damn it floating point!

As much as I think ⎕IO should be 0, it is quite convenient being able to do (≢⌷⊢)

I have too many 1-⍨≢⍵s in my code now

<moon-child> somebody here linked a list of all matching unicode delimiters (e.g. {} [] () «» etc.) a while ago; anybody still have that link?

APL may be the best for arrays but it sucks for strings! would rather use python

is there a function that lists contents of a namespace?

couldn't find anything on APLcart

<moon-child> one option (admittedly somewhat hackish): )cs yournamespace and then )vars / )fns / )ops / )classes / )obs

very minor graphical issue in ]html: the buttons at the top don't adjust their size so you get this

@rak1507 what kind of content are you looking for? ns.⎕nl -2 will give you the variables, ns.⎕nl -3 will give you the functions, -9 will give you objects (namespaces, etc). And you can specify finger granularity with decimals, check out the documentation for ⎕NL. You can even give it an array as an arg, so ns.⎕nl - ⍳10 is a common way to look at what the namespace contains

⎕JSON is also a new quick and nice way to view contents of simple namespaces (no functions)

⋄ns←⎕ns '' ⋄ ns.(x y z)←1 2 (⍳3) ⋄ ⎕JSON ns

@Josh D
Illegal code
VALUE ERROR
VALUE ERROR

oops, not quite sure how to make the bot print those multiple lines

@rak1507 a string is just an array of characters :). lol, I hear the complaints about strings in APL a lot, but I never had a string problem I couldn't solve in APL. Sure, when I first started, forgetting to ⊂ character arrays to make a "string" was a hiccup, but imo it fits nicely with the array/depth paradigm.

I had a thought the other day -- what if double quotes just automatically made characters inside a "string" so "abc" would equal ⊂'abc'

<moon-child> jsoftware.com/papers/array1.htm#5 relevant

Didn't give it much thought though, seems like a waste with no real benefits

Oh wow, that seems like a much better use of the double quotes

are dictionary native structures in J ?

<moon-child> no

<moon-child> (but they are in k, and symbols there have a similar role. Syntax is `foo instead of "foo")

Ah ok, ty

I'm just realizing the thought of single quotes for characters and double quotes for strings most definitely came from when I wrote C years ago.

Even though C doesn't really have "strings", still just arrays of characters with a null terminator.

<moon-child> well, it's also cultural. US uses ""; canada (where iverson comes from) uses ''

Good point, didn't think of that

<moon-child> (personally I'm more inclined towards single quotes for use in prose, but I don't think strings are important enough in apl to use up such a visually distinct character)

@rak1507 ⊢⌿ not good enough?

@JoshD year and a half ago. (I still think that having "…" be ,'…' just for single character charvecs is worth it, and the need for ⊂ should be fixed in the builtins needing it (which isn't really an option for Dyalog, but that's not my problem))

@JoshD yeah, but interpolation is a pain

@ngn now that you say it, it's so obvious! thanks

@RomillyCocking do you know this one: what's a good anagram of "banach-tarski"? :)

this is probably more of a maths question than an APL question but does anyone know if there's a way to find the derivative of a function at a point iteratively? would be nice if there was some way to do ⍣≡ so it converges

ofc I can do something like {¯0.00001÷⍨-/⍺⍺¨0 0.00001 + ⍵} but for exponentials that will probably be off by a fair amount

{dx←1E¯20 ⋄ dx÷⍨-/⍺⍺¨⍵+dx,0} will probably do I guess

I'm playing with parsing JSON expression trees from JSON namespaces instead of JSON matrix, and the only way of working with it that currently jumps into my head involves recursion, and potentially very nested vectors. Does anybody here have any performance trivia / ballpark estimate off the top of your head - do you think this would be eventually worse than index-searching and exists-fn on simple arrays I have to do in the matrix approach?

ie. I haven't started to do much of anything yet and I'm already drowning in (⊂'foo'),∇¨e.children

@ktye yeah a dfn probably

@rak1507 Slightly better is mathspp.com/…

what is it with APLers and making incredibly useful pasting features!

thanks, I didn't think of that but now you say it, it is obvious

IIRC the formula I shared is of 2nd order, whereas the one you are using is just first order.

Of course, the formula is only of 2nd order IF the derivative exists :P otherwise the numerical results are bogus.

well yeah lol

@rak1507 by bogus I mean that they don't make sense mathematically, because they might still evaluate to a nice number. That is what I mean.

yeah, I know

alright alright, mb.

try sin(1/x) for example

@ktye specifically at $x = 0$, otherwise there is nothing wrong with that function :)

@RGS fwiw you have a useless erroring g at the end of editor_mode(tex_area.value? 2 : 0);g which makes keyboard shortcuts not load

@dzaima THANKS

lol it is right there in the console. I did find weird the shortcuts not working XD

@dzaima btw thanks for dzaima/paste :)

:)

@dzaima Hard-refresh and it's fixed. Thanks for that.

<klg> for complex differentiable functions one can use something like {⍺←1E¯13⋄ ⍺÷⍨11○⍺⍺ ⍵+¯11○⍺}

klg I think I'm missing something... You only use the function ⍺⍺ once. Is your dfn supposed to approximate the derivative of ⍺⍺?

*The derivative of ⍺⍺ on point ⍵.

<klg> yes, derivate of ⍺⍺ at ⍵

@DyalogAPL How does your dfn do that? It looks like your dfn computes mathspp.com/…

<klg> you're missing the 11○

Yup you are right. So we are at mathspp.com/…

Correct?

<klg> it's called complex step differentiation and is much more robust than finite differences with regard to choice of a step size; the drawback is requirement of the function being complex differentiable

<klg> yes, correct

If the formula is correct, this looks neat! Approximating a derivative without even having to compute a difference.

I wonder why I never came across this. This doesn't look that advanced.

whoa, that is really interesting

@dzaima Fixed.

@Bubbler Fixed.

The 3270 font now supports APL and BQN out of the box!

Samples shown on my fonts page now.

@xpqz ⎕FMT always returns a character matrix, and can format numbers in various ways ― og and it "evaluates" control chars. ⍕ can only round and/or use exponential format while returning a non-nested character array of any rank.

@rak1507 Switch to ⎕FR←1287 and it'll give exactly 0.

@rak1507 What does that even mean? You could erase all names with ⎕EX⎕NL-⍳10 or simply save a clear workspace on top of the file.

Ah yeah that makes more sense

moon-child: Unicode: Brackets, Quotes «»「」【】《》

@rak1507 Not our fault. Simply how Windows draws its buttons on narrow windows. Happens to other applications too, if they allow making their window that narrow.

Ah

If you're trying to do things like parametrized SQL expressions, the SQAPL interface does allow for bound variables in your statement.

I'd rather do f'<path d="M {W} C {X}, {Y}, {Z}" stroke="black" fill="transparent"/>' than '[WXYZ]'⎕R{⍕⍎⍵.Match}⊢'<path d="M W C X, Y, Z" stroke="black" fill="transparent"/>'

The whole thing is {W X Y Z←⍵ ⋄ '[WXYZ]'⎕R{⍕⍎⍵.Match}⊢'<path d="M W C X, Y, Z" stroke="black" fill="transparent"/>'}

if there's a nicer way to do this, please inform me

@dzaima oh wow, that's cool. Seems like this is a common idea that pops into our heads. There definitely seems to be untapped potential with double quotes

@dzaima that's true. but that's the tradeoff of allowing mixed arrays of different types. I think in development, it pays to know what variables are numeric and then adding the ⍕ should come naturally. If i find myself doing a lot of (⍕i) i'll just assign it as n←⍕i or something

I don't think that was the issue

@rak1507 I guess {W X Y Z←⍕¨⍵ ⋄ '<path d="M ',W,' C ',X,', ',Y,', ',Z,'" stroke="black" fill="transparent"/>'}

W X Y and Z are numeric vectors

oops missed the ⍕¨

that's ok but not much better than the regex one imo, at least then it's clearer to me anyway

and to get rid of remembering commas

{W X Y Z←⍕¨⍵ ⋄⊃,/'<path d="M ' W ' C ' X ', 'Y', ' Z '" stroke="black" fill="transparent"/>'}

if you're doing this a lot, it might pay to make a simple utility function

you can probably even make the interpolate function smarter by using ⎕DR (or maybe even ⎕VFI) to only convert numeric variables to char

would have to benchmark the performances

@JoshD seems Dyalog is smart enough to make ⍕ on a character vector a no-op

yeah, i would imagine that is the case

another problem with string interpolation in other languages is now you have a new keyword/character to escape

which actually ties in with the discussion of double quotes. it's kind of nice not having that as a reserved character because it's used a lot in HTML. if that was included as a symbol, i guess that would make HTML string construction even more complicated because it would have to be escaped now

not necessarily? you could use 'text with a " in' fine right?

You guys might like these.

I guess you're right, " in single quotes will be fine. But I'd imagine there would be some fallout somewhere. Maybe on things like expression executors now need to escape them.

@JoshD Yeah, this was discussed during the BAA open session Thursday. It is currently quite simple to determine if a particular character is "quoted away" or not. With a double-quote string (whether a normal vector, enclosed vector, or new type) it becomes a much more complex task. Not sure if that is a good argument against adding, but worth mentioning.

Nice Adam, there's your interpolation as other languages have it. I still think it's messier than using other solutions like dzaima's (e.g. you need to escape the backtic key now)

@Adám I saw them, if I was doing a lot of interpolation I might do that, but for one svg element it wasn't really worth it

I once used {(W X Z)←⍵ ⋄ ∊⍕⍤⍎¨@(∊∘'WXZ')'<path d="M W C X Y Z" stroke="black" fill="transparent"/>'} and found a bug in @…

when i did this before, i used dyadic functions with this, so the left argument contained the literals, and the right contained the numeric values

and you insert the ⍕ of ⍵ between ⍺

just thought of doing it this way, maybe there are more efficient ways

⋄ 'var one = ' ';var two = ' ';'{v←⍕¨⍵ ⋄ ∊⍉↑⍺ v}1 2

@Josh D
┌→────────────────────────┐
│var one = 1;var two = 2; │
└─────────────────────────┘

Nobody mentioned {W X Y Z←⍵ ⋄ ∊⍕¨'<path d="M 'W' C 'X', 'Y', 'Z'" stroke="black" fill="transparent"/>'}

right, super ravel (∊) is extremely helpful with this issue

lol, super ravel

It is unstoppable. It'd be nicer to have a single-level join than this overpowered thing.

@RomillyCocking i forgot to reply to this one: ngn.bitbucket.io/k/…