The APL Orchard

12:55 AM

anyone know of any good video tutorials on using the Dyalog UI covering basics like loading a workspace, configuring APL settings, how to use the debugger, etc.

Adám

@gwa000 Not yet, but we're about to launch a series covering exactly that.

gwa000

adam, don't want to pin you down but are we talking a few days, weeks, months? if it is too far off i'll start reading some documentation although my goal was to hopefully watch some videos late at night.

att

7:11 AM

Is there way when displaying multidimensional arrays to alternate the directions of axes

PyGamer0

@att what do you mean by that?

att

so for example `2 2 2⍴0` might be
```
0 0 0 0
0 0 0 0
```

dzaima

@att just transpose the array before displaying it

att

only the last axis goes horizontally, from what I've seen

dzaima

oh. if you want multiple axis horizontaily, you'll need to do something more fancy

att

7:14 AM

ah, ok then

dzaima

then there's the question of whether 2 2 2⍴⍳8 prints to
1 2  3 4
5 6  7 8
or
1 2  5 6
3 4  7 8

RikedyP

9:13 AM

definitely months before a substantial series, but have a preview of some drafts. If you have feedback it's appreciated. 5 minute vids is the goal: https://youtu.be/sSfspfApNNg
https://youtu.be/TEz0b7mHQbM
https://youtu.be/-sz-MFAAUOU
https://youtu.be/stIo5mNLgWA

Richard

12:13 PM

@RikedyP thanks! Very usefull, watching them now

Adám

12:26 PM

Announcing: Dyalog Version 18.4.1

PyGamer0

@Adám What on earth lol

oh i understand

i like it

Adám

Also: TC39 has moved APL language syntax in JavaScript to stage 5

Richard

1:21 PM

@Adám nice! Will there also be a new keyboard supporting this?

Adám

2:05 PM

@Richard Already rolled out, in fact: Hold the ⊞Win button and tap .

Richard

p.s. I'm ready for the quest or did I miss something?

Jeff Zeitlin

@Adám - /me snrchs with the effort to not laugh out loud, after carefully noting the date

Adám

@Richard 15:00 UTC.

Richard

ah :) Daylight Savingtime started here last sunday...

PyGamer0

why do people still follow day light savings

Richard

2:11 PM

Like the rest of Western Europe ... ;)

PyGamer0

it literally makes no sense

Adám

@Richard Not Iceland.

Richard

Also an Island

@PyGamer0 Some fierce debates about this as you probably know. For the same reason nothing will change

PyGamer0

> fierce debates

i dont know

Richard

The EU had a proposal in 2018 to cancel de DST. Debates started in every country. With the new chairman, which changes every year or so, being Portugal, it is off the agenda again. Original reason was saving energy by shifting the clock, so most of the active part of the day is in daylight.

On the other hand people claim it is bad for their biorithm and have a jetleg after the shift.

Maybe just like the ⎕IO←0 discussion :)

Fmbalbuena

2:38 PM

@Adám Is 🌜🤔🌛 in 18.4.1 equivalent to {⍵}?

PyGamer0

yeah

Adám

@Fmbalbuena Yes. The entire table is:

📢!
🌳#
🐣&
✍'
🤱(
🎅)
🤯*
🧮+
😍,
🥖-
🔴.
🌿/
💂:
👉<
🙂=
👈>
🎲?
🌍@
🔎[
🤰\
🔍]
🌜{
📏|
🌛}
🙅~
🍑¨
🥶¯
🧭×
🪓÷
🏃‍
🏴‍
🏁←
👍↑
🥅→
👎↓
😳∇
📜∊
🤐∘
🐜∧
🛶∨
🙏∩
🦄∪
😠≠
😊≡
😡≢
☝≤
👆≥
🔒⊂
🔓⊃
🔏⊆
🤣⊖
⏩⊢
🔙⊣
🔡⊤
🔢⊥
🚧⋄
📈⌈
📉⌊
🚀⌶
👇⌷
🔑⌸
🐇⌹
📝⌺
🙃⌽
🧬⌿
🌱⍀
🎶⍉
🛫⍋
🛬⍒
💾⍕
💭⍝
📩⍞
🌴⍟
🔧⍠
😬⍣
😮⍤
😱⍥
😵⍨
🥰⍪
😶⍬
⛵⍱
🚫⍲
📚⍳
📐⍴
🤔⍵
👀⍸
🥺⍺
📤⎕
🥧○

PyGamer0

nice

except for few, i like it

🧮😵🌿🙃1 5 3 8 2 3 4 🥶5 🥶1

@Adám what is today's quest?

Richard

problems.tryapl.org/psets/…

rabbitgrowth

Love the readability of 🎲

xpqz

2:55 PM

Having 💭 denote a comment would be awesome

rabbitgrowth

Thought ⍨ and ⍣ would be 😕 and 😘

Adám

Welcome to the APL Quest! Today's quest is It Is a Moving Experience:

> Write a dfn which produces n month moving averages for a year’s worth of data.

xpqz

+⌿÷⊣

PyGamer0

My solutions:
APL 18.2: {⍺÷⍨⍺+/⍵}
APL 18.4.1: 🌜🥺🪓😵🥺🧮🌿🤔🌛
ngn/k: {(+/'x':y)%x}

rabbitgrowth

I started with {⍺÷⍨⍺+/⍵} and was pleased to discover that it could be simplified into +/÷⊣ :)

Adám

3:00 PM

Yup, that is indeed very nice.

rabbitgrowth

Is it better conceptually to write +⌿÷⊣ even if I'm sure the argument is a vector?

PyGamer0

huh no one posted an 18.4.1 solution yet, seeing as its the future of Dyalog APL :P

i tried making a bqn solution, failed

Adám

@rabbitgrowth Either is fine, but maybe one day you'll copy your snippet to a higher-rank context.

Nobody thought to do (+/÷≢)¨,/?

What happens if you ask for n>12?

PyGamer0

@Adám i though about it last week

Adám

@Ioanna Hi there. You're clearly interested in mathematics. Maybe in APL too?

Ioanna

3:06 PM

Hi! Yes, I am definitely interested in APL

Just hard to find the time to learn more, but here I am now :)

Adám

Cool. We're just in the middle of discussing this problem.

Ioanna

Thanks for the link - taking a look

xpqz

@Adám length error?

Ioanna

hm perhaps I should try to understand the solutions mentioned above

Adám

@xpqz Right. Now, what do you think should happen?

Richard

3:08 PM

@xpqz but not for n=13

Adám

Exactly.

For 13, we get an empty vector of averages, essentially saying that there are no length-13 sub-vectors.

I'd argue that for n>13, that should be the answer too. How might we implement this?

Brian

This is why I argue that windowed-reduction should be changed.

Richard

conditional statement? Can;t figure out something else

Adám

@Richard How about clamping the left argument to the maximum allowable value?

Fmbalbuena

@Adám What a waste of time to learn all.

Adám

3:14 PM

@Fmbalbuena wdym?

Fmbalbuena

@Adám I don't want to change because this is waste of time to learn...

Adám

Check the date (and minor version number).

Fmbalbuena

@Adám ... okay, this helps me?

Adám

@Richard OK, so how would it look with a conditional?

Fmbalbuena

@Adám Oh, thanks

Richard

3:17 PM

⍺<≢⍵:⍬

xpqz

{(13⌊⍺)÷⍨(13⌊⍺)+/⍵}

yuck

rabbitgrowth

(⊣⌊⍴⍤⊢)(+⌿÷⊣)⊢ seems to work, but I feel like there's a better way to do this

Adám

@xpqz Hold on, if (13⌊⍺)+/⍵ gives an empty vector, who cares what you divide it by?

@rabbitgrowth You could swap arguments on the middle tine: ⊢(+⌿÷⊣)⍨⊣⌊≢⍤⊢ or even ⊢(+⌿⍨÷⊢)⊣⌊≢⍤⊢ but this uses 12 instead of 13, which isn't right; the result will be the overall average instead of the empty vector.

Richard

{⍺>≢⍵:⍬ ⋄ ⍺÷⍨⍺+/⍵}

Adám

That's quite readable.

xpqz

3:21 PM

I was just about to say the same.

rabbitgrowth

Ah yes, should have used ≢

Adám

1+≢ in fact.

There's a silly way of making that a tiny bit shorter.

rabbitgrowth

Are you swapping the arguments just to avoid the parens?

Adám

Me personally? Yeah. That's my style.

Richard

yes I am

any other ways to cope with n=14?

Adám

3:25 PM

Than the clamping with ⌊?

Richard

yes

Adám

You could catch the error.

xpqz

0::⍬

Richard

how does that work? You had some video's about that but didn;t watch them yet

Adám

Better would be {5::⍬ ⋄ ⍺÷⍨⍺+/⍵}

Richard

3:27 PM

is 5 the error code for ' Range Error' ?

Adám

:: takes error numbers on the left (0 for all errors, 5 for only length errors) and a return expression on the right. The expression is evaluated and returned if any such error happens subsequently.

xpqz

      ⎕em 5
LENGTH ERROR

Adám

Sometimes, I wish there was an "Adverse" operator, which applies/returns its right operand if the left operand errors.

Richard

why is n=13 returning ⍬ en n=14 the length error? What is special about 13?

Adám

13=1+12

Richard

3:28 PM

yes and 14=1+1+12

gwa000

exactly the type of videos i was asking about.

thanks!

Adám

@Richard Yeah, this is what Brian meant. He'd like any too-large left argument to give the empty result.

Richard

ok

rabbitgrowth

I'm staring at the function bar in RIDE trying to figure out what that silly way is

Richard

just for fun, is there an other way to implement n+/⍵ e.g. with outer product?

Adám

3:32 PM

@rabbitgrowth My idea is for replacing 1+≢⍤⊢ in ⊣÷⍨⊢+⌿⍨⊣⌊1+≢⍤⊢

@Richard You could use Stencil, but it gets awkward. @xpqz spoke about this at APL Seeds '22.

Richard

@Adám ah missed that.

xpqz

Soon available on catch-up

Richard

thx

Adám

_Adverse_←{⍺←⊢ ⋄ 0::⍺(⍵⍵⊣⊢)⍵ ⋄ ⍺ ⍺⍺ ⍵}
(+/÷⊣)_Adverse_⍬

⍺(⍵⍵⊣⊢)⍵ is a trick to allow both function and array right-operands. In our case, it could be just ⍵⍵

Bojan Petrović

maybe something like this can be used for windowing: (¯1+⍳5) ∘.+ ⍳5

Adám

3:36 PM

@BojanPetrović Ah, yes, you could index with that.

Richard

yes was tryinbg something like that

nice!

Adám

However, it goes too far. You need to compute how many sub-vectors you'll need.

rabbitgrowth

Ah, is it ≢⍤,?

Adám

Indeed!

gwa000

@RikedyP will there also be a separate set of videos using the new emojis adam referenced in 18.4.1? and we'll need a whole new set of manuals, etc. :>)

rabbitgrowth

3:40 PM

That was fun to figure out.

Ioanna

I'm still having fun trying to understand the several proposed solutions (total noob)

Richard

{×(⍳⍺),(⍺-⍨≢⍵)/0}

for masking

Adám

And then rotating that?

Richard

yes maybe. Still thinking :)

Adám

@Richard {(≢⍵)↑⍺/1}

Richard

3:44 PM

ah :)

Adám

Fun alternative: {1/⍨⍺,⍺-≢⍵}

(You can also use \ instead of /)

xpqz

@Ioanna The clearest one is probably {⍺>≢⍵:⍬ ⋄ ⍺÷⍨⍺+/⍵}.

Ioanna

thanks for the tip @xpqz

Adám

@xpqz Agreed, but how would you adopt it to allow higher-rank right-arguments?

Brian

+⌿

xpqz

3:47 PM

What Brian said

Adám

That's not enough. ⍬ is wrong.

E.g. for 7 and ⍪⍳5 it should give 0 1⍴0

xpqz

{⍺>≢⍵:0⍴⍴⍵ ⋄ ⍺÷⍨⍺+⌿⍵}

Adám

That's the same as ⍬

The idea is that we need to empty out the leading axis.

xpqz

(0,1↓⍴⍵)⍴⍬

Adám

Sure, or just 0⌿⍵

xpqz

3:53 PM

ofc

Adám

btw, if it wasn't for this, 0@1⍴⍵ might be better, as it replaces the value in-place.

That's 0@0⍴⍵ for the ⎕IO←0 crowd.

xpqz

{⍺>≢⍵:0⌿⍵ ⋄ ⍺÷⍨⍺+⌿⍵}

nice

Adám

I'd say that's the industrial quality version.

Richard

:)

xpqz

Two points to @Richard

Adám

3:55 PM

⊣÷⍨⊢+⌿⍨⊣⌊≢⍤, is the golfer's version.

xpqz

What if ⍺ is also higher-rank?

Or at least a vector

Adám

How would you make sense of that?

Sliding window of varying size…

xpqz

like a product -- mapping different values over the cells of ⍵

Richard

an matrix of 'sales vectors' with a corresponding ⍺

xpqz

^^ yes

contrived, admittedly

Adám

3:57 PM

{⍺÷⍨⍵+⌿⍨⍺⌊1+≢⍵} would be the no-branch-prediction-failures-fanatic's version.

Shall we say we've exhausted this problem?

xpqz

yeah

Adám

See you next week for Solution Salvation.

xpqz

Sadly I'll miss next week's masterpiece. I'll leave this one here for you for then: ⌹

Richard

thanks again. Looking forward to the video again

@xpqz :), but now without

Adám

@xpqz Don't tell me you'll be playing dominoes instead of attending!

rabbitgrowth

4:04 PM

> Hint: this is the easiest of all of the problems presented here.

?!

Kamila Szewczyk

easiest provided you know the builtin

silently hoping someone implements the Cramer method or Gaussian elimination

rabbitgrowth

4:15 PM

@xpqz Lol, I thought that was a hint, like "here's an essential ingredient you need to solve this problem"

xpqz

@rabbitgrowth it sort of was..

LdBeth

4:42 PM

it is only recently I realized Xm≡(⌹Ym)+.×Ym+.×Xm

att

7:27 PM

@dzaima I was thinking alternating axes (the second)

dzaima

then you can (⊂⍤2)2 2 2⍴⍳8

att

7:42 PM

{⊂⍤2⍣(⊃⌊2÷⍨⍴⍴⍵)⊢⍵} more generally then

dzaima

⊃⌊2÷⍨⍴ → ⌊2÷⍨≢

att

why ≡/?

dzaima

oh that's font misrendering the Tally character. this is how it should look like

att

ah

Brian BED

10:12 PM

      'C:\Users\brian\Persinal\Scripts\APL\more\jónRói'⎕NPUT⍨∘⊂'test'
FILE ACCESS ERROR: C:\Users\brian\Persinal\Scripts\APL\more\jónRói: Unable to create file ("Access is denied.")
      'C:\Users\brian\Persinal\Scripts\APL\more\jónRói'⎕NPUT⍨∘⊂'test'
                                                            ∧

why am i getting access denied even as administrator?

like... what i want to do is write to a text file

or even be able to make one

dzaima

maybe the directory doesn't exist? does it work with an ascii filename? how about a relative path?

Brian BED

oh wait... its a clear ws

is that the issue maybe?

and i don't know what you mean by ascii filename

dzaima

without non-english characters like ó

probably not the cause as everything should understand unicode nowadays, but worth a try