« first day (2640 days earlier)      last day (9 days later) » 

6:21 AM
@Adám ah, similar to how Key in Kap is a function?
I feel like the problem with just returning the sections is that you lose border info
 
True, but you can compute that based on the parameters you gave.
 
yeah, but it's kinda annoying
unless I add a border option that is "no border"
 
But what does that mean?
 
smaller arrays around the edges
 
Ah, that is an option, of course.
 
6:35 AM
I would also like to add a "custom fill" option, because no prototypes means that the default cannot be 0's (or spaces or whatever)
would a doubly enclosed array work?
 
Not sure what you mean?
 
I'll try to make an example when I get somewhere I can use my computer, it's hard to explain on mobile
oh, but Stencil right now uses rows, not enclosures, right?
 
Yes.
Btw, current (operator) Stencil, as an entire argument that is unused. It could be used for a fill value. Alternatively, the row containing edge handling spec, could have (0 fill)
 
how would that work with the fill mode extension?
oh, can I choose a different mode for each axis?
 
The first row indicates window sizes
The second row indicates step sizes
The third row indicates initial offset
The fourth row indicates edge handling (0=Zero, 1=Replicate, 2=Reverse, 3=Mirror, 4=Wrap, `0 fill`=fill with given value)
@RubenVerg Yes, that's already the case. Hence 3 3 for GoL. I've not thought about what happens outside the corners if you fill with two different values along the two axes. Maybe indeed better to only allow a single fill value, as variant or left arg.
 
6:43 AM
maybe the former axis wins?
I think I like the ability to have different fills on each axis
@Adám initial offset is also an extension right?
 
Yes, and probably the most needed one, with 0 being very common but painful to currently obtain.
@RubenVerg I guess having a predictable precedence here allows you to use to decide which one you want to win.
 
right.
I like discussing how primitives should work with you (:
is an extension similar to what you describe ever coming to dyalog, in your opinion?
(same question for all of your many other extensions, especially those in Vision which seem more likely to appear in it)
 
The pure extension to offsets and maybe edge handling? It is already on our internal list of language extensions that we are currently considering. But I'm way too busy with more important stuff atm.
x⊇y and f⍥k are already likely to be pushed off from 20.0. I'm fighting to keep
 
ah, that's sad. what are the more important features that are being developed instead, if you can talk about them?
(when is 20 planned to release, btw?)
@Adám does this list contain anything that's not mirrored in any of your extensions? (assuming it's not secret features, of course!)
 
7:28 AM
@RubenVerg Compatibility features for migrants from APL2 and APL+Win.
@RubenVerg Not really planned yet, but I'd guess mid-2026.
@RubenVerg I'll have a look, but doubt it.
 
@Adám ah, boring stuff (:
@Adám is there a large userbase of people coming from apl2 and a+w?
 
 
2 hours later…
9:43 AM
@RubenVerg Yes, right now, because IBM has sold APL2 to a company that isn't making it available, and APL2000 isn't going to release any new (or 64-bit!) versions of APL+Win (instead telling people to migrate to APL64 doesn't quite have APL+Win's quality). So if we put in the effort now to catch these customers, we can look forward to a substantial and reliably stable income, which we can then allocate e.g. to new features.
 
 
5 hours later…
2:29 PM
How can I create a sliding window with increasing size on a vector? I tried
((2 3 4)∘.,/⊢) or
((⍳4)∘.,/⊢)
but that is not working, probably because the length of the resulting vectors are all different
 
 
2 hours later…
4:35 PM
2 3 4{⍺,/⍵}¨⊂1 2 3 4 5?
 
 
2 hours later…
7:00 PM
@LdBeth Why are you wrapping the ,/ function in a dfn?
@Richard Do you want the window to increase as it progresses (as indicated by your "increasing size" or just all the combos (as indicated by your "∘.")? If the latter, then, remember that for each of the sizes, you want the entire right argument, so you need to enclose the right argument. Also, remember that operators bind from the left, so you need ∘.(,/) instead of ∘.,/ which becomes (∘.,)/. Also, as @LdBeth does, you can do with a simple ¨ if your left argument is a vector.
 
7:30 PM
@Adám Ah, my impression told me n f / is a function, now I remembered f/ is a dyadic function (bivalent, actually)
 
 
2 hours later…
9:40 PM
Hi everyone, new to this channel. I was introduced to apl recently by @Richard's interview on youtube. Still looking for a proper introduction to the language. Working through the wiki now.
 
10:04 PM
@user24320873 Welcome. The wiki is a good place to start. The blue "Hello world" box on the front page should give you what you're looking for.
 

« first day (2640 days earlier)      last day (9 days later) »