A look at BQN with my favorite mindset, negativity.

@Marshall wdym by "that are only paired for their glyphs and not for any other reason"?

@dzaima Both meanings somehow relate to the symbol, but they don't relate to each other.

@dzaima The glyph is mnemonic for both monadic and dyadic valences, but otherwise the two valences are unrelated functions.

@Marshall cant do something interesting without some challenges

@Marshall ah, managed to not read the title of that :|

im interested in the APL seeds class, because I think I will start to appreciate some of the design decisions more

@cannadayr Maybe, although my concept of Seeds was more about how to implement than what to implement. I'll definitely keep writing about why I made certain choices in BQN. Maybe there could also be some structured discussion that's dedicated to design at some point.

"Can't return from inner functions" - i don't think i've ever needed this. But if it's really needed i'd prefer either operators or proper control structures

"Tacit and one-line functions are hard to debug" - with bytecode a char-by-char debugger should be possible?

(@Adám how can I get a string representation of an operand ⍺⍺ inside a dop?)

@dzaima In BQN you can do things like lists of functions as switch/case statements that makes it seem more important. The reason I'm leaning away from control structures is that the function replacements are more flexible. They just have the return problem.

i personally need the foundation of "how" (especially where it relates to backend array manipulation) to get the "what"

@RGS aa←⍺⍺ ⋄ ⎕NR 'aa'

@cannadayr Yes, they're definitely connected that way.

@Adám awesome, thanks!

@dzaima And {(x) ∇ y→… for sharing code?

i don't see how sharing code would otherwise work (assuming you can't check for variable existence)

Would x←⊢ be allowed?

@Adám probably not in BQN, but in APL i guess it'd be the implementers choice

@Marshall "Comparison tolerance" - i think there really only needs to be one thing about it - an equality checker with an epsilon arg (that's 3 args though, unfortunately)

"Poor font support" - BQN386 :)

@dzaima When you can't choose the font, of course.

I'd say a fuzz operator would do.

@Marshall (obviously)

"Inverse is not fully specified" - i actually really like relying on it, and at least the way i implemented it in dzaima/APL it's very well-defined (aka each function can define it to be whatever it wants ¯\_(ツ)_/¯)

"Rank/Depth negative zero" - dzaima/APL separates 0 and ¯0 (they equal one another, but still are different) so when i finish implementing ⍥ i'll definitely make ¯0 work :p

"Must read the body to find explicit definition's type" - again, dfn headers (i think allowing no headers is still good, maybe not for operators)

@dzaima That could be precarious though. If you arrive at a 0 through arithmetic the sign might not be what you expect.

@Marshall true (solution - just don't do arithmetic for those args)

"The directions of ⊏⊐ and so on were mainly chosen to line up with ∊" i don't see the need of resemblance

"Syntactic type erasure" - maybe •𝕗 etc for a bool if that's the type given?

@Adám after looking at it zoomed in for a while, except “≡≢, i don't think mine looks worse. Maybe there's a bigger difference at smaller sizes

@dzaima It is definitely a hinting issue.

@Daneolog Hi there. Interested in APL?

@Adám it is, hence why i said smaller sizes. but in your picture scale matters much more

also what happened with the underline in ≥..

anyways, pushed whatever i had changed (+ fixing ≥), if you want to, you can look at hinting because i just can't test anything

I really don't know anything. I tried the auto-hinting, but that was awful.

@Adám ah, that's pretty much what i would do too, except see results ¯\_(ツ)_/¯

(just to make sure, you did do autohint and then autoinstr?)

@dzaima Can't remember. It was a long time ago.

@Adám I was just peeking in to see some discussion since I'm somewhat new to code golf, but I'll see if I have any questions! thanks!

@Daneolog For general discussions on code golf, the Nineteenth Byte is the right room. This room is dedicated the practical APL programming language, both when used for golfing, and when not.

@Adám and more often than not the discussions are not about golfing :P

@RGS for APL Orchard or the Nineteenth Byte? (or both, likely)

@Wezl I meant this room, I can't speak for TNB because I am never there, but maybe it also applies to TNB room

Say I have an operator _T expecting a single character on the left and an operator _S_ expecting on the left and right things of the form 'x'_T

And I want to apply _T to a series of characters and then "join them together" with the _S_ operator I have. I wish I could write something of the form _S_ / _T¨ someChars but this doesn't really work :P

But I also don't want to write down dozens of _Ts and _S_s...

one possibility would be to generate the string of code I want to have and then ⍎ it... but this also doesn't sound that elegant

(totally unrelated but why can I type _T and view its source but if I type _S_ I get a SYNTAX ERROR?)