The APL Orchard

Razetime

6:52 AM

funny idea: execute with axis value

Kamila Szewczyk

@ngn i'm trapping domain errors, because the streams are "infinite", which means they're going to halt eventually

and they'll halt because of a domain error in the iterator or mapping function

i don't want it to be handled by something else / displayed

@Adám hmm, right, i was thinking if i can somehow avoid using tradfns, but this feels really weird for me. there is no way to mutate a vector a few stack frames above in a dfn?

ngn

7:40 AM

@KamilaSzewczyk sounds fragile. what if a domain error is signalled for a different reason?

Kamila Szewczyk

yeah, it's quite fragile for now

i'm trying to rewrite this all as a class

ngn

@KamilaSzewczyk i don't see how tradfns or oop can help you, to be honest

Kamila Szewczyk

the API will be a bit more elegant, also i can hold some state variables

ngn

@KamilaSzewczyk can you post the source code in copy-pastable form?

Kamila Szewczyk

old?

okay, wait a second

ngn

7:43 AM

yeah, like in yesterday's screenshot

Kamila Szewczyk

pastebin.com/wZdUKswa

ngn

thanks

Kamila Szewczyk

no problem

ngn

what argument do you give to Stream?

Kamila Szewczyk

5 200 Stream '2×⍵' '1+⍵' '⎕←⍵'

ngn

7:46 AM

ok, that gives me a 0

Kamila Szewczyk

you could try forcing it to collect the results, by swapping out the 1st string as '⍣2×⍵', but that doesn't work

@ngn because the stream doesn't get collected

i needed to implement better limit, some sort of a working and non-erroring collect

∇ init aStart;aIterator will this work as a valid declaration of a tradfn taking two arguments?

i've seen it in some code somewhere, but i think aIterator here will be a local and not anything else

Bubbler

That declares aStart as the name for ⍵ and makes aIterator just a private variable

Kamila Szewczyk

okay, so i'll move it to the beginning

ngn

@KamilaSzewczyk i must be missing something. you don't call app at all

Kamila Szewczyk

@ngn if you ask for collection, then i append a finisher everywhere which calls app with the argument

ngn

7:58 AM

the first of ⍵ doesn't start with ⍣, so the part of the code that says "app⍵" is never reached

Kamila Szewczyk

@ngn in this case, yes

5 200 Stream '⍣2×⍵' '1+⍵'

this would work, if my code worked

which unfortunately doesn't seem to be the case

Razetime

@dzaima is it possible to have ranges to and from infinity in dzaima/APL?

ngn

@KamilaSzewczyk ah alright, now i get a value error for app

Kamila Szewczyk

if you have any tips regarding code style, i'll take everything

because i'm dead lost in my own code

ngn

@KamilaSzewczyk let me first understand what it's doing :)

i'm a slow reader

Kamila Szewczyk

8:01 AM

i'm both a slow reader and thinker

writing this code took me about two days

which is ridiculous

Adám

@Razetime What would that do? And can't you use Rank?

Razetime

Rank?

I want to use it for a question regarding interval notation

@Adám Try it online!

it takes union of each list, and then intersects everything after that

Adám

@Razetime Still doesn't answer what it would do.

Razetime

I want to use … with infinity and then perform set operations

Kamila Szewczyk

~~use my streams when they're done :^)~~

ngn

8:07 AM

@KamilaSzewczyk very strange: when i put ⎕←⍬ at the beginning of inner (in lim), it works

Kamila Szewczyk

what

Bubbler

(chat markdown for strikethrough is three dashes)

Razetime

@KamilaSzewczyk sure

@ngn how did you figure this out

Kamila Szewczyk

i can't even begin to understand why that would change anything

maybe i found third dyalog bug while writing my code :)

Razetime

@Adám This is the question

ngn

8:09 AM

@Razetime i tried my secret super-advanced debugging technique: ⎕←⍺⍵

Kamila Szewczyk

sadly not all software is idiotproof

@ngn i was debugging using ctrl+enter

Razetime

@ngn genius

Kamila Szewczyk

but sometimes the interpreter segfaulted

ngn

i'm sure @Adám will bring this to the attention of the relevant developers

it does look like a bug

Adám

@Razetime I still don't get where ⍎ with axis comes it.

Kamila Szewczyk

8:17 AM

@Adám what do you think of that code

Razetime

@Adám oh that was just a random shower thought™️

Adám

(What‽ Unicode has both ™ and ™️‽ What's wrong with them‽)

Razetime

but, think about it: execute with axis

Adám

⍎ doesn't even work on rank≥2, so axis makes no sense.

ngn

@Adám one of them should've been TMN :)

Adám

8:21 AM

If it executed each row of a matrix, then sure, ⍎[1] could execute each column, but seriously, that's like never useful.

Razetime

@Adám i know but funny command work on matrix

      A←⍉↑'2+3+4' '4+5+6'⋄⍎[1]A
  9 15

Adám

More interesting:

      (⍎,⍎[1])2 3⍴⎕D
12 345 3 14 25

Razetime

I'm gonna make my own dialect, with only this part changed

dzaima

@Razetime definitely no. And even if you could, no operations could be done on them as they'd take forever

Razetime

What would be a good name?

APLWithExecuteFirst

@dzaima alright :P

I just want to get it under 34 bytes fnsdfjsn

Adám

8:30 AM

@Razetime But why not simply make ⍎ a leading axis function and then use the regular Rank instead of the anomalous Axis? Execute←⍎¨,⌿

@KamilaSzewczyk @ngn I'll report it.

Kamila Szewczyk

I actually mailed earlier revision of that code to the support

because it segfaulted the interpreter

Adám

Good. Thank you, but this is a different issues. Do you want to email this too, or do you want me to log it?

Kamila Szewczyk

i'll mail it in a second

Adám

Super.

Kamila Szewczyk

> We have made important fixes to 18.0 in the time since then, so please visit our website and download our latest available Dyalog for Linux from our Download Zone.

hmm

Adám

8:43 AM

This latest bug reproduces in my month-old 18.1

Kamila Szewczyk

okay, i'll install that

where can i download 18.1

there's only 18.0 on the website

Adám

You can't. It hasn't even entered Beta yet!

Kamila Szewczyk

I got this error message when I opened RIDE:

> Could not load bridge interface functions. (Error 0x00000000: "libhostfxr.so: cannot open shared object file: No such file or directory")

Adám

Yeah, very annoying. I added a workaround to that yesterday.

Bubbler

@Razetime Your current answer is not even correct

Kamila Szewczyk

8:46 AM

okay, it seems gone

Razetime

@Adám comedy is more important than utility

Adám

@Razetime Funnier statement if you remove both "the"s.

Kamila Szewczyk

@Adám mailed

Adám

Thank you.

Kamila Szewczyk

for now, is there a workaround which won't obscure the console?

Adám

8:51 AM

Yes, ⎕←⍬⊤⍬

Kamila Szewczyk

hmm, okay, thanks

Bubbler

9:04 AM

@Razetime Fixed with the same byte count for you

Razetime

@Adám how do I make ⍞ output to the output window instead of debug?

Bubbler

@Razetime Why do you want that?

Adám

@Bubbler Because output without line break to stdout.

Razetime

I want to output without newlines to STDOUT for this

Bubbler

Outputting to STDERR is allowed by default, if you're solving a Code Golf challenge

Razetime

9:06 AM

Never output to stderr, unless EOF is read.

that's in the challenge description

Bubbler

wtf.

Razetime

or alternatively

how do I make ⎕← output without newline

Adám

Answer for both questions is that you can't directly do so. However, maybe you can redirect stderr to stdout…

Alternatively, you should be able to use ⎕ARBOUT to send raw bytes to the stdout stream.

Razetime

@Bubbler thanks a looot

@Adám this would make the ⎕: prompts come up in stdout

Adám

Wait, you have ⎕s too? Just use ⍎⍞.

Bubbler

9:14 AM

@Adám Can you? I tried various X values but nothing worked

Adám

I don't know much about this. Let me see if I can get Andy in here.

Razetime

@Adám wait how exactly?

Adám

He's in a meeting. I've ordered a notification when he's available.

@Razetime Getting input from ⎕ is almost the same as ⍎⍞

Razetime

oh ok got it

ok now how do I redirect our stderr stuf

Adám

You have to do that from the command line when you start APL.

Razetime

9:17 AM

oh well then

so much effort aaaaa

Adám

You're doing something unusual.

Razetime

well, onto the next question

old questions are bees

Bubbler

Just skip challenges with messy I/O. That saves time a lot

... or anything before 2016 or whatever

rak1507

9:32 AM

every time I look through the challenges in the SEDE query they all seem impossible :((

or very hard

Razetime

@rak1507 use the random question script that redwolf made

I've found a bunch of 200s from there

rak1507

@Razetime oh do you have a link?

Razetime

in The Nineteenth Byte, Dec 27 '20 at 14:32, by Redwolf Programs

Random Question 1.4 (Random Question S-1.4)

rak1507

dumb question how do I use user scripts

Razetime

@rak1507 also, here's my list of possibly solvable questions

@rak1507 chrome -> tampermonkey

rak1507

9:36 AM

wow that's a lot

Razetime

firefox -> greasemonkey

Dyalog APL

<moon-child> I thought greasemonkey had come under new ownership and was no longer trustworthy

Adám

I use Tampermonkey on FF.

Dyalog APL

<moon-child> and tampermonkey works on ff now

Bubbler

I have ~50s on my list, all for 200 and some 400

Razetime

9:37 AM

oh great I';; switch to that

Bubbler

though I haven't tried much recently

rak1507

wow lol I don't keep actual lists I just skim through the query

Razetime

if you do finish one of those in apl, just comment on the list and i'll add a checkmark

dzaima

alternatively, violentmonkey if you don't like that tampermonkey isn't open-source

rak1507

sounds like a good idea

@Razetime I would feel bad using your list

right which one of the monkeys do I use, too many options

Razetime

9:38 AM

the point is to prevent boredom, not hoard rep, so go ahead :P

@rak1507 tampermonkey

Adám

@Razetime I thought the point was to get ready for The competition…

Razetime

uhh it was initally to find easy 100s

then it became a graphical output only list

and now I just use it for boredom

If I was getting ready for the competition, it would make more sense to attempt actually hard math based questions rather than these

rak1507

the competition... I should probably do some actual proper hard APL practice so that I can do the hard ones

ngn

writing comments is more important than maths :)

Adám

For phase II, you'd want to exercise writing good APL. So yes, including comments.

I guess you could use CR for feedback…

ngn

9:41 AM

good as imagined by the organizers

Adám

Good as decided by the organisers.

Kamila Szewczyk

i don't know how should i prepare for the competition

probably i'll try myself at it but nothing will come out of it lol

dzaima

@ngn well, yeah. Everyone's gonna have a different definition of "good", you literally cannot satisfy everyone

Razetime

prepare for the competition by spamming andriy makukha

Adám

@KamilaSzewczyk Solve the problems from previous years.

Bubbler

9:42 AM

Is APL.TA dead btw?

ngn

@dzaima that's why it's not an objective criterion

Razetime

read the winners blog posts

Adám

@Bubbler Reminds me that I've not been on TA since like, forever.

dzaima

@ngn you're obsessed with having objective criterions :p

Adám

@Bubbler And no, it isn't, we've just not gotten around to putting stuff there yet, but are planning on using it.

Kamila Szewczyk

9:44 AM

the 1st phase doesn't seem difficult

so doesn't the 2nd phase from 2020

ngn

@dzaima cgcc is. and programming competitions in general are.

Kamila Szewczyk

true that

Bubbler

@KamilaSzewczyk Last one is draconian

dzaima

@ngn cause that's what the organizers decided. Code golf can have an objective definiton, "Good" cannot

rak1507

@dzaima I'm sure you're aware that programming competitions can have objective criteria

Bubbler

9:46 AM

Most programming competitions judge only right or wrong, with time limit

This is objective

Kamila Szewczyk

i think objective criterions for contests are essential

Bubbler

Other competitions with human judges are not

Adám

Then think of it more like a talent show.

ngn

in which the talent is not even shown? (i.e. solutions not published)

rak1507

In which the talent is writing an obscene amount of comments?

dzaima

9:48 AM

@Bubbler right, but pure performance isn't the goal of the APL problem solving competition. You can disagree with that it isn't, but within the limits that it is, there's not really much choice on how to do scoring

ngn

@dzaima performance??

rak1507

@dzaima correctness and then speed to break ties

Razetime

@Adám also writing efficient code is a bit far from code golf(although I could ignore that)

rak1507

more or less

Bubbler

Actually I don't like the most programming competitions way, and I'm fine with APL one

ngn

9:49 AM

@dzaima it's about correctness first of all

dzaima

@ngn ok right, i did forget that people can write incorrect submissions. But that shouldn't be the majority

rak1507

@Bubbler well you would be, you weren't the one who might have lost out on a prize due to subjective marker opinions

dzaima

(of course, one should judge being wrong much more harshly than there not being any comments)

ngn

@dzaima exactly

Adám

@dzaima Oh yes, and we do.

Razetime

9:55 AM

I wonder if there's a person who submitted lookup tables for all the questions

rak1507

lol

Bubbler

Personally I'd recommend getting more familiar with math challenges, and some linear-algebra

Razetime

as I expected

also getting familiar with writing efficient APL

Adám

@Razetime You can't, because phase I uses random arguments, and for phase II we don't publish what we'll test with.

rak1507

linear algebra 😱

Razetime

9:56 AM

@Adám there's always a big enough lookup table for any given problem

Bubbler

As an example, I used ⌹ for 2020 Phase 2 Q9 to solve an important step

Adám

@Razetime We limit submission size…

Razetime

boooooo

rak1507

@Bubbler if that's mobiles, it does tell you to use that, so it's quite obvious

ngn

@Adám are you going to publish the tests after the contest?

rak1507

9:57 AM

iirc

Bubbler

But it takes a little bit of (mathematical) effort to use it without errors I think

or to understand why it works in certain cases

rak1507

it requires less insight than other linear algebra might require though, which is the hard bit (to me)

Bubbler

Also some basic number-theory and maybe combinatorics

APL is a mathematical language after all

Knowing to write some regex helps too, mostly for string processing problems

Razetime

One thing I have learned properly from code golf is regex

ngn

@Bubbler if anyone knows how to write regex, it's rak1507 :) (judging by aoc20)

rak1507

10:27 AM

lol

if the questions are too hard for me I can always try and persuade my smarter friends to learn APL and get the referral prize

RikedyP

10:42 AM

@rak1507 While you really should understand the problem and solution well to write the code and comment it such that we can understand your understanding, there is really no way for the judges to prove that you figured out the solution without talking to people / googling (I'm also not sure that's a criterion, but obviously copying the dfns library verbatim isn't gonna be considered great). This is not official advice.

rak1507

to me talking to people and googling are two very different issues, googling is researching yourself, and talking to people is basically cheating to me

RikedyP

@rak1507 Then we are all hopeless

rak1507

Arguably last years winner won because of his googling and research into the different algorithms for question 8

RikedyP

@rak1507 It is not a test under exam conditions, people have months

The crux of the "good" to me really is the (fundamentally subjective) aspect of good programming - expressing ideas to other humans in the form of executable code

rak1507

Yeah that makes sense

RikedyP

10:45 AM

For which you almost always have to make trade-offs in the real world between time spent coding, execution speed, memory usage, readability and maintainability etc.

chrispsn

@Adám sorry to drag comments from earlier - i found this comment from AW interesting re: depth and the history of k development web.archive.org/web/20041020215135/http://www.kx.com/listbox/k/…

RikedyP

@Bubbler Well partially all it takes is for people to post their specific questions as a thread there and post a link here, but this place seems to have evolved quite naturally over the years so I'm not gonna be the one dickhead going "please, please guys... hey guys can you post your Qs over there please". Really I should be doing that for both TA but also getting people to repost certain questions on stack overflow as well.

chrispsn

even with @ngn's 'deep where' it can sometimes be awkward to target parts of a nested structure, but i guess the point in k is that keeping things flat often keeps it simple

i haven't had a chance yet but i'm looking forward to digging into aaron/johnE/arthur's work on trees

ngn

11:57 AM

@chrispsn that archive is a real treasure

xpqz

@chrispsn Love the fact that AW's obsession with conciseness extends to a one-letter email address.

Kamila Szewczyk

how does one ⎕FIX a class?

is it possible to stop the status output window from appearing?

and have the errors logged elsewhere

Adám

@KamilaSzewczyk ⎕FIX ':Class myclass' ':field public answer←42' ':EndClass'

xpqz

@RikedyP Having completed the beta test of the comp, that was a concern I raised -- it wasn't quite clear to me if this is a problem-solving competition, or a "beauty contest". If a set problem can be solved trivially, or significantly easier by using a routine found in the 'dfns' workspace -- arguably the closest thing Dyalog has to a standard library -- to me, that's a great, practical, production-oriented path to take.

If you don't agree (the competition management committee) - then don't set problems where this is the case.

ngn

12:12 PM

@Adám ping?

Adám

@ngn me or Dyalog?

ngn

dyalog

Adám

I don't know. I assume not. And afaik, nobody has ever requested that we do that. You are free to use the official contact email to request otherwise.

ngn

:D

Kamila Szewczyk

one mistake i constantly make while writing APL is vectors of strings

Adám

12:18 PM

@KamilaSzewczyk We were just discussing this annoyance today. For now, you can hack it away by editing ide.js where there's only one mention of StatusOutput. I assume that making that function be a no-op solves the problem.

xpqz

@RikedyP as to plagiarism/research -- that's a grey area, right? Half a century's worth of the best APL code snippets are to be found in APLCart. It's part of my workflow now to first try to articulate what I want to achieve and enter that into APLCart. Is that acceptable?

Kamila Szewczyk

for some reason i end up iterating on individual characters, sometimes i end up joining them, sometimes i end up splitting the string and catenating each letter in boxes

the nest operator seems to be a godsend in most cases

Adám

@KamilaSzewczyk You're welcome.

rak1507

@xpqz most things in APLcart are pretty short snippets/idioms, it's not as if a complicated problem will be easily solved by looking in it

xpqz

Sure. What about Rosetta Code?

rak1507

12:22 PM

Again, it probably has some ideas you can use but not full (or very good) solutions

To me all online resources are fair game, but asking others for hints/tips shouldn't be allowed, but other people might think differently

ngn

@xpqz the dfns workspace is better for more advanced stuff

Adám

@KamilaSzewczyk Trick: {⎕SE.Dyalog.Utils.repObj⍎⍵⊣'∊'⎕ED⍵}'cv'

xpqz

@ngn There sure is a wide variety in there. Some of it your doing, I note.

ngn

@xpqz hm, not really. i have only a few small contributions.

Kamila Szewczyk

new syntax for streams, for now

map/each/filter/limit now works, collect also works

ngn

12:27 PM

oop - no thanks

Kamila Szewczyk

lol

i wish APL had more scala-ish model of objects, not classes

i'm using a class only to keep my state in a single place, pretty much nothing else

Adám

@KamilaSzewczyk Maybe consider a namespace?

Kamila Szewczyk

smart, but then you can't have streams in streams

Adám

Because…?

Kamila Szewczyk

i wish i could get rid of this ugly eval, but it doesn't look like i can get rid of it

@Adám because namespace variables are static, not bound to a single stream

you can't interleave operations on two streams

Adám

12:32 PM

@KamilaSzewczyk And it works with classes?

Kamila Szewczyk

yes

you just create a new instance

Adám

@KamilaSzewczyk Just create a new namespace, no?

Kamila Szewczyk

the only other choice is passing around the state as a boxed array through a couple of dfn's

ngn

@KamilaSzewczyk can we see your source code?

Kamila Szewczyk

sure thing

Adám

12:33 PM

Not as a screenshot, please.

Kamila Szewczyk

pastebin.com/ADXR3TrC

it's not perfect, at least for now

but it's much more cozy to work with for me

ngn

i should have said, can we copy-paste your source code :)

Kamila Szewczyk

why would you copy paste this crap

i plan on putting it in my APL dialect prototype soon-ish

anyways, yes, CC0 for that one

@Adám i was feeling that huge code walls may be annoying to someone

autism moment, kind of

ngn

@KamilaSzewczyk to try to understand what's that thing that cannot be done with dfns, but can be done with oop and tradfns

Kamila Szewczyk

obviously it can be done with dfns, but dfns are very annoying to work with

Adám

12:36 PM

@KamilaSzewczyk SE chat limits the display size, with a link to show all.

Kamila Szewczyk

right, i forgot about that

ngn

@KamilaSzewczyk ok..

Kamila Szewczyk

here are a few reasons why i picked this over dfns:

1) the code feels more readable than keeping the state in a single boxed array
2) no scoping problem, no dyalog bugs to work around
3) tradfn's don't return after a single non-assignment statement

rak1507

you can get around that by doing _←whatever

Kamila Szewczyk

for comparison, the small dfn you've seen took me around a day to write, and the thing you've seen a second ago took me 10 minutes to write, debug and test

rak1507

12:38 PM

or using ⊣

ngn

@KamilaSzewczyk 0) ? 1) readability is in the eye of the reader 2) that's a bug, i believe 3) i agree

Kamila Szewczyk

spot the ⎕IO←0 warrior

lol

rak1507

lol

Kamila Szewczyk

the thing i don't like about the current solution is it's verbosity

this felt almost as bad as java

rak1507

yeah it does look a bit grim

Kamila Szewczyk

12:41 PM

:If 0≠≢'aStart'
    current←⊃aStart
    :If 2=≢aStart
        iterator←2⊃aStart
    :Else
        iterator←'⍵+1'
    :EndIf
:Else
    current←1
    iterator←'⍵+1'
:EndIf

i had no idea how to make it not-bad

ngn

@KamilaSzewczyk write 1000 programs like that one, and you'll be able to do it in minutes :)

Kamila Szewczyk

fun ~3.5 years ahead.

ngn

maybe more :)

rak1507

0≠≢'aStart' huh?

did you mean to put quotes around that?

Kamila Szewczyk

oh no