The APL Orchard

nathan rogers

00:00

name me an implementation of any language that doesn't find itself at the brunt of lots of complaints about how things work

ngn

@nathanrogers bqn?

nathan rogers

bqn has users?

Adám

@nathanrogers bf

ngn

@nathanrogers 2 of them :) which is more than avg for a programming language

rak1507

@Adám I thought you said before a lot of people who work on APL code bases dislike it?

Adám

00:04

@rak1507 Where?

rak1507

hang on I'll try to find it

Adám

Doesn't sound like a very "me" thing to say.

rak1507

I'm paraphrasing

Adám

How would I know? I barely know anyone who works on real code bases.

phantomics

I bet a lot of outsourced devs hired to update APL codebases by managers who don't know the difference between APL and Java have had coronaries when they saw the code

Adám

00:06

@phantomics Because they were all over 80?

phantomics

Because they never saw a ⍴ or a ⌈ in their Java and Python bootcamps

Adám

@phantomics Surely managers don't hire Python devs for Java work.

phantomics

Don't underestimate them, non-technical management can make all sorts of entertaining decisions

rak1507

frustratingly, I can't find it, maybe I was dreaming but I'm sure someone at some point mentioned something along those lines

Wezl

@rak1507 I'm sure that's true ::)

ngn

00:22

@phantomics judging by the number of apl-like impls, most people think that it is :)

rak1507

this is very annoying, can anyone else remember something like this? I'm sure at some point someone said something along the lines of people who work on the legacy code would prefer porting to a new language but that's not possible

ngn

00:52

@rak1507 this sounds close

rak1507

nah, not that, something else, I must have imagined it lol

ngn

if it was someone from dyalog, it must have been framed very differently

Marshall

@rak1507 Morten says something like all Dyalog's major clients have tried to switch to other languages at some point and given up. I don't remember anything about individual programmers though.

rak1507

ah well maybe I'll stumble across it in the logs one day and then I'll look like less of an idiot

phantomics

APL: the quicksand pit of languages?

Marshall

00:59

@phantomics I mean, only after you write a few million lines of code with it.

Adám

@Marshall Wouldn't that be the case with any language?

rak1507

is there a way to search for symbols/punctuation?

Adám

In SE chat, you mean?

ngn

@Adám if it's a parsable language it could be transpiled, at least in theory

rak1507

@Adám yes

Adám

01:01

@ngn Can you tell me any PL that can reliably be transpiled to another language, even though neither language was designed with that goal in mind, nor are the two languages trivial variations of each other?

ngn

@Adám you're sneaking in extra conditions there :)

Adám

@rak1507 No. But I think @dzaima "downloaded" the entire transcript so it could be searched.

@ngn Only to exclude silly answers.

user

@Adám This is probably not what you're talking about, but Scala can be (somewhat reliably) transpiled to JS, as can a bunch of other languages

phantomics

Lisps can be transpiled into lots of things, in some cases requiring that you use a particular dialect to be transpiled into the target language

dzaima

@Adám (i tried to search for what rak was trying to find with it, but failed to find anything)

ngn

01:04

@Adám the perfect example is scheme - simplest of all syntaxes, well-defined semantics

phantomics

THe Lisp machines had compilers for C, Fortran and other languages that generated Common Lisp code

rak1507

@dzaima would it be possible for you to grep for ...

Adám

Being able to transpile the syntax isn't enough. PLs are a lot more than syntactic rules.

phantomics

@Marshall Are there seriously million-line APL codebases out there?

ngn

@Adám on the other side of the transpiler - probably the jvm, as there are so many languages targetting it (though it would probably be a compiler in that case)

Marshall

01:06

@phantomics Just going by memory here, but yes I believe there are a few.

dzaima

@rak1507 4486 messages

rak1507

possible for you to send them in a file somehow?

phantomics

@Marshall Wow, I'd think a few thousand lines should be enough for a strong AI

user

@ngn The distinction between transpilers and compilers is somewhat blurry. I've seen tutorials on how to program in JVM bytecode directly.

dzaima

@dzaima (also includes ones that have been truncated with '(see full text)' because i was too lazy to get those; remind me to do that tomorrow pls)

Adám

01:08

@dzaima Would it be possible for you to create a public repo with the content, and then update it periodically?

ngn

@user true. i've programmed in jvm bytecode directly but it was long ago.

dzaima

@Adám after i load the truncated ones tomorrow

@rak1507 do you need the ids, text, or html?

rak1507

just text would do

I'm not even sure if there was a ... in the message but lets hope so

dzaima

@rak1507 here

rak1507

thank you :D

dzaima

01:12

updated to include usernames because that's probably useful

@dzaima (actually no it doesn't, those have …)

rak1507

brief skim over things and I've not found it, I feel like someone in an asylum trying to prove that they are sane

Adám

@rak1507 At least Nathan agrees with you.

rak1507

does he remember it too?!

Marshall

@rak1507 I really do believe you but I think it would be best if you stay here with us for a while.

rak1507

lol

Adám

01:20

4 hours ago, by nathan rogers

I've truly stepped into another timeline

rak1507

chat.stackexchange.com/transcript/message/56647450#56647450 YES! I AM SANE!

(I don't think that was the message I was looking for, but it's at least something)

Adám

@rak1507 That doesn't have anything alluding to how difficult or not it would be to migrate.

rak1507

yeah, that wasn't what I was looking for, but it is at least some evidence of wanting to migrate from APL

Adám

So?

I could have told you that (again).

rak1507

your message said 'vendor locked-in' people were not likely to complain, isn't migrating to a new language the biggest complaint there is?

Adám

01:30

Isn't considering migration but deciding to stay the biggest compliment?

rak1507

depends on their reasoning for not migrating

Adám

Right. So without knowing the details, there's really no conclusion to be made.

rak1507

true

ngn

@Adám like, people who tried to escape from prison but couldn't, must be loving prison so much

Adám

Afaik (this is one of the customers I've actually visited), they are quite happy with Dyalog APL as part of their stack.

@ngn No, not "couldn't", "didn't".

ngn

01:32

tried to, couldn't, so didn't

Adám

Who said they tried to?

They planned to, but they never made a real effort.

dzaima

@Adám I would assume they would be pretty far in their plans if you managed to get to know about it

ngn

@Adám "arranged leadership for the explicit purpose of facilitating migration of their own things away from APL" - i interpret that as "tried to"

Adám

That was a step that would have made it easier for them to make the move, but even though they successfully laid down that easier path, they still decided to stay.

dzaima

@Adám you think they'd set all that up just to "decide" to stay?

they surely would decide whether it's a good idea before setting up separate leadership for the project

Adám

01:36

@dzaima Yes. As I remember it, they did actually port some APL code (to C++, I think) but found that APL was valuable enough to keep for some of the other code. So not because they couldn't migrate APL code.

Elias Mårtenson

02:56

I need some advice, in case any of you are awake.

rak1507

as long as there is homework to procrastinate doing I will be here

Elias Mårtenson

OK, thanks. :-)

dzaima

am here too

Elias Mårtenson

Here's the problem: Since I have now implemented Dyalog-style trains, I need a way to assign them a name. Similar to Dyalog FN ← ×+÷

dzaima

@dzaima (decided to upgrade (well, rewrite) the orchard scraper now because why not)

Elias Mårtenson

02:59

Now, in Kap, functions are defined using: ∇ a fn b { a+b }

Or the short form: ∇ fn {⍺+⍵}

rak1507

how do you do inline functions? (or can you?)

Elias Mårtenson

Multiple arguments (actually destructuring lists) are implemneted as such: ∇ FN (a;b;c) { a+b+c }

dzaima

@rak1507 like regular Dyalog, assigning only is special

rak1507

ok

Elias Mårtenson

My problem: There is no natural way to include the Dyalog-style function assignment of trains in this syntax.

rak1507

03:01

∇ fn (train)

Elias Mårtenson

There are no syntaxtic reasons why FN ← + couldn't be allowed. The problem is that while it woul dbe Dyalog-compatible, it would be very confusing since function declaration is a compile-time operation, and variable assignment is done runtime.

So using the same syntax for both would imply that you can reassign a function at runtime, which is not the case.

rak1507

hmm, so there's no eval then I assume

Elias Mårtenson

So the question is: What syntax would be nice and concise for this type of assignment?

dzaima

@EliasMårtenson you could have assignment be statically parsed specially just like ∇-assignment

Elias Mårtenson

I've considered things like: FN ∇← +

dzaima

03:03

@EliasMårtenson probably what rak suggested (∇ fn (+⌿÷≢))

Elias Mårtenson

@dzaima Yes, indeed I can and that's one of the options I have thought of, but see my previous message.

@dzaima That would be problematic from a parsing perspective, because the leading part: ∇ fn ( is a valid function declaration

rak1507

what about ∇ fn +/÷≢

different from the lack of parens but maybe that's equally as annoying to parse

dzaima

@rak1507 no better than with parentheses (and actually worse)

rak1507

yeah :/

dzaima

@dzaima oh, a problem with that is that the train is actually a thing that's evaluated whereas with dfns, they're just stored

Elias Mårtenson

03:06

@rak1507 Also problematic since + is a function which is a regular symbol. Consider this: ∇ FN FOO {...} if FOO is a function, this would be a short-form assignment, if it's not, it would be the long-form.

I have considered ∇ FN ← +

rak1507

I don't like this very much, but what about ∇ fn ⍬ (train) because it takes no explicit arguments

dzaima

@dzaima (scraper finished reading the 162400 messages yay, 67MB uncompressed)

Elias Mårtenson

@rak1507 That would be possible, in which case the parens wouldn't actually be needed.

rak1507

I feel like that way is too verbose

Elias Mårtenson

Oh wait, they would

dzaima

03:09

@rak1507 at that point ∇← is better imo

rak1507

agreed

you say it's compile time, but what do you do about something like value ← ⎕ ⋄ f ← value+⊢ in dyalog

where the train depends on the value at the time and not whatever is associated with the 'value' name at the time of execution unlike dfns

Elias Mårtenson

@rak1507 You can't do it like that. The compiler must statically determine what is a function and what is a value. However, all is not lost. You can use lambda functions for that: value←XXX ◊ ⍞value + ⊢

Where XXX could be for example: λ+×÷

rak1507

wait, what is a lambda function

Elias Mårtenson

Basically the λ turns a function into a value, and ⍞ turns a value into a function

dzaima

@EliasMårtenson also consider: ∇ fn (-⍣(⎕←2))

rak1507

03:13

@EliasMårtenson oh, neat

why bother having compile time trains then

dzaima

@rak1507 using ⍞ on every train invocation is a bit lame

Elias Mårtenson

@rak1507 A few reasons. Terseness, Dyalog compatibility, performance. A train can be type-optimised without doing flow analysis like you have to on a regular function.

dzaima

@rak1507 i don't think the trains would be compile time

Elias Mårtenson

Also, when I put up the web-based kap demo page, the first thing that happened was that people started complaining that it didn't handle trains :-)

rak1507

is there a reason 0≡≢scalar btw

Elias Mårtenson

03:16

I personally don't use them (yet, perhaps I will start now)

rak1507

1 2 3+,1 errors at least, yay :D

Elias Mårtenson

@rak1507 Wait, shouldn't it?

rak1507

it doesn't in dyalog but that is a mistake

dzaima

@EliasMårtenson it doesn't because Historical Reasons in Dyalog and GNU and most other impls

Elias Mårtenson

@dzaima OK, I agree that's inconsistent. What are those historical reasons?

I'm guessing APL2 did too?

rak1507

03:19

if you really want to make ngn happy you should make ⍳,5 equal to ,¨⍳5 too

dzaima

@EliasMårtenson there wasn't any easy way to get from a 1-item vector to a scalar. So they, instead of adding a function for that, decided to make 1-item array behave like scalars

@rak1507 yeah that too exists for the same reason

Elias Mårtenson

@rak1507 That does sound logical. Now I have to check my code why Kap doesn't do that... I must have hadded a special check for it?

rak1507

why does ⍳2⍴2 give an error 'Error: Index out of bounds: Reading index 0 from scalar'

Elias Mårtenson

Ah. Yes. Now I see what I'm doing.

The fact that ⍳,5 works that way is a historical accident due to how I implemented the extension of it to support non-scalar arguments.

rak1507

⍳⍳5 causes the page to disappear!

Elias Mårtenson

03:23

There is the offender: github.com/lokedhs/array/blob/master/array/src/commonMain/…

@rak1507 Yes. It's a bug. The pretty-printer crashes on 5-dimensional values.

rak1507

ah ok

don't suppose you need many of them anyway

Elias Mårtenson

@rak1507 You should be able to see a very clear error in the JS console :-)

And true, this is the first time I've actually seen i thappened. I even forgot that this limitation existed.

rak1507

there's a fun error on ⍳⍬ too

dzaima

@rak1507 ⍴ seems quite broken - (2⍴2)⍴2 errors too

Elias Mårtenson

Argh... Hang on :-) I need to have time to implement your examples as test-cases :-)

rak1507

03:27

dyadic f⍣g doesn't seem to exist

(sorry for all the random nonsense I'm firing at you :D)

Elias Mårtenson

@rak1507 Correct. Not implemented yet.

Not at all nonsense.

rak1507

I'd say ⍳⍳5 falls squarely into the nonsense category

Elias Mårtenson

I need this. I'm just translating it to testcases for the test package so I'm sure to fix it.

@rak1507 Well it shouldn't crash :-)

rak1507

is ⊃(1 (2 3)) not padding intentional?

dzaima

@rak1507 (fwiw imo it's okay to error on that, though traditionally it doesn't; dzaima/APL errors too)

Elias Mårtenson

03:36

Yay. github.com/lokedhs/array/commit/…

Fixed it.

@rak1507 Yes. It's intentional.

However, the intent isn't based on anything else than my reading of the ISO spec.

I believe ISO says it should error.

Elias Mårtenson

03:50

OK, I changed the behaviour of ⍳,5 as an experiment (all I needed to do was remove some special-cased code and fix a related bug in a different place). The only test case that fails is the one that explicitly tests for this specific case.

Should I commit that change?

It's consistency vs. compatibility

dzaima

@EliasMårtenson I mean, you're one here to decide what KAP is for. I'd very much go with consistency

Elias Mårtenson

@dzaima Sounds good enough for me.

Marshall

@EliasMårtenson I think ISO only defines scalar, and not singleton, extension. From the APL2 documentation it looks like they treat one-element vectors as scalar but not higher-rank singleton arrays.

Elias Mårtenson

Done: github.com/lokedhs/array/commit/…

Note how the fix literally removes a special-case. I like that.

dzaima

announcing: dump of The APL Orchard

phantomics

03:56

Why is "1 2 3+,1" supposed to error?

dzaima

@phantomics because there's no reason for it to not error

the reason it should error is the same reason 1 2 3+1 2 should error

phantomics

Are singleton vectors not supposed to be treated like scalars for this purpose?

There is no clear way for 1 2 3+1 2 to resolve

dzaima

@phantomics that's the reason many APLs allow them, but singleton extension is regarded by many as bad

phantomics

I'd be interested to hear more about the rationale

dzaima

@phantomics it's just pointless when scalars exist. Not necessarily rationale, but a message from Dyalog's CTO

("when we only had ⍴" referring to times before ≢ existed, and to implement ≢ you had to do (0⍴0)⍴⍴x)

phantomics

04:04

Thanks, I'll consider making it a bug in April

Marshall

@phantomics The reason you would need singleton extension is that you have axes in your data that don't correspond to anything you might use as an index. Singleton extension makes these disappear but it would be much better, and less bug-prone, to write your code in a way that doesn't introduce these superfluous axes.

phantomics

I suppose you could have singletons whose number of axes is meant to convey an additional piece of data

dzaima

@dzaima (maybe my "by many" claim isn't very founded. regardless, i believe it's bad, and others do too)

Marshall

@phantomics Yes, but it doesn't seem like you would want the interpreter automatically throwing that data out for you.

phantomics

Right

Should scalar ops with axes be able to scale singletons? I.e. (,1)+[1]⍪⍳3

I'm aware that the scalar + axis syntax is not well regarded

Marshall

04:13

@phantomics If ⎕IO is 1 (so, pairing a length-1 axis with a length-3 one if I got the shapes right), then I'd say that shouldn't be allowed for the same reason.

phantomics

It's pairing a length-1 axis with another one

Elias Mårtenson

@phantomics It's a bug in kap :-)

phantomics

since ⍪⍳3 makes a vertical table

Marshall

@phantomics + with axis should use the axis as indices into the axes of the higher-rank argument to get axes to pair with the other argument, so you get a shape of 3 1[1] or ,3. Am I making an error there?

phantomics

@Marshall my mistake, that should have been [2]

Marshall

04:20

Regardless, the rule should be that pairing no axes with some axes is always fine, and pairing axes of different lengths is never fine.

@phantomics Okay, then it's two length-1 axes and should definitely work (it would work if they had length 2!).

phantomics

@Marshall That's what I thought, if you're not specifying it like that then I can see how arbitrarily extending a singleton is unacceptably vague behavior

Also I figured out that algo from earlier, now [↑ mix] is fully parallelized

Elias Mårtenson

04:41

@phantomics What do you mean by that?

phantomics

05:03

@EliasMårtenson I was asking about how to implement sorting the dimensional bases used to transcode the row-major index within the mix function

nathan rogers

05:22

('e'∨.=⊃) this gets a cell of values, picks the item in the cell and asks are there any = t the letter e

cell input looks like this

┌─────────┬─────────┐
│e 461 550│h 564 900│
└─────────┴─────────┘

how can I generalize this to accept a right argument for the letter 'e' or any other letter?

rak1507

a right argument?

nathan rogers

er, left

'e' should be left argument

I tried just ⊣ which I was certain was the answer

but it isn't

I think because {⍺⊃⍵} on the right

rak1507

∨.=∘⊃ is probably simplest

nathan rogers

⊣∨.=⍥⊃

oh that works?

oh, that works :P

rak1507

with ⊣ you would have to do something like ⊣∨.=⊃⍤⊢

nathan rogers

05:27

thanks

rak1507

np

Russtopia

06:20

@Marshall Hi, GNU APL dabbler here, I can confirm it doesn't support control structures, only classical branch

rak1507

06:33

for this codegolf.meta.stackexchange.com/questions/5243/… the SEDE query also includes questions in 2020, I assume they are not valid for the bounty though?

also how come ×⍨⍣¯1 ⊢ 4 doesn't work in dyalog extended

dzaima

12:47

@rak1507 because × and ⍨ are overridden, and ⍣¯1 does really funky things to get ×⍨⍣¯1 to work normally; `×`⍨⍣¯1 ⊢ 4 works

Adám

@dzaima ''⍴ worked but was ugly and not intuitive.

rak1507

13:33

@dzaima ah, thanks

rak1507

14:22

@Adám can you confirm whether or not challenges posted during 2020 are eligible for your bounty? The SEDE query data.stackexchange.com/codegolf/query/1156913/… uses <= 2020, should this be < 2020?

Adám

14:54

@rak1507 Now updated to be clear that it includes 2020.

rak1507

oh, it does? cool!

Adám

15:08

@dzaima Any reason GH pages is not enabled?

dzaima

@Adám the whole 60MB download on every page reload thing

(it'd be less compressed, but still)

Adám

@dzaima Sure. Or you could warn, then click-to-load.

dzaima

@Adám nah. Downloading isn't that hard

nathan rogers

Any way to get the utterly first item of a nested vector?

dzaima

@nathanrogers ⊃⍣≡

Adám

15:11

@nathanrogers ⊃∊ is shorter but much less efficient.

nathan rogers

for the case where the utterly first item is ⊃⊃⊃ does ⊃⍣≡ make any difference?

dzaima

@nathanrogers ⊃⊃⊃ will probably be a couple nanoseconds faster

nathan rogers

hha, k

dzaima

(of course assuming that "utterly first item" is a simple scalar)

nathan rogers

yes

nathan rogers

15:31

I have a question about variable encoding

variable meaning the encoding only marginally differs from another solution

eval ← {                             ⍝ ...
    ~×≢⍵ : 1⊥,⍺                      ⍝ ... Not any? Then sum reduce
    i←ind ⊃⍵                         ⍝ ... Current update indices
    o←¯1+'f n'⍳⊃⊃⊃⍵                  ⍝ ... Is on?
    'e'=⊃⊃⊃⍵ : ({2+⍵}@i⊢⍺) ∇ 1↓⍵     ⍝ ... Add 2
    (0⌈({o+⍵}@i⊢⍺)) ∇ 1↓⍵            ⍝ ... Add 1 or ¯1, with 0 floor
}
eval ← {                             ⍝ ...
    ~×≢⍵ : 1⊥,⍺                      ⍝ ... Not any? Then sum reduce
    i←ind ⊃⍵                         ⍝ ... Current update indices

These 2 are mostly the same besides the value of o and the left arguments to @

Adám

@nathanrogers Sounds like you need an operator :-)

nathan rogers

How would you go about expressing this in such a way such that it is clear those differences

Adám

Create a single operator that takes two operands to specify o and the @ part, then define the two evals in terms of that operator.

nathan rogers

What I'm drawn to is some kind of macro, but obv that isn't available. So perhaps a text representation with arguments with an ⍎

dzaima

@Adám it seems here are 3 variable parts though

Adám

15:35

@dzaima Oh, right. Well spotted.

nathan rogers

@Adám but so far as I can tell there's 4 parts. o← 0⌈ o+ and 2+

I guess I could merge the o← 0⌈ and o+

that should work

rak1507

using ~×≢ and 1⊥?!

Adám

@dzaima Ah, but each o is only used in combination with one particular version.

@rak1507 I'd write 0=≢ and +/ for clarity.

rak1507

I thought he would have too

nathan rogers

I guess 0⌈ can be applied universally

because when toggling and seeting to 1 or 0 that'll never change

ok no here's the problem

({o+⍵}@i⊢⍺)) uses the argument from @

(o@i⊢⍺)

depends on the argument from the outer scope

on the second one, I can simply (⍺⍺ ⍵) to get the ⍵ used to defin o in (o@i⊢⍺)

but for the first one, how do I return the function o + to apply to ⍵ of the @ function?

s06_2 ← {({(¯1+'f n'⍳⊃⊃⊃⍵)}∘+s06_1a 2+) ⍵}

I think that works?

but I'm getting wrong answers

Adám

15:49

@nathanrogers Maybe I can have a look later, but I have to play with my children now.

nathan rogers

I think I got it

well I'm close anyway. I named the out n←⍵ and then call (n ⍺⍺ ⍵) so the other functions don't break but that's not quite it, and I'm getting errors o the other

gotta run I'll check in after

nathan rogers

19:19

DOMAIN ERROR
s06_1a[5] 'e'=⊃⊃⊃⍵:((⍵⍵@i)⍺)∇ 1↓⍵        ⍝ ... Toggle? Apply ~ to i in ⍺
                    ∧
               ⍵⍵
{~⍵}

yeah I'm not sure why this is a domain error

Adám

@nathanrogers What's up with that ⍵⍵ and {~⍵}?

nathan rogers

I need to pass 2 operand functions

⍺⍺ and ⍵⍵

Adám

Sure, but they are not part of the error message, are they?

nathan rogers

⍵⍵ is the operand function {~⍵} for one case, and the other case ⍵⍵ is {2+⍵}

Adám

What is i at the time of the error?

nathan rogers

19:26

well it was working the previous refactor step where I replaced the dfns with the operand reference

i and ⍺ look right

⍵⍵ was the only thing I changed

xpqz

Is there a way to pick a field out of an array of namespaces without doing {⍵.myfield}¨arr?

Adám

@xpqz arr.myfield

xpqz

That sounds too good to be true.

nathan rogers

no you're right, looks like I did a goof

Adám

@xpqz It is always fun to watch APL do C# better than C#.

xpqz

19:31

I was thinking that's exactly what I wanted, but didn't bother trying, because "surely there's no way"

nathan rogers

s06_1←{({'n'=⊃⊃⊃⍵}s06_1a{~⍵})⍵}

is there a problem with that operator?

Adám

@xpqz another fun one is obj1.(obj2.something obj3.else) etc. and all those objs can be arrays of objs.

@nathanrogers Depend what you want it to do, of course. Parens are not needed here, btw.

nathan rogers

(1000 1000⍴0)(⍺⍺ eval⍵⍵) ⍵

Adám

In fact, unless s06_1 will be called dyadically, s06_1←{'n'=⊃⊃⊃⍵}s06_1a{~⍵} should be enough

nathan rogers

being passed this way, but ⍺ is undefined

getting VALUE ERROR

Adám

19:34

@nathanrogers You're giving us way too little context to be able to help you.

How about posting all the relevant code together?

nathan rogers

mok

s06_1a←{                                  ⍝ 6_1 Recursive Solution
    ind  ← {,{(⊃⍵)∘+¨⍳1+⊃|-/⍵}1↓¨⍵}      ⍝ ... The indices to set
    eval ← {                             ⍝ ...
        ~×≢⍵ : 1⊥,⍺                      ⍝ ... Not any? Then sum reduce
        i←ind (n←⊃⍵)                       ⍝ ... Current update indices)
        'e'=⊃⊃⊃⍵ : (⍵⍵@i ⊢ ⍺) ∇ 1↓⍵        ⍝ ... Toggle? Apply ~ to i in ⍺
        (0⌈({⍵ ⍺⍺ n}@i⊢⍺)) ∇ 1↓⍵               ⍝ ⍝ ... 0 or 1 at indices in ⍺
    }
    lights ← 1000 1000⍴0

ignore comments

they haven't been updated because I can't get the operands workign

oh

⊢

nah ⍺ VALUE ERROR still

Adám

OK, now we just need some quiet time to look at your code.

nathan rogers

actually the comments still match

because this is the first case. second case is this
s06_2 ← {⍺+¯1+'f n'⍳⊃⊃⊃⍵} s06_1a {2+⍵}

,⊂('e' 0 0)('h' 100 100)

data is in the form of this

with any number of pairs

Adám

@nathanrogers I get:

      s06_1 data
10201
      s06_2 data
20402

nathan rogers

no value error?

Adám

19:48

Nope.

nathan rogers

then perhaps its happening further down in the recursive call

sec

let me see if I can define an input that triggers it

Adám

What is your ⎕IO⎕ML?

nathan rogers

⎕IO←0

⎕ML←1

Adám

I was using ⎕IO←1

Nah, same results with ⎕IO←0

nathan rogers

rgr, sec

s06_1 (('f' 0 0)('h' 100 100))(('e' 0 0) ('h' 100 100))

that should do it

f for off

'n' = ⊃⊃⊃⍵ should return 0

because its not 'n' for on its 'f' for off

Adám

19:51

OK, now I get a syntax error on s06_1a[6]

nathan rogers

and I'm not trying to be terse, ust the first usable regex I could match :P

0⌈ shouldn't have any effect on this case

because its 0 1 or ~

rak1507

what problem are you doing?

Adám

@nathanrogers ({⍵ ⍺⍺ n}@i⊢⍺) cannot be right as that would give a monadic operator as left operand of @

nathan rogers

(0⌈({⍵ ⍺⍺ n}@i)⍺)∇ 1↓⍵

well that gives me a syntax error

Adám

Yes, because an operator cannot take an operator as operand.

Maybe you meant (⍺⍺{⍵ ⍺⍺ n}@i⊢⍺) instead?

nathan rogers

19:54

oh, right

⍺⍺ needs to be named

oops

Adám

Or just pass it in.

dzaima

@Adám (or ⍺⍺∘n@i⊢⍺)

Adám

@dzaima Ninja'd. Yes.

nathan rogers

alright that did it

Adám

Cool.

nathan rogers

19:57

thanks for that. I already went round the bend with it for too long :c

oh, here's a question

does ⍺⍺ and ⍵⍵ need to be recursively passed into ∇?

Adám

Btw, you don't need to assign n←⍵ now.

nathan rogers

oh, right

that's cool

rak1507

what is the full problem?

nathan rogers

adventofcode.com/2015/day/6#part2

you need to complete part 1 to get part 2

Adám

@nathanrogers Not entirely sure what you're asking but in a dop, ∇ is equivalent to (⍺⍺∇∇⍵⍵) or (⍺⍺∇∇)

nathan rogers

19:59

I mean, I'm calling (⍺⍺ eval ⍵⍵) and eval is recursive

do I need to pass (⍺⍺∇⍵⍵) or no

Transcript for