The APL Orchard

nathan rogers

20:00

I guess its self evident

because its working now

Adám

@nathanrogers ∇ is (⍺⍺∇∇⍵⍵) so if you want to invoke the same operator with the same operands, ∇ is enough. Otherwise use ∇∇ if you want to use the same operator with different operands.

nathan rogers

I see

s06_1a←{                                 ⍝ 6_?
    onoff ← ⍺⍺                           ⍝ ... 6_1: 0/1 ⋄ 6_2 1/¯1
    tggle ← ⍵⍵                           ⍝ ... 6_1: ~⍵  ⋄ 6_2 2+⍵
    ind  ← {,{(⊃⍵)∘+¨⍳1+⊃|-/⍵}1↓¨⍵}      ⍝ ... The indices to set
    eval ← {                             ⍝ ... Recursive Solution
        ~×≢⍵ : 1⊥,⍺                      ⍝ ... Not any? Then sum reduce
        i←ind (n←⊃⍵)                     ⍝ ... Current update indices)
        'e'=⊃⊃⊃⍵ : (tggle@i⊢⍺) ∇ 1↓⍵     ⍝ ... Toggle? Apply ~ to i in ⍺

There's the final solution for each one that passes tests

the ind function to a while to intuit because the size was super big. even on smaller tables output looked right, but it wasn't until I did ...1+... that it produced the right answer

Why can I only edit for like 30 seconds after sending a message?

Adám

@nathanrogers I would personally find this easier to read if you used a naming convention to differentiate arrays/functions/mops/dops

nathan rogers

mops/dops?

Adám

@nathanrogers So that I cannot agree with your statement, only to have you change it and make me look like I agreed with something very different. (Also, it isn't 30 s.)

@nathanrogers monadic operators, dyadic operators.

@nathanrogers My personal style guide on naming conventions.

nathan rogers

20:12

I tend to be more slapdash until I know what things are, and then try to refactor into names that work

Adám

@nathanrogers Though you didn't name the operand in this case, you could have, and that might have helped you spot that you were giving an operator to @

nathan rogers

Right, that wasn't something I thought about until now

Adám

Now you know for next time.

nathan rogers

Do guards always return?

dzaima

@nathanrogers yep, at least in Dyalog

nathan rogers

20:21

even with assignment

dzaima

yep

Adám

@nathanrogers Yes, then the result is shy.

nathan rogers

how can I map one of 2 functions to a given name

depending on conditions

I always want to call some form of an expression, but the f may be one or the other

dzaima

@nathanrogers {cond:f⍵⋄g⍵}

nathan rogers

no, I don't want to evaluate

I want the function

{cond : f ⋄ g}

I know that isn't the way to do it

but how is the way

rak1507

20:23

     solve←{
         m←0⍴⍨2⍴1000
         on←⍎⊃⍵ ⋄ off←⍎1⊃⍵ ⋄ toggle←⍎2⊃⍵
         +/,m⊣{0⊣m∘←0⌈(⍎3⊃a)@((⊂⊃⌽b)+,⍳1+⊃-/b←⍎¨0 2⊃¨⊂a←⌽' '(≠⊆⊢)⍵)⊢m}¨⊃⎕NGET'201506.txt' 1
     }
solve '10~'
solve '+∘1' '-∘1' '+∘2'

not very fast... probably a better way

Adám

@nathanrogers Where do the two functions come from?

nathan rogers

they're the operands from ealier

Adám

@nathanrogers How about ⍺⍺{⍵:⍺⍺ ⍵ ⋄ ⍵⍵ ⍵}⍵⍵∘cond ?

nathan rogers

(0⌈f@(inds⊃⍵)⊢⍺) ∇ 1↓⍵

this is the expresion I want

its the same in both cases

f might be ⍺⍺∘⍵ or it might be ⍵⍵

in this case onoff or toggle

f←{cond: onoff∘⍵ ⋄ toggle} doesn't work

dzaima

(BQN, cond⊏f‿g :) )

nathan rogers

20:26

I need to select the function that should go there

Adám

@nathanrogers Right, because you need to pass the argument to the function.

nathan rogers

so you're saying there is not a way to optionally bind the name of f to one of a few functions

rak1507

⍎cond⊃'fg'

nathan rogers

are there any guards that DONT exit

Adám

(0⌈⍺⍺∘⍵{⍵:⍺⍺ ⍺ ⋄ ⍵⍵ ⍺}⍵⍵∘cond@(inds⊃⍵)⊢⍺) ∇ 1↓⍵ should work.

rak1507

20:28

embrace the evals

nathan rogers

I get that it should work, but it isn't very elegant, so I'm trying to rephrase the question ot be specific

Adám

@nathanrogers No, then you have to wrap in another layer of dfn. If you really need this, you seem to be doing structured imperative programming, so use a tradfn.

nathan rogers

it seems like I'm doing functionally recursive programming

dzaima

@Adám wanting to pick one of two functions is a pretty normal thing to do imo

Adám

Anyway, what I wrote should work.

rak1507

20:29

@Adám he's trying to do functional programming not imperative

Adám

@dzaima Sure, but that wasn't what I answered. I talked about diverging branches coming back together again.

rak1507

personally I'd give up any principles and use ⍎ but that's not for everyone

dzaima

@Adám functional programming doesn't not allow that

Adám

@nathanrogers Does my expression work?

@dzaima OK.

nathan rogers

yes it works

but I'll go with @rak1507 suggestion

eval ← {                             ⍝ ... Recursive Solution
    f←⍎'onoff∘⍵' 'tggle'⊃⍨'e'=⊃⊃⊃⍵   ⍝ ... Select function
    ~×≢⍵     : 1⊥,⍺                  ⍝ ... Not any? Then sum reduce
    (0⌈f@(inds⊃⍵)⊢⍺) ∇ 1↓⍵           ⍝ ... Set onoff at i in a, with 0 floor
}

honestly I'd take an array of functions over trains

Adám

20:32

If you went with mine, I'd recommend breaking up that long lin.

@nathanrogers Not sure what that means.

rak1507

first class functions would be great but it'll never happen :(

dzaima

(also fwiw, {cond: f ⋄ g} works in dzaima/APL :) )

rak1507

wouldn't really make sense syntactically

Adám

@dzaima As did it in a conference edition of Dyalog APL.

nathan rogers

@Adám I wnat to have a list of function onoff tggle and select which one

onoff tggle[cond]

for instance

Adám

20:33

@nathanrogers Sure, I get that, but I'm not sure what the connection is to trains.

nathan rogers

the syntax would be overloaded with train syntax?

you have to pick one with strand notation

dzaima

@nathanrogers well, you still couldn't create an array of both functions and arrays

nathan rogers

or at least some kind of "symbol" notation

@dzaima right you'd need some kind of lisp-style or k-style symbol reference notation

rak1507

in k functions are first class anyway

dzaima

or just a non-strand array notation (like BQN)

Adám

20:36

Right. BQN is good.

nathan rogers

@rak1507 yeah I like your input parsing method better

I used a regex, but yours with just splitting on space works out just fine

rak1507

yea regex seems slightly overkill but for something more complex splitting probably wouldn't work

ironic because in AOC last year I probably overused regex if anything haha

nathan rogers

i06 ← ('.? ?[0-9]+,[0-9]+'⎕S{(⊃,{⍎','⎕R' '⊢1↓⍵})⍵.Match~' '})¨2015 #.day 6

prit-tee overkill

Adám

@nathanrogers Isn't ⍎','⎕R' '⊢1↓⍵ the same as ⍎1↓⍵?

nathan rogers

idk, is it?

let me check

I thought i needed to get rid of ,'s

Adám

20:49

@nathanrogers Because…?

nathan rogers

yeah seems to work

because ,'s are numbers

and I'm not super familiar with ⍎ spec

Adám

@nathanrogers ⍎ ≈ APL

nathan rogers

but yeah, that works

('.? ?[0-9]+,[0-9]+'⎕S{(⊃,{⍎1↓⍵})⍵.Match~' '})¨2015 #.day 6

Adám

@nathanrogers {⍎1↓⍵} is ⍎1↓ no?

nathan rogers

sorr that's a train

Adám

20:51

Oh, right.

So (⊃,{⍎1↓⍵}) is 1(⊃,∘⍎↓) with ⎕IO←1

Or (⊃,∘⍎1↓⊢) with any ⎕IO

@nathanrogers Any reason you use [0-9] instead of \d?

dzaima

(dzaima/APL, ⍎⍢(1↓) because of course i'll choose to mention under)

nathan rogers

not specifically

Adám

@dzaima Oh, I so wish for structural under…

dzaima

@Adám (though this isn't purely structural under since ⍎ can change the item amount)

oh :/

Adám

I was just about to ask "Won't under fail then?"

dzaima

20:57

@Adám it fails only on a single item. 'a' 1 2 34≡(⍎⍢(1↓))'a1 2 34' works fine

the problem is that ⍎,'1' creates a scalar, whereas ⍢ wants back a vector

Adám

@dzaima So ,⍤⍎ works?

Shouldn't a scalar distribute?

dzaima

@Adám yep

@Adám if ⍎⍢(1↓)'a1' was 'a' 1, it'd break the single under rule, i.e. 1↓'a' 1 isn't the result of ⍎

nathan rogers

21:13

man this is getting obnoxious

@Adám I've got some kind of weird stateful problem. When I clear the workspace and link in the directory again, the function doesn't work... until I modify it, then save it and run it twice, THEN it works

any idea where to begin with troubleshooting?

Adám

@nathanrogers Trace through runs until right before it works, then look at what changes when it starts working.

nathan rogers

ok no, just modifying the file does it

um... do assigning operators in a namespace file count as "evaluating code" in a namespace...

myop ← f dfn g

@Adám

Adám

@nathanrogers That doesn't look right. What are f and dfn and g? (Again, a naming convention would help.)

nathan rogers

that doesn't matter I'm asking if that's legal to do in a namespace file

or is that tantamount of "evaluating code" in a namespace

dzaima

@nathanrogers is dfn a function? an operator?

Adám

21:28

I cannot parse the statement without context info. APL is not statically parseable.

@nathanrogers I assumed myop is an operator, is this correct?

nathan rogers

ok fine

myfn ← f op g

harga garga

is that legal in a namespace file

Adám

Oh, now that makes more sense.

nathan rogers

or not

Adám

I think it is technically legal, but I wouldn't do it, because it depends on the order in which things are declared.

rak1507

⋄ ∊'your code ',('is' 'is not'[?2]),' valid'

nathan rogers

21:30

look I'm completely stuck

Adám

Nominally, namespace scripts are considered ordered, but when you edit a single function, the system can "patch" that back into the namespace without evaluating everything. I think.

nathan rogers

I do a fresh link
run code
unexpected errors
make any arbitrary edit to the namespace file
run twice
it works

@Adám the point to clarify is I'm not making an edit to a single function, but I can make any arbitrary edit to anywhere in the file

Adám

@nathanrogers Please cut it down to a minimal repro and log an issue against Link.

nathan rogers

run the same expression twice and it works

i minimal repro involves an entire library

I'll just create a repository for the whole thing I guess

dzaima

@nathanrogers something that should've been asked way earlier: what exactly are your "unexpected errors"?

Adám

21:32

@nathanrogers That sounds unlikely.

nathan rogers

LENGTH ERROR
run[32] ∨/~passed:((⊂⍵),error passMSG passed),(⊂⍉,/↑failMSG¨failed) ⍝ any failed

Adám

There's missing one line from that message.

Could if be that you mutate some state when you run, and only after two mutations are the arrays conformant?

nathan rogers

      )clear
clear ws
      ]link.create # 'c:\d\aoc'
Linked: # ←→ c:\d\aoc
      aoc2015 t.test's06_1'
LENGTH ERROR
run[32] ∨/~passed:((⊂⍵),error passMSG passed),(⊂⍉,/↑failMSG¨failed) ⍝ any failed
                                                 ∧

Adám

@nathanrogers Still missing a critical last line of that error, no?

nathan rogers

no

Adám

21:35

@nathanrogers no yes. TFTFY.

dzaima

@nathanrogers did you do printf-debugging (i.e. add a ⎕← after ,/ to see what input it doesn't like?)

nathan rogers

yes

Adám

OK, again, we're missing the context.

nathan rogers

I've taken a look at all of the inputs

Adám

So on runs where it works, and on runs where it doesn't work, are the inputs the same?

dzaima

21:36

@nathanrogers and did you understand why ,/ was failing?

nathan rogers

and to make it work, all I need to do is add a new line anywhere in the namespace file I've been working in

Adám

@nathanrogers OK, and what change do you see to the intermediary values when you do that?

dzaima

1) replace ,/ with {⎕←⎕SE.Dyalog.Utils.repObj⍵⋄,/⍵}
2) reload the namespace
3) run the function twice, see if the outputs differed

nathan rogers

so values is meant to be the output of a function, and somehow its collecting something that isn't the output of a function, its collected the internal data of the function

Adám

@nathanrogers "values"?

nathan rogers

21:39

values ← (⍺ trap ⍵)¨inputs ⍝ collect results
passed ← expected≡¨values ⍝ 1 for pass 0 for fail
error ← (⊂,⊂⎕THIS.config.error)⍷values ⍝ which tests crashed ?
∘∘∘
failed ← (~passed)/↓inputs,expected,⍪values ⍝ select failing cases

these are the lines that preceed the failing one

Adám

Can you please just follow dzaima's 1-2-3 guide?

nathan rogers

the first 2 times, values is a giant matrix, which is the ⍺ in the function causing the failure

yes, I did that

and that isn't the issue

the isue is that values is something entirely unexpected

s06a←{                                   ⍝ 6_? ⍵ ←→ The range to modify
    ~×≢⍵  : 1⊥,⍺                         ⍝ ... Not any? Then sum reduce
    onoff ← ⍺⍺ ⋄ toggle ← ⍵⍵             ⍝ ... Functions for each case
    range ← {,{(⊃⍵)∘+¨⍳1+⊃|-/⍵}1↓¨⍵}     ⍝ ... The range of inds to modify
    ⍺     ← 1000 1000⍴0                  ⍝ ... A light matrix
    f     ← ⍎'onoff∘⍵' 'toggle'⊃⍨'e'=⊃⊃⊃⍵⍝ ... Which action for this range?
    (0⌈f@(range⊃⍵)⊢⍺) ∇ 1↓⍵              ⍝ ... Set value at indices, 0 floor

Adám

OK. So go put ⎕← or that special dfn dzaima wrote here and there until you identify the exact point where two consecutive runs differ. Then report back to us.

nathan rogers

listen

dzaima

so work backwards and see where something differs

nathan rogers

21:41

this is the function I was working on

Adám

Good for you. Now can you follow the sound advise that we're giving you?

nathan rogers

let me finish

Adám

Why? We answered your question:

26 mins ago, by nathan rogers

any idea where to begin with troubleshooting?

nathan rogers

1. link the workspace
2. run the test on this functiON
3. fail, values ←→ ⍺ in this dfn
4. edit the namespace containing this dfn anywhere not in the dfn
5. run it again
6. 1st run :: values ←→ ⍺ in this dfn
7. 2nd run :: values ←→ the expected result

yes, and I'm trying to answer

this is the thing that is wrong

values, used to formulate the failing line

is the wrong "value"

that is the thing that is wrong, how does editing a namespace file change that?

i mean, arbitrarily, just adding blank lines anywhere in the file fixes it

Adám

We're missing the context (again!)

Where does the line values ← (⍺ trap ⍵)¨inputs appear?

nathan rogers

21:45

right, because there's an entire testing framework that lays behind this

dzaima

@nathanrogers I assume that code is within some other function that we know nothing about?

nathan rogers

inside a testing tool

like I said

minimum repro involves a testing library

Adám

We gave you a universal method for figuring this out on your own.

If you want us to figure it out for you, you've got to give us the whole story.

nathan rogers

yes and I answered the question

Adám

Did you remember the answers, though?

nathan rogers

21:46

yes

dzaima

@nathanrogers what is "this dfn" there? the test function, s06a or something else?

nathan rogers

first 2 runs a 1000 x 1000 matrix

last run an integer

Adám

@nathanrogers Great, then work backwards from there.

nathan rogers

what's backwards? the inputs are identical

the function is identical

the output is different

Adám

Sure, but clearly, at some point, the two runs diverge.

nathan rogers

21:48

yea, and that newline sure has a lot to do with how its being evaluated

Adám

We want to know the point where the same code gives two different results with the same input.

dzaima

@nathanrogers you've found that values changes. Have you looked at the reason values changes?

Adám

So, your homework is: Identify the earliest expression that can give two different results depending on how many times it is run or depending on if you've edited in a trivial change.

nathan rogers

right, I have looked and the inputs and expected outputs look exactly the same

this is the earliest

that's what I'm saying

this line is the result of calling the function on the same inputs

Adám

btw, when you made the trivial change, did you try altering a comment rather than inserting a line?

dzaima

21:50

how about a ⎕← for values both after it is just defined, and in the place it's used?

nathan rogers

yes that works

modifying a comment

Adám

@nathanrogers Works, as in triggers the change?

nathan rogers

yes

(,⊂(('e' 461 550) ('h' 564 900)) (('f' 370 39) ('h' 425 839)) (('f' 464 858) ('h' 833 915)) (('f' 812 389) ('h' 865 874)) (('n' 599 989) ('h' 806 993)) (('n' 376 415) ('h' 768 548)) (('n' 606 361) ('h' 892 600)) (('f' 448 208) ('h' 645 684)) (('e' 50 472) ('h' 452 788)) (('e' 205 417) ('h' 703 826)) (('e' 533 331) ('h' 906 873)) (('e' 857 493) ('h' 989 970)) (('f' 631 950) ('h' 894 975)) (('f' 387 19) ('h' 720 700)) (('f' 511 843) ('h' 581 945)) (('e' 514 557) ('h' 662 883)) (('f' 269 809) ('h' 876 847)) (('f' 149 517) ('h' 716 777)) (('f' 994 939) ('h' 998 988)) (('e' 467 662) ('h' 555 957…

look at the last result

the final elemnt

(,⊂(('e' 461 550) ('h' 564 900)) (('f' 370 39) ('h' 425 839)) (('f' 464 858) ('h' 833 915)) (('f' 812 389) ('h' 865 874)) (('n' 599 989) ('h' 806 993)) (('n' 376 415) ('h' 768 548)) (('n' 606 361) ('h' 892 600)) (('f' 448 208) ('h' 645 684)) (('e' 50 472) ('h' 452 788)) (('e' 205 417) ('h' 703 826)) (('e' 533 331) ('h' 906 873)) (('e' 857 493) ('h' 989 970)) (('f' 631 950) ('h' 894 975)) (('f' 387 19) ('h' 720 700)) (('f' 511 843) ('h' 581 945)) (('e' 514 557) ('h' 662 883)) (('f' 269 809) ('h' 876 847)) (('f' 149 517) ('h' 716 777)) (('f' 994 939) ('h' 998 988)) (('e' 467 662) ('h' 555 957…

contrast that with the last element here

which is an integer

not a 1000x1000 matrix

same input

Adám

Well, a 1-element vector, but OK.

At which primitive do runs diverge?

nathan rogers

values ← (⍺ trap ⍵)¨inputs

 r←(namespace trap fn)args
 :Trap 0
    ⍝ Attempt to run fn inside namespace with args
     r←(namespace⍎fn) args
 :Else
    ⍝ Report if error occurs
     r←,⊂⎕THIS.config.error
 :EndTrap

Adám

21:56

@nathanrogers Last I checked, that's not a primitive.

nathan rogers

okee

Adám

At which primitive do runs diverge?

nathan rogers

r←namespace⍎fn args then

Adám

fn is a character vector?

nathan rogers

yes

dzaima

21:57

@Adám i assume nathan verified that ⍺ ⍵ trap inputs there are all the same, but it ends up differing. the trap definition given is the faulty thing here

nathan rogers

's06_1'

Adám

@nathanrogers Hold on, that doesn't appear in your code!

nathan rogers

:C I don't follow

Adám

@nathanrogers What don't you follow? What I said or what dzaima said?

dzaima

@nathanrogers (you mean with the parentheses, right)

nathan rogers

21:59

oh yes

sorry

r←(namespace⍎fn) args

Adám

That's not a primitive!

dzaima

@Adám ⍎ is definitely a primitive

Adám

@dzaima Sure, but NR didn't say that.

nathan rogers

you're asking where inside the evaluation of fn it differs?

Adám

@nathanrogers So, if you split the line into two:

temp←namespace⍎fn
r←temp args

Does temp differ between runs?

nathan rogers

22:00

ok

Adám

ok? It was a yes/no question.

dzaima

@Adám (just r←(⎕←namespace⍎fn) args would also work)

Adám

@dzaima There are cases where that ⎕← might mess up things.

nathan rogers

i mean ok I'm gonna go do it

dzaima

@Adám how so?

Adám

22:03

@nathanrogers Great. Take your time. I'll check in tomorrow morning.

@dzaima A bug in the interpreter.

dzaima

@Adám yaaaaay

nathan rogers

oh hey

something new happened

I put ∘∘∘ after temp and r in this example, continued execution first time it exited correctly

the run errored

@Adám WHOAH

THIS is really wierd

      aoc2015 t.test's06_1'
 T IS:  ,  #.aoc2015 .  s06_1←{
                            lights←1000 1000⍴0
                            eval←{
                                ~×≢⍵:⍺
                                'e'=⊃⊃⊃⍵:~⍺
                                onoff←'o'=⊃⊃⊃⍵
                                a←onoff@,({(⊃⍵)∘+¨⍳⊃|-/⍵}1↓¨⊃⍵)⊢⍺
                                ⍺ ∇ 1↓⍵
                            }
                            lights eval ⍵
                        }
LENGTH ERROR
run[34] ∨/~passed:((⊂⍵),error passMSG passed),(⊂⍉,/↑failMSG¨failed) ⍝ any failed

I don't have this version defined ANYWHERE

this is an OLD version

before I got part 2 of the problem

even after closing and reopening Dyalog

dzaima

maybe you have it in another file?

nathan rogers

I don't have the word "lights" anywhere in my source code, nor "eval"

nope

I SEE

its been defined inside a directory for some reason

it created a directory for my namespace with this dfn in it

dzaima

yay

nathan rogers

22:13

I didn't define that

manually or automatically. I probably opened the )ed on this dfn at some point

yeah ok, problem solved

dzaima

yay

nathan rogers

thanks a ton @Adám for walking through... I never would have thought to print the namespace⍎fn, I would have expected it to give me back some kind of reference to a namespace, not the text of the function

nathan rogers

22:36

@rak1507 I really like that

rak1507

thanks

nathan rogers

and how you made the arguemtns uniform

in both cases, that's slick

s06a←{                                   ⍝ 6_A Generalize both solutions
    ~×≢⍵  : 1⊥,⍺                         ⍝ ... Not any? Then sum reduce
    onoff ← ⍺⍺ ⋄ togl ← ⍵⍵               ⍝ ... Name functions for each case
    range ← {,{(⊃⍵)∘+¨⍳1+⊃|-/⍵}1↓¨⍵}     ⍝ ... The range of inds to modify
    ⍺     ← 1000 1000⍴0                  ⍝ ... A light matrix default value
    f     ← ⍎'onoff∘⍵' 'togl'⊃⍨'e'=⊃⊃⊃⍵  ⍝ ... Which action for this range?
    (0⌈f@(range⊃⍵)⊢⍺) ∇ 1↓⍵              ⍝ ... Set value at indices, 0 floor

That's what I'm calling my final version

I tried to copy your +∘2 for part 2

but it causes errors without ()

and (+∘2) is the same length as {2+⍵} so, meh

rak1507

yep

I would guess yours is faster than mine but I haven't tested

nathan rogers

idk, I can test yours quick

rak1507

just from not having to store a bunch of stuff

I reckon there's probably a cleverer way to do it though

nathan rogers

22:39

nah

rak1507

like, if you know a certain region is on, it's obviously easy to calculate the area, so maybe it's still doable to calculate the total area even with different overlapping rectangles

nathan rogers

I like the way you handled inputs

yeah, like collapsing multiple in a row of the same type

rak1507

really my inner dfn should've been two or three lines to make it clearer, inline assignment isn't great

nathan rogers

um

can I load one thing from dfns without eliminating the entire workspace?

dzaima

@nathanrogers eliminating?

nathan rogers

22:45

yeah when I )load dfns the entire workspace is replaced

dzaima

@nathanrogers ⎕cy'dfns' (and 'specificThingFromDfns'⎕cy'dfns' if you only want a specific thing)

nathan rogers

* Benchmarking "solve '01~'"
             (ms)
 CPU (avg):  2219
 Elapsed:    2217
      ]runtime "#.aoc2015.s06_1  ('.? ?\d+,\d+'⎕S{(⊃,∘⍎1↓⊢)⍵.Match~' '})¨2015 #.day 6"

* Benchmarking "#.aoc2015.s06_1  ('.? ?\d+,\d+'⎕S{(⊃,∘⍎1↓⊢)⍵.Match~' '})¨2015 #.day 6"
             (ms)
 CPU (avg):  2516
 Elapsed:    2522

@rak1507 yours gooder

rak1507

I imagine the regex takes some time

nathan rogers

yeah I can try without loading in the file

CPU (avg):  2485
 Elapsed:    2484

oof

@dzaima so when adam says "BQN Is Good"

what does that mean

dzaima

@nathanrogers presumably, that BQN is good (it has first-class functions and syntax for arbitrary type arrays)

nathan rogers

22:55

how do you convince yourself that the literal notation is good? Because I don't see myself getting past that

1_2_3_4_

dzaima

@nathanrogers that's not talking about the 1‿2‿3‿4 notation, but ⟨1,2,3,4⟩

nathan rogers

I mean the former

@Marshall how is not simply inverting the case where 1 is a 1 element vector 1 2 is a 2 element vector but having some kind of prefix for scalars like .1 is a scalar or something

dzaima

yesterday, by dzaima

@ngn I'm not exactly a fan of ‿-notation, but the alternatives aren't good (always using ⟨⟩-or-whatever-notation is a lot of overhead, and k-style array tokens and APL strands both make for weird syntactic things)

nathan rogers

how does that not solve the problem

or why not strand notation inside of those brackers?

brackets

1 is 1, ⟨1⟩ is a 1 element vector

⟨1 2 3 4 5⟩

why are those not valid alternatives

dzaima

@nathanrogers how about ⟨2+2,3+3⟩?

nathan rogers

22:58

2 + (2, (3 + 3)) ?

dzaima

@nathanrogers , in BQN is equivalent to ⋄

nathan rogers

er no

dzaima

so ⟨2+2,3+3⟩ is ⟨2+2⋄3+3⟩, and both evaluate to JSON [4,6]

Transcript for