manifestly better than a moderately obscure CAS in the year 2000" to "APL is always better", but APLers sure seem inclined to follow that logic.

rak1507

yep

ngn

so the image for how apl sees apl was right :)

2

Marshall

@ngn Been thinking there should be a BQN extension available for this table. BQN obviously sees itself as the alien from Alien; from APL's perspective it's probably more like E.T.. For K, Jabba the Hutt?

ngn

20:21

@Marshall jabba - haha, maybe, but it needs to be more rectangular and have boxes, and ideally strange symbols in the same picture

.. and trimmed from the right, the way bqn chars appear in a terminal by default :)

Marshall

@ngn Oh, like the chinese in j→apl. I hadn't noticed that. Nice.

ngn

@Marshall also k→j, not only fat clown, but fat clown with .-s and :-s :)

@Marshall what would be bqn→k?

Marshall

@ngn My opinion of K is not negative enough! I guess I see it as kind of limited or cramped, maybe like a mime in a small (imaginary) box.

Wezl

@ngn source?

ngn

@Marshall i was expecting something like a drunk football hooligan shouting abuse at everybody else :)

@Wezl own work

Marshall

20:33

@ngn Whitney at least is far too quiet about it.

ngn

@Marshall very true

Marshall

A good inverse-alien would be to make APL an early hominid or other human predecessor, but I'm not sure this works for J and K.

Wezl

could also add an "Everyone else" category

rak1507

I thought about that but I don't think there'd be enough difference between how does [a language] see everyone else

nathan rogers

@ngn much to my chagrin that symbol resolution point is moot

rak1507

20:37

@Marshall APL - neanderthal?

Marshall

@ngn Or a unicycler. Admirably simple form of transport; faster than walking; still not the best way to get around.

Wezl

21:02

@nathanrogers &prolog

rak1507

21:44

jsoftware.com/papers/KEIQA.htm#Clinton I think this is about a different ken iverson...

nathan rogers

23:44

CMC: Find an array oriented way to solve this "iteration game"

I don't know what this is, and someone asked me what is an idiomatic way to write this in APL, and I don't really know what an iteration game is, thought maybe I'd toss it to someone here who might actually know what that is

The optimization puzzle so far as I understand it is get to f0>= 1000 in the fewest moves

https://chat.stackexchange.com/transcript/message/57323387#57323387

UPDATE: Incremental game, not iteration game.

dzaima

@nathanrogers (fwiw, you can reply to yourself by manually copying the message ID from the URL and writing :1234, or by using a userscript)

Transcript for