@cannadayr This line is setting the value of a variable: a is the value array for the block and i is an index into it, so a[i] is the variable's value. So [a,i] is basically two values simulating a pointer into the variable values.
BQN is all about immutable data, but the VM implementation is not.
@cannadayr d is whether the assignment is a definition/declaration (←) rather than a modification (↩). The variable needs to be undefined iff (===) d is true. Then id is the assignment target and v are the values to set it to.
@cannadayr Yes, the point is that you're modifying the variable value, which is stored with the block. So a is just a copy of a mutable array.
So an instance of a block, which is buried somewhere in a closure in D(), has a variable array, and a is a copy of that array. Update a and the copy with the block instance will also get the new value.
thx, ill need to mull it over a bit and cont to pull apart the js. everytime i think i understand javascript hah...
on plus side i got this earlier (likely wrong): ``` 16> erlang:apply(lists:nth(1,bqn:smap(B,S)),[[]]). E:[{array,148,0,undefined,1000},[]] S:[] C:{15,1} E:[{array,148,0,undefined,1000},[]] S:[#Fun<bqn.10.91007752>] C:{22,3} E:[{array,148,0,undefined,1000},[]] S:[#Fun<bqn.10.91007752>,{array,148,0,undefined,1000},0] C:{11,6}
my interpreter port breaks in set() as well, it can set a single value but trying to run the full compiler breaks at the assert eq(id.sh,v.sh)
it's not the set() doing it, I'm suspicious of the way one is 21 things and the other is 42 things, and it appears to be 21 things alternating with 1,2,3,4,5... I stopped with no good ideas how to make progress into it
@Marshall good to hear that confirmed as a possibility; powershell does the equivalent of throwing ravel into your code almost (but not quite) everywhere. I've squashed that a few times
@TessellatingHeckler The variable references don't really need to be pairs (they just need to be recognized by get and set), so if there's a way to encapsulate them then just use that.
@TessellatingHeckler The list for multiple assignment gets created in case 4. Variable references should never escape to have actual functions called on them, so that should be the only place to fix. I assume that's where they're getting squashed together.
@Marshall gah, case 4 has caused me a bunch of problems already, I totally believe it's still wrong. It is doing "pop N items, reverse them, push them back individually not as a single nested group", yeah?
@cannadayr PowerShell currently; if I can both (make it work and understand it) that might lead to C#, and IIRC you're disinterested in .Net in all forms
@Marshall gotcha; yes I am using list() to push them back at once, and that's how I got to be able to test basic list store/retrieve
@TessellatingHeckler If you're getting a length-42 list, it has to be case 4 that's breaking, because calling the instruction with an argument of 21 should always produce a length-21 list.
@TessellatingHeckler It stands for "get environment". It starts at the current environment (the data for a block instance) and takes its parent i times, then returns it.
Variables are referenced by the depth of the environment where the variable was defined, then the index in that environment.
I changed the numbers from 1,2 to 77,88 so they stand out more from indexing. First case 4 leaves the stack with a single item, list 77,88, that looks good.
Then gets to case 22, drops into ge() with e as a nested emtpy list something like [[[],[]],[]]
@TessellatingHeckler That looks about right. e should have two parts: the variable array and the parent. The variables are a and b, which aren't initialized yet, and it has no parent so the parent doesn't matter.
what happened on case 21:{let v=ge(e,num())[num()];assert(v!==null);s.push(v);break;} to make it worth putting an assert there? If I "fix" variable assignment, I get a null v here.
@Razetime You can assign a keyboard shortcut to that too. It is called FX. Edit → Preferences → Shortcuts → Search: FX then hover over the line, and click ➕ then press your desired shortcut.
@TessellatingHeckler the idea with the ___M functions (LOCM=22, ARRM=4) is that they create assignable/mutable arrays, that "mutate" the thing the ___O-version would've otherwise returned. In my impl they're a new type (SET_s for reference) completely separate from regular values
({𝕨𝕗_𝕣_𝕘˜⁼𝕩: 2} is the most "complex" header that can get processed currently, but its execution isn't yet implemented)
so GNU APL allows a 5th argument with bracket axis, right? How about a b⌾(c _𝕣_ d)e:? :P (and a b⌾(c _𝕣_ d˜)e: for squeezing another variation in the same function, for extra horror value)
@dzaima actually, a b⌾(c _𝕣_ d˜)e: doesn't exist. Got it confused with a⌾((b _𝕣_ c)⟜d)e: or something like that i really don't know anymore
Inspired by this.
Given a string as input consisting of only upper and lowercase alphabet characters, wicka-wub it.
How do I wicka-wub a string?
The example text used is "DJMcMayhem".
Split the string before each capital letter, so you get ["D", "J", "Mc", "Mayhem"].
Next, take the two halve...
pushed inverse headers. No modifiers currently because i can't decide what to do about providing 𝕊 to them (not as in what should it be, but whether to pass the Md[12]Derv to all callInvs or always re-derive it (as it's only passed for regular calls))
CMC: Write a function that returns 'https://dyalog.com/download-zone.htm#termscheck' as its result. Rules: The last line of the function must be https://dyalog.com/download-zone.htm#termscheck unless you write a dfn, in which case that should be the second-to-last line, and the last line must be }. The required line must actually be executed. No early termination allowed.
@dzaima so, for a non-result-enclosing dfn to work, either 1) there's some magic way to redirect the dfn output from the result of //…; or 2) my logic about what //x does is wrong
@dzaima (double-double-quotes is a bit excessive there. just regular double-quotes is enough)
@dzaima 106, still with a side-effect of outputting garbage
@dzaima 111 with hiding that garbage
@dzaima 100 with outputting garbage (¯1 this time)
@ngn finally looked at that; #.htm←com← would only assign in # for one of the assignments (i just assumed that #. for assignment wasn't needed); also why 1⌷⊖?
so my latest solutions: 114, dfn: https://tio.run/##SyzI0U2pTMzJT///qKMt@VHbBAtDy0c92xQe9c5VyC8tSUksSU1RCPH0/58GlKsuARGPelflPOpuedS7C4hqueKBYo96@w6t0NDIedTRBTamb2qi5qOuthyglLqCet2j3hXGj7qagcJ5Repp6po6h1aog6RAtKGyYYlyiSGI/airSR1osKE6V0ZJSUGxlb4@xGl6yfm5@in55Xk5@YkpulX5eal6GSW5yiWpRbnFyRmpydlctf//Aw0HGpkGdPgaAA 107, tradfn: https://tio.run/##SyzI0U2pTMzJT///qKPN@VHbBAtDy0c92xQe9c5VeNQ31Rko3F4EFE7jKgGS1Y96Vz3qXVPL9ai379AKDcNHbZM1ih51dIH19k111HzU1QZSra6gXveod4XJo65moHBekXqauqbOoRXqICkQfWhFyaEVyiWG6lwZJSUFxVb6@hBH6CXn5@qn5Jfn5eQnpuhW5eel6mWU5CqXpBblFidnpCZncwGd8/8/0EyQkwA
<phantomics> One other thing: Dyalog throws an error when you enter 5⍳7, but GNU returns 2, is there any harm in taking the latter approach when I mostly follow Dyalog?
Similarly, monadic ⍳ treats a 1-element vector as a scalar, so now it is inconsistent on vectors of length 1 (returns an array of vectors in all cases except when its argument is a scalar or a 1-element vector).
@dzaima Yes, and you can easily get ∊∘, if that's what you want, but not the other way around.
phantomics: I don't know of a summary, but you may have luck on APL Wiki's individual primitive pages. Also, btw, 5⍳7is a RANK ERROR in APL2, so that's a GNU APL special.
@Adám i guess i should trust you on that "yes", but i'm still highly skeptical about it (probably just because i highly prefer vector-of-column data structures to rank-2 arrays)
I wish you could use APL in competitive programming, was just doing one and I was struggling to get a working answer in both python and haskell, but it's so simple in APL :(
@Adám i would assume that'd mean they'd need to add checking code, which means writing APL, which they definitely can't do. And they definitely have many many more popular languages they'd actually be able to add before APL
supporting APL for 0-1 people is probably not worth it
Yeah, it just needs to be able to run code and check against stdout. According to my friends I could make a post or email them and they might consider it, so I might as well
@Adám it'd definitely be their first time considering legality of using a programming language. Also, since the finals aren't online, they'd need to set up APL input system for the competitors
@TessellatingHeckler In the JS runtime, null marks an undefined variable. Line 21 gets the value of a variable, so it needs to be defined. Even though declarations are all resolved statically, a variable can be undefined at runtime because inner blocks see all declarations from the parent block, even from statements that come later. So {a}⋄a←6 should trigger that error.
@TessellatingHeckler Is this when running the full runtime? Note that the value for no left argument (Nothing or ·) has to be different from the one for an undefined variable.
In my Go runtime I didn't make the distinction and just removed the test, but that means it doesn't detect uninitialized variable errors. One reason I never published that runtime.
@Marshall I have blindly ported the old JS interpreter to PS, and running it tries to run the full compiler and can't. I've been gradually making simple bytecode in the JS one and porting it accross and trying it and fixing the errors. The current case 21 error is if I try and run the full compiler. With {a}⋄a←6 either I've got the bytecode wrong or it's breaking earlier
When I started, I had no clue how it worked and no way I could write it in anything else, now I have some idea how it works and am very tempted to restart in a more sensible language, but this feels pretty close
I should probably say "glyph mapping" instead of ⎕ML because it's faster than opening an APL-enabled editor, typing it wrong before remembering APL keyboards use backtick instead of backslash, and copying into here.
<phantomics> I have a button on my desktop bar that runs the setxbkmap command, but the changes it makes are lost when I switch input modes
<phantomics> I have a Japanese mode as well as the English mode, switching into it and back returns me to the default mode where the winkey works normally
when v was an empty array my assertion test was failing incorrectly; with that fixed, "{a}⋄a←6" runs and outputs 6, but it doesn't affect the full compiler run erroring in the same place, for a different reason
@dzaima I did say that; it comes out with v=@() (empty array) which means v is defined and no error; x←{a} ⋄ a←6 ⋄ x appears to hit case 21 with v ending up as nothing, assertion fails
powershell has a problem with arrays of one thing being flattened into scalars, and this has me feeling that is happening somewhere, where I'm either losing something that should be an empty array, or making an empty array where there should be nothing
@Marshall another question, writing it out because i just got done looking at it and its fresh on my mind...
some of the core runtime functions (ie "(x,w) => x.sh?1:0 ") take 2 arguments x and w. but some of them test if w is defined and will return different results.
will these functions always be called with 2 arguments (but one is undefined), or is this relying on javascript defaulting an argument to undefined?
i ask because in erlang functions have to be called with the right arity, while javascript doesnt.
@cannadayr in a language without default args, you should have all functions always take 2 args (and call always take 3 args), and always call them with both
Inspired by this.
Given a string as input consisting of only upper and lowercase alphabet characters, wicka-wub it.
How do I wicka-wub a string?
The example text used is "DJMcMayhem".
Split the string before each capital letter, so you get ["D", "J", "Mc", "Mayhem"].
Next, take the two halve...
