The APL Orchard

ngn

18:03

@xpqz btw, i read your day19 but i think i'd better try to solve it without regex

xpqz

@ngn yeah, regexing bypass a lot of difficulty, I think. Although the problem seemed tailor-made for it.

ngn

@xpqz do the loops in part2 imply a context-free grammar, or is it still a regular grammar? i forget what we studied at uni, it was too long ago..

user

What do # and ## mean in Dyalog? Something to do with the current namespace?

ngn

18:18

@user # top, ## parent (help.dyalog.com/18.0/#Language/Introduction/Namespaces/…)

user

@ngn Thanks

Adám

@user Do you know you can type things like these, then hit the F1 key?

user

Cool, I didn't know that

Adám

Alternatively, you can enter ]help thing

user

Nice! That's going to make life a lot easier.

Adám

18:22

Even APLcart could give a hint here (and link to documentation) even if that isn't its main purpose.

user

Thanks, that's really useful.

By the way, is there some sort of builtin to show an array as a string so I can execute it and get it back again?

ngn

@user excellent question! :D

Adám

@user ⎕SE.Dyalog.Utils.repObj

ngn

@user i've been arguing for a long time that that should be the default output format

Adám

Hm, maybe I should various bundled utils to APLcart…

user

18:25

@Adám Thanks! This just saved me a ton of work!

Looks like SE has a ton of other useful stuff too.

Kamila Szewczyk

i wrote my first program with a custom APL dialect

⎕(+-|⍨)¨2⍫4

Adám

What does ⍫ do?

Kamila Szewczyk

range, this one generates 2 3 4

Adám

Why not …?

Kamila Szewczyk

one byte shorter than 1+i3

… would be much better here, yeah

Adám

18:28

That's what Extended and dzaima use.

ngn

range end should be exclusive

Adám

@KamilaSzewczyk aplwiki.com/wiki/Range

Kamila Szewczyk

@ngn I like my range ends inclusive

ngn

i can't argue against that

Adám

@ngn Does TMN's {a, a+1, a+2, … , n} include n?

ngn

18:34

@Adám is that an argument for end-inclusiveness?

Adám

Yes.

ngn

@Adám then does {a, a+1, a+2, ..., n-1} include n?

Adám

No. So in APL should a…n-1 include n? How about a…n?

ngn

does ⍳n include n (⎕io←0 assumed)? in a triganle rendered for computer graphics, why do the rightmost pixels not get rendered?

Adám

… is not ⍳. Not sure what the triangle has to do with anything.

ngn

18:37

think of dyadic ⍸. do the ranges ⍸ splits the integers into have their rightmost ends?

Adám

The rightmost (possibly only) one does. Since there's only one range here too, it should too.

ngn

the point i'm trying to get across is that a sequence of ranges: range(a[i], a[i+1]), range(a[i+1], a[i+2]), etc shouldn't overlap, otherwise there's a lot of +1 / -1 like with ⎕io←1

Adám

In Extended, they don't: 3…7…1 has no duplicates.

dzaima

@Adám that's a completely separate thing

ngn

@Adám idk what that means

dzaima

18:40

(fwiw i do remember needing to do an ugly 2…∘(-∘1)/ a couple times)

Adám

@ngn Try it online!

ngn

@Adám but

Adám

Now that monadic ≠ is gone, who's for monadic < and > as decrement and increment?

@ngn Surely. It returns the answer most people would expect, methinks.

ngn

@Adám and 10 is a dupe

Adám

@ngn The two ranges are unrelated.

ngn

18:43

ok, i've made my point

dzaima

@Adám but if they are, the dupe is annoying

Adám

@ngn Joining ranges.

dzaima

@Adám again, the question is about separate ranges. also that example is entirely pointless afaict

Adám

With < as decrement, …∘< wouldn't be so bad. Neither would …∘> of course, but I think … is more intuitive as-is.

ngn

@Adám dyalog is getting fat. you add something - you should remove something, i think.

Adám

18:49

I refuse to have this discussion again. Waste of time.

However, I very much agree that a new clean implementation (see starboard) should have a rationalised vocabulary.

ngn

@Adám yes, i *-ed that too :)

rak1507

github.com/rak1507/Advent-Of-Code-2020-APL/blob/main/… I did day 21, do you have an APL solution @ngn (or anyone else)

ngn

@rak1507 only k but with comments. i'm lagging behind with the translations to apl, and i'll probably never catch up as the problems are getting harder.

rak1507

yeah they really are, had a look at 19 and 20 and they look pretty tough, will tackle them just now

x^x^y the only bit of k code I can understand!

ngn

i still haven't solved 19 in a satisfactory way

rak1507

18:53

your code looks a lot simpler than my APL

ngn

@rak1507 i'll have to read yours carefully before i can judge

rak1507

there was another problem where I used a similar thing, 16

dzaima

random results of searching for … in my files:
https://dzaima.github.io/paste/#0VVJNb9NAEL3nVwy9dBe79a7bSMmhiIa0aZw2/IFenOCkRold2a4AtZUQIJSaGPWCUgkhobZATnDol1SpQkp@CftLeOuAEJZlz7zdmffm7Qbq7XGR1Ls31O4jDFbU0elkvDdDWgRILwx/tnKEqWFqmWo4sPi@YR0wFhiST29UNkZZu8UXDGsyZkK9/BrwlV/HVzptt9TgG2KVXR3qrFAoqPcf6o/RWuRNO4gYuzcdqcGJyi6kwVV2rtIzUsM71BEzLRCnr@6p7FKlp1j9aHKuseGdxlX2Ha/gpgTv5Afq87ZSCAGincTve3MdKs@BmO0zBjVLZXX0mSPIE1sdnf9NpF16wMFwa0CuSs/l8qZKv@imaLUKdfOVrXVnw3Ga9U2nuqE/NWdtzWk2nEaj6VTrjUa1UnOcWq3yqFJtNquNdTyrzfr8IayAj3oAZmuHigKkJ4tmSSdSCu1L9ome7sUJRW7wJOxTECYeJSHFXq9TgIi8Dl8dz8pKukwfnK1ljjFLdjnjG…

rak1507

Looking back, I actually used exactly the same code, apart from the variable name

ngn

ugh.. ⎕io=1 :( i wasted a few minutes wondering what's wrong :)

rak1507

19:02

haha

would be independent of ⎕IO if I used ⊃∘⌽¨ instead of 2⊃¨ or something

Adám

In a rationalised APL, ⎕IO would definitely need to go… Cue discussion on what it should be.

rak1507

2!

ngn

i'd recommend just using 0 but @Adám would probably attack that :)

rak1507

split the difference and make it 0.5

Adám

Yeah.

ngn

19:04

what i call the "solomonian solution"

rak1507

What's that a reference to

Adám

@ngn That's a misrepresentation of the Bible.

Judgement of Solomon

The Judgement of Solomon is a story from the Hebrew Bible in which King Solomon of Israel ruled between two women both claiming to be the mother of a child. Solomon revealed their true feelings and relationship to the child by suggesting the baby be cut in two, each woman to receive half. With this strategy, he was able to discern the non-mother as the woman who entirely approved of this proposal, while the actual mother begged that the sword might be sheathed and the child committed to the care of her rival. Some consider this approach to justice an archetypal example of an impartial judge...

rak1507

Ah

ngn

@Adám why misrepresentation? the article does say he was "suggesting the baby be cut in two"

afaicr the bible says that too

Adám

@ngn Yes, but that was only to get the true mother to react with "better a suboptimal outcome rather than something that is bad for everyone".

Having 0.5 or variable ⎕IO is bad for everyone.

He never actually intended to cut the baby.

ngn

19:08

@Adám well, exactly. i don't think anybody's seriously suggesting ⎕io←.5

rak1507

no but ⎕IO←2 on the other hand...

Adám

@ngn No, but clearly some thought a variable ⎕IO was a good idea. Acording to Solomon, the true APLer (?) should have been content with a suboptimal fixed origin.

ngn

i want 0. only 0. i'm not a true apler, i guess

rak1507

What if there was no way of directly indexing and ⎕IO became unnecessary anyway

Adám

Hm, an APL without indexing. So you'd only use ⊃⍤⌽ etc?

And things that now return indices would return bit masks?

rak1507

19:12

Ah yeah I forgot about things like ⍒, oh well

Kamila Szewczyk

is there a smart way of summing in a diagonal line in a matrix? for example, for 3 3⍴⍳9 it should return 15

rak1507

+/1 1⍉m

Kamila Szewczyk

why? it works

ngn

+/0 0⍉m

Adám

hehe

@KamilaSzewczyk Did you try APLcart?

rak1507

19:13

^ I accidentally had ⎕IO←0 and thought it didn't work

ngn

:P :D

Kamila Szewczyk

hm I may have missed that one

Adám

Whoever added )ORIGIN really did APL a disservice. Better fixed ⎕IO←1 than this mess.

rak1507

+/m⍉⍨2⍴⎕IO

dzaima

or how about ordinals are a separate numerical tower (just as characters are in BQN)? :p

Adám

19:14

+/m⍉⍨⊃⍳1 1

Kamila Szewczyk

@Adám what does )ORIGIN do

it spews out Incorrect command for me

ngn

@rak1507 in your solution: ⍵/⍨1-⍨2×.. could be ⍵×[2].. or something equivalent with rank (⍤)

Adám

@KamilaSzewczyk It was the way to set the index origin before the invention of quad-names. It isn't included in modern implementations.

Notice that almost all )commands have quad-name equivalents. The )commands are mostly kept for backwards compatibility, and a little bit of convenience.

Kamila Szewczyk

the fact that ⎕IO doesn't accept rationals is kind of disappointing

I always wanted to index from 3.14

Adám

@dzaima Didn't dzaima accept all numbers for ⎕IO at some point?

rak1507

19:16

@ngn hmm yeah

Adám

@KamilaSzewczyk Try it online!

dzaima

@Adám it did (but some builtins would assume it was just 0 or 1 (or an integer))

Kamila Szewczyk

@Adám i kind of knew that, because I've seen 'a'...'z' working

but I was thinking more of vanilla APL and having the rational index origin affect everything like the normal index origin does

rak1507

I wish there was a more straightforward way to map 0 1 to ¯1 1, I find it to be very useful a lot of the time

Kamila Szewczyk

at least I can do ⎕IO←2 :)

Adám

19:18

One problem would be what value to use where we currently use fractional indices. Integers?

ngn

@rak1507 -¯1*

rak1507

oh that's clever, thank you

simplifies a bunch of my desmos graphs as well!

ngn

@rak1507 common golfing trick

Adám

Common in production code too.

I used that in code I just wrote last week, for the 2021 competition.

@rak1507 I'll add it to APLcart.

rak1507

2021 competition 👀

I'm going to try and get some of my friends to sign up to it but I'm not sure how enthusiastic they will be

Adám

19:22

This rounds features sanity test of phase II submissions.

rak1507

Like online submission? Cool

How does that work with multi line dfns?

Adám

What's the problem?

Kamila Szewczyk

@rak1507 back when the 2020 competition started I didn't know APL pretty much at all, but I plan on trying it out next year

ngn

@rak1507 you know how many people i've managed to interest in apl? 0.

Kamila Szewczyk

I interested two people in J

rak1507

19:23

I was thinking if you used the same thing for phase 1 there might not be enough space to fit it in, but I suppose other text boxes exist

@ngn :( lets hope I can have some better luck

maybe the 'you can win money!!!' thing might appeal to them

Adám

No more generation of a giant namespace script to upload. There'll be a multi-line text input area which you can paste into, drag-and drop a file onto, or click a button to upload a file which will then populate the input area.

rak1507

That sounds way better

ngn

@KamilaSzewczyk that's great

rak1507

@ngn I think maybe 5 or so people have tried APL to the extent of doing a tiny amount of stuff in tryapl or whatever, with something more solid to work towards I think 2/3 people doing the competition would be optimistic but achievable

ngn

@rak1507 oh, from the competition, right

Adám

19:27

While we won't run the submissions against test cases, we will analyse what you submit to see if the required functions are present and support the required calling syntax.

rak1507

So users are still expected to test stuff themselves? Fair enough

Adám

For phase II, yes, as it is intended to be a bit more like "work" or a project IRL.

Only downside is that your top level functions must be explicit, but you can always wrap a derived function.

rak1507

What does that mean, the final functions can't be composed?

Adám

@rak1507 Btw, you may benefit from phrasing it as ¯1*~

rak1507

Ah yeah nice then I can use ≠ instead of =

Although ngn suggested doing it differently anyway

Adám

19:31

@rak1507 ?

rak1507

16 mins ago, by ngn

@rak1507 in your solution: ⍵/⍨1-⍨2×.. could be ⍵×[2].. or something equivalent with rank (⍤)

ngn

@rak1507 another suggestion in part2: ⊃{⍺,',',⍵}/ -> 1↓∊',',¨

Adám

@dzaima Am I drunk?

dzaima

@Adám :D

ngn

lol

rak1507

19:35

@ngn ah yeah that's clever but I prefer being able to visually tell that it's being joined

ngn

@rak1507 ok. btw, k has "separator"\ and "separator"/ for split and join. i miss them in apl.

rak1507

that's nice

Adám

20:05

Any feedback?

Wezl

good, except for too many pixels

I like the color

do you have a vector version?

Adám

Yeah.

@Wezl The ones spelling "APL"? Can I lower the count?

Wezl

no, it was a joke

Adám

Oh.

Wezl

but actually yeah, you can remove the second row from the top

Adám

20:17

Closing the gaps in A and P?

Wezl

 #  ### #
### ### #
# # #   ###

Adám

That's pretty unreadable.

Wezl

## ## #
## #  ## :)

Adám

I could offset the horizontal middle part of A and P half a pixel down, but that breaks the parallel a Boolean array.

(unless that implies double resolution)

Wezl

don't bother

Adám

20:22

But APL handles non-square matrices just fine.

rak1507

how come (,'a')∊,¨'ab' is false?

dzaima

@rak1507 you probably want (⊂,'a')∊,¨'ab'

rak1507

Yep, thanks

Adám

@rak1507 Because ∊ treats its arguments as collections, so it looks at elements of each, and the elements of ,'a' are just 'a' which doesn't occur in the right argument.

ngn

21:25

my favourite for now:

Wezl

Nice! here's my modified version:

ngn

gliders :)

Wezl

it kind of spoils the simplicity

ngn

@Wezl if it makes it more likable for more people, it's fine

we should make a poll or something to measure popularity

Wezl

the backround squares could represent some value in binary, right now they mean nothing

Konrad 'Unrooted' Klawikowski

22:31

hey @Adám can we continue tommorow?

I finish my semester tommorow with a physics exam and I'll have one month of holidays, seems like a good amount of time to learn more APL

rak1507

can anyone think of a nice way of generating all ways of choosing n from k

so given 2 5 it'd return (1 2)(1 3)(1 4)(1 5)(2 3)(2 4)(2 5)(3 4)(3 5)(4 5)

3 5 would be (1 2 3)(1 2 4)...(3 4 5)

Bubbler

@rak1507 dfns.cmat

rak1507

Ah, shows up in APLcart for 'combination' but not for 'combinations' (@Adám)

thanks

ktye

22:52

is "shriek" a known term in apl? it's used throughout the src code of apl\360 for "!"

dzaima

@ktye APLCart and Dyalog help know about it but I definitely didn't

Adám

23:20

@rak1507 Fixed.

@ktye Yes, fairly common name among old-timers. Works with RIDE's `` input too.

@Konrad'Unrooted'Klawikowski Sure.

Transcript for