The APL Orchard

@nathanrogers Uh, Expand is a function, it can't take an operand. In principle, one would extend Scan to be Expand when the derived function is used dyadically expansionVector(fillElement\)mainArray but that begs the question of how to specify fills for all other filling functions, e.g. ↑.

Also, what if you want a non-constant fill?

⋄ 1 1 0 1 1 0 1 0 1('ABC'{⍺⍺@(⍸~⍺)⊢⍺\⍵})'abcdef'

Dyalog APL

@Adám abAcdBeCf

nathan rogers

Then specify a fill vec that matches the number of 0 values

Adám

5:26 AM

OK, but that also implies that you have to enclose any non-scalar fill value (as indeed you have to in my above solution).

Furthermore, this only works for a Boolean expansion vector. How about ⋄ 1 1 ¯1 1 1 ¯2 1 ¯3 1\'abcdef' ?

Dyalog APL

@Adám ab cd e f

nathan rogers

Either a scalar expanded to fill or a vector of equal length to the expanded slots

Same as before

Adám

So this would need what, a 3 or a 6 element fill?

nathan rogers

3 in the case of the boolean vector

This is not a bit vector

So 6 in this case yes

It seems like type checking is not out of the question

Adám

What do you mean?

nathan rogers

5:31 AM

Check whether alpha is a bit vector or not

And branch thusly

Adám

I don't understand, what would be different in the two cases?

nathan rogers

Bit vector requires a default vector the number of false values, vectors containing negative numbers take a default operand length of the sum of negative numbers

Or scalar in both cases

Cases

Adám

Well, that isn't two branches, since you can mix zeros and negative values.

nathan rogers

Right but its either a bit vector or it isn't

Adám

Sure, but I still don't understand why a bit vector needs special treatment.

nathan rogers

5:34 AM

Can you mix 0 in with negative numbers?

I need an example

Adám

For sure, and positive numbers too. \ just requires an integer scalar/vector as left argument.

@nathanrogers ⋄ 3 1 0 1 1 ¯2 1 ¯3 1\'abcdef'

Dyalog APL

@Adám aaab cd e f

nathan rogers

Right

Adám

But the spec could be simple: Left operand has to be conformant with the number of fills inserted.

nathan rogers

So count of 0 + the sum of negatives

Would be the length of inner operand default

Or a scalar

Zeroes

Hard to type on mobile

Adám

5:38 AM

@nathanrogers Like ⋄ 3 1 0 1 1 ¯2 1 ¯3 1('ABCDEF'{⍺⍺@(⍸(1⌈|⍺)/0≥⍺)⊢⍺\⍵})'abcdef'

Dyalog APL

@Adám aaabAcdBCeDEFf

nathan rogers

That's ugly to me compared to x('A'\)y

Adám

Ugly or long? I find it rather elegant considering \'s odd definition. In any case, the non-Boolean case is rare, and the Boolean case is pretty simple: fill@(~exp⍨)exp\data in 18.0

@nathanrogers And again, how would we specify fill for / ↑ ⌺ ⍴ ⍤ ⊃ ? J uses a magic operator to specify fill…

nathan rogers

5:58 AM

expand←{(⍵,(-/≢¨⍺⍵)/⊂⍬)[⍋⍒⍺]}

this is what I'm doing right now. you could very well index your default value with ⍸~⍺ to obtain the correct default values

expand←{(⍵,defaults[⍸~⍺])[⍋⍒⍺]}

The specific use case is providing default values where 0=X.⎕NC name

I have a list of names and I compress the list to evaluate X⍎¨names

but I can't use each name if it doesn't exist, so I need to provide a default value for each term based on its data structure

how do I do an eval @Adám?

Adám

@nathanrogers Not ⍎?

nathan rogers

I mean in chat

Adám

@nathanrogers Just begin a code block with ⋄

nathan rogers

⋄{names←'ABCD' ⋄ defs←8×⍳4 ⋄ {⍵{(⍵,defs[⍸~⍺])[⍋⍒⍺]}ns⍎¨⍵/names}∊0≠ns.⎕NC¨names}(ns←⎕NS'').(A B D)←⍳3

Dyalog APL

@nathanrogers Illegal code

nathan rogers

6:09 AM

haha

it doesn't like it

Adám

@nathanrogers The bot is limited like Try APL.

nathan rogers

but you can see the inner function there {(⍵,defs[⍸~⍺])[⍋⍒⍺]}

Adám

@nathanrogers You can create a namespace of default values, and the replace the non-default values: 'defaults'⎕NS X

nathan rogers

fair enough

is there some reason why this form of expand is less good that @?

seems pretty direct IMO

probably that it isn't general enough to solve the negative case

Adám

6:37 AM

@nathanrogers Not really. And you can write defs[⍸~⍺] as defs/⍨~⍺

@nathanrogers Sure, but if it works for your needs, go for it.

@nathanrogers Did you understand the ⎕NS method for default values in a parameter namespace?

nathan rogers

not especially, I hadn't looked at the docs yet

Adám

@nathanrogers Here:

      (⍎'defaults'⎕NS⍬).(a b c d)←⍳4 ⋄ 1 ⎕JSON X
{"a":10,"c":30}
      (⍎'X'⎕NS ⍬).(a c)←10 30 ⋄ 1 ⎕JSON X ⍝ user input
{"a":10,"c":30}
      {'X'⎕NS⍵⊣'X'⎕NS defaults}⎕NS X ⋄ 1 ⎕JSON X
{"a":10,"b":2,"c":30,"d":4}

If you don't care about destroying defaults in the process, you can replace the last statement with:

      'X'⎕NS⍎'defaults'⎕NS X ⋄ 1 ⎕JSON X
{"a":10,"b":2,"c":30,"d":4}

rak1507

1:45 PM

{(9 11○⊂)¨(*⌾⊃⍵)×⌾/¨⍵⊂⍨(≢⍵)⍴⍳2} @Razetime here's a dfn for that rotation, just need to format it correctly

I think it works, in theory

Razetime

2:30 PM

@rak1507 um

which one exactly?

don't remember

rak1507

4

Q: Rotate Cartesian coordinates.

Write a program rotates some Cartesian coordinates through an angle about the origin (0.0,0.0). The angle and coordinates will be read from a single line of stdin in the following format: angle x1,y1 x2,y2 x3,y3 ... eg. 3.14159265358979 1.0,0.0 0.0,1.0 1.0,1.0 0.0,0.0 The results should b...

James Heslip

Is there any way to recover work from a ⎕OFF? Someone just lost a lot of code...

Morten Kromberg

@JamesHeslip I'm afraid not. Have to hope the code will be better the 2nd time round, it often is, at least if you slept the night before...

James Heslip

2:50 PM

@MortenKromberg Figured as much, but thought it was worth asking. It's fine- it didn't happen to me at least. A colleague was following the tutorial outlined in the Interface Guide (page 61), and fell victim to the ⎕OFF quit function in the temperature converter app.

Razetime

@rak1507 oh, for imaginary coordinates

cool

Try it online!

Made this one for palindrome distance

Anything I can imporve here?

9

Q: Palindrome distance

Find what is the distance for a given string to its closest palindrome of the same length. For this task I decided to give the characters further away from the string's center more weight (think of it as contributing more torque), proportional to their distance to the center. Let's define the pal...

code-golf string palindrome

Adám

3:10 PM

@JamesHeslip Consider autosaving.

Elias Mårtenson

⋄ 2*1000

How does the bot work?

Razetime

⋄2*1000

Adám

@EliasMårtenson Use backticks around your code. (Also, it doesn't pick up edits.)

Razetime

i think the bot isn't active now..?

Adám

Oh, indeed, it seems to be MIA

@Moonchild Any idea where the bot went?

Razetime

3:13 PM

but yeah it parses anything in backticks , that starts with⋄

Adám

(I should have opted for running it on a Dyalog server…)

@Razetime It should also work on code blocks that begin with ⋄ now.

Razetime

oooo

tried a different way, but longer

Elias Mårtenson

⋄ 2*1000

Anyway, I wanted to know if Dyalog supports bignums.

Anyone know the answer?

Razetime

@EliasMårtenson ⎕FR and ⎕PP are there

aplcart.info/?q=big%20int#

and there's a bigint arithmetic lib

Elias Mårtenson

I see.

I was thinking about implementing full bignum support in KAP, but I think I won't. :-)

dzaima

3:30 PM

@EliasMårtenson (fwiw dzaima/APL has bigints)

Adám

@EliasMårtenson Roger is working on bigrat(ional number ― a superset of bigint) support. You can follow his progress on the forums.

NARS2000 has too (and bigrats), and so does J.

Elias Mårtenson

Thanks.

Razetime

What are some good ways to shorten this?

Elias Mårtenson

I use Common Lisp most of the time, and having transparent bignums are useful. But I'm not entirely sure it's wirth the effort to implement it. Have you felt that it's ever been a life-saver in NARS?

dzaima

@Razetime use a train instead of a temporary x

Adám

3:34 PM

@Razetime Try it online!

@Razetime With ^^: Try it online!

train

i have 22

23 here

@dzaima 20

OK, I give up. I have dishes to wash… and I'm not supposed to be working today.

Razetime

3:40 PM

lol cya

dzaima

@EliasMårtenson bigints aren't needed often, but when they are, there's pretty much no way around needing them. So it depends on for how much you want to use the lang for. Most of my usage has been for base conversion

@dzaima (1: move the | somewhere else (in Adám's latest solution); 2: there's a way around needing that -∘)

Razetime

a way around it...

oka, down to 22

@dzaima the |, however isn't too obvious to me

Ok, now at 21

dzaima

@Razetime a prerequisite is having the train form for the starting part (i.e. no dfns)

Razetime

hmm i got to 21 with dfn

This is the same thing I did, but with a train

but it's 22

dzaima

@Razetime think about it :) (aka you're most definitely on the right track)

Razetime

3:54 PM

trust me, I'm really trying (:

dzaima

as a hint, my solution is a subsequence of yours

Razetime

4:28 PM

nice

Razetime

4:49 PM

How good is RAD?

dzaima

5:27 PM

@Razetime nice idea, but unfortunately somewhat unfinished

rak1507

7:34 PM

wonder how terrible of an idea writing an apl derivative in haskell is

Marshall

7:49 PM

@rak1507 I did early drafts of I in Haskell. Not a bad idea.

rak1507

wow, cool

dzaima

8:01 PM

well then

Marshall

@dzaima Your JIT compiler seems to be working in reverse.

dzaima

@Marshall my to-java compiler isn't even touched

Marshall

I'm seeing the same effect with that code.

dzaima

and i can't really blame javas JIT too due to that being a pretty convincing linear increase in runtime

Marshall

@dzaima I assume it's really some kind of resource leak.

dzaima

8:06 PM

@Marshall there is indeed RAM usage leaking (linearly too)

now to find where

Marshall

Memory usage doesn't increase when you wrap it in braces, so maybe the definitions (folded constants?) are getting leaked?

Could be something to do with REPL scoping.

dzaima

40k strings leaked per execution ¯\_(ツ)_/¯

ah yes
https://dzaima.github.io/paste/#07ZK9TsMwEMf3PoXVqVEl0gCVaCc@VoSEhNht50it@iPYCRWdOiFADDwAA2LiASqxgVRG3iJPgsuSa1@BkzL8/E/Ozp1/jLHQ3D0395@9bNA/TCL8vKzem8XbqQ3ddOIMpPmcK8PTQlXp8flZGrxM5Y64tt0OY6zXPD2ulrGsv1p@fzQPrwnLBs3iazducXKhDLDQGQ5H6yem2d5wlGXZ/s7BujYJUnUuuVdcaAjjGPUGCattHSCPizBm3Ht@G1FfjWMeoIpsectywn0kFBS6XYD3tZ1aN7N/322WSZdDmwh0gDCIJWK0gygQC8ToeOEQK8Q5YosYNSZmiMuW0a/dqM2GvNOoITQThTlDi1Ci7io3BavmsDE@WxvwSm6NzwPP4xuUaBAbdUYFwys5gVx4LqdQha1byUFq7gFNolzfJOrImVLhhkgBUoAUIAVIAVKAFCAFSAFSgBQgBUgBUoAUIAVIgX@qwFHFCu0E1yxIV8Iv#BQN

TessellatingHeckler

I'm a bit surprised that the compiled BQN byte code includes lambdas from the surrounding language in the list of Objects.

the Javascript and dzaima/BQN Java versions use the same bytecode, what about those functions?

dzaima

@dzaima two problems - variable overriding is stupid (i knew this very well), and •CTIME compiles the code as if it were to be executed in the global REPL scope and so needs to set vars

@dzaima another yay

Marshall

8:24 PM

@TessellatingHeckler Those are the values of primitives provided by the runtime. You could also represent them as their own type, as long as whatever method you use to call functions knows what they are.

Only the bytecode itself is shared between implementations. In a portable format, the objects would be serialized as well, but there's nothing like that now.

TessellatingHeckler

@Marshall I see, ok

dzaima

yay
https://dzaima.github.io/paste/#0lZQ9jtQwFID7nMLaakYrJbbjxMlU/LQICQnR246ZsSZxgp1hxVZbIUAUHIACUXGAlbZjpaHkFjkJTqbIm6lI5e8957P9nq0ghMTw8RuJUIDoNISYnuJpWHtlojfCGSFr7TchtcJrdLAHr6sQiA2i0dMebetWihp51XZ61HxYZvj0e0Xw9ZN1gL/fj7@Gu58vrL9Kdm2jk@pWmEYkW9Mnz169TLxTiYrlO3s12qvh65fjfdCuj/d/HobPP9aIYIyHu0caFnn@2jQa@YhwgjlhY5bkOSlYXLJ0gU5zxrKQeSRpxnGJY5qTRXqOy2zSKWcZiSmni/SUcDbppCxyGhc0W6ZzWp52D0EZs2KZnudZetLzIhtbxxfpPMV80lnwWcxZ@b@vJTyODRLOiQ8B67ebMO11H9iKmdVOuEAgsa3nQDt3sHvb3tjpu3NNtZWeMxJsIBvACjBYQW4BS8Bge9kCNoArwBYwKEzeAO5mBkd7b84Lcm0NCgI9MZAJCHwHquvbvbbmVp@1zx4a7Yy6aJ/TogozIFNreeY1xjeiVztdSSfUXvf@…

voidhawk

9:56 PM

Is there a better way to concatenate along the leading axis than something like ⊃⍪/⊂⍤¯1 ? I think there's some fractional axis specification magic here for catenate/laminate/reduce, but I'm not finding anything

Marshall

@voidhawk Raze but not in Dyalog. If you find anything good put it on that page though.

Moonchild

@Adám Interpreter ICE. (In rakudo, not dyalog.)

Marshall

Oh, but you just want to merge the two leading axes so it's a bit simpler. The fast way would be a reshape, but I don't know anything elegant.

@voidhawk Now I remember. It's ,[0 1]. Adjust for ⎕IO to taste.

voidhawk

@Marshall Ah right, thanks. There's usually some bracket-axis specification thing when I want to work with axes somehow (insert a new one, combine existing etc.). But to be honest using the bracket notation makes me feel kind of... gross.

Though I have started using ,[0.5] when I want to laminate two vectors

Marshall

@voidhawk Especially considering it's ,/ in J or ∾˝ in BQN (laminate's ,: or ≍).

Transcript for