« first day (1832 days earlier)      last day (822 days later) » 

thejonymyster
thejonymyster
1:08 AM
is there any way to see how a train is evaluated for any given like, train
im trying to break down functions from the cart to try and understand them but i cant pick them apart
 
thejonymyster
thejonymyster
1:31 AM
in particular this: (10∘⊥⍣¯1)N, to get the digits of N. the usage of ¯1 here within a train doesnt match up with what i currently know about trains / what i thought i understood from what i read
 
Josh D
Josh D
1:48 AM
:60298219 try running this in the session:
]box on -trains=tree
 
thejonymyster
thejonymyster
thank you kindly
 
Josh D
Josh D
then just enter your train in the session, without the N
 
thejonymyster
thejonymyster
oh wow, ⍣ was at the center
would not have expected that ever
but yea thank you thats exactly what i was looking for
yaay
 
Josh D
Josh D
And ¯1 is a special right operand to the power operator, which applies the functions inverse. the Docs for Power will explain this
 
thejonymyster
thejonymyster
aha
 
Josh D
Josh D
1:50 AM
no problem!
 
thejonymyster
thejonymyster
wait but thats not the power operator, thats repeat?
but now i know how to mess with it to get a better understanding, so thats ok yay
oh, it does say power here, hm
 
Josh D
Josh D
It is the power operator, you can apply a function N times, -N will make it apply the inverse of that function N times. Or you can apply a function until a certain condition is met
 
thejonymyster
thejonymyster
i was looking at the tryapl description and it was different from the dyalog ide one, interesting
 
Josh D
Josh D
Check out https://help.dyalog.com/18.0/index.htm#Language/Primitive%20Operators/Power%20Operator.htm

https://xpqz.github.io/learnapl/iteration.html#power
 
thejonymyster
thejonymyster
ahhh
that makes more sense
 
Josh D
Josh D
2:00 AM
Although the use case in the train you're looking at isn't iteration, just to apply the function's inverse
 
thejonymyster
thejonymyster
right, cause its only -1
so is it inverting ∘? or the whole of 10∘⊥
 
Josh D
Josh D
 
thejonymyster
thejonymyster
ooh :]
 
Josh D
Josh D
yeah it's inverting 10∘⊥
there's a few other styles you can use ]boxing with to see the train in different ways
]box -??
in the session will show you
 
thejonymyster
thejonymyster
thank u kindly
very helpful
 
luser droog
luser droog
2:19 AM
peekaboo
Stumbled upon this chatroom a few days ago. Howdy everyone!
I've got a small project to make an APL interpreter that looks like the code in the original 1962 book. After hacking up a javascript keyboard, I found Adám's keyboard program and followed links over to the APLwiki which led me here. now.
Are there any plans for more APL Seeds? I'm studying the transcripts from the earlier ones. But more is more, you know?
 
Brian BED
Brian BED
2:40 AM
{(?10 10⍴1000)<¨⍵}100 10
this does not work... anyone know how to make it work? like what i want to happen is it makes multiple arrays and each has a diffrent percentage of ones based on the input, but i can't get the ¨ operator to do < on each element of ⍵.
 
Razetime
Razetime
3:11 AM
@BrianBED the ¨ operator when given two arrays will try to "match" their elements together
{(⊂?10 10⍴1000)<¨⍵}100 10 or {(?10 10⍴1000)∘<¨⍵}100 10 should be fine
 
LdBeth
LdBeth
(⊂?10 10⍴1000)<¨⊢ for tacit
but beware the "once only" problem
 
 
5 hours later…
Adám
Adám
8:11 AM
@thejonymyster You mean the name "Repeat" vs "Power"?
@thejonymyster In addition to what Josh said, you can do 'tc'⎕CY'dfns' then insert tc to the immediate right of every function you're interested in. This will print the argument(s) and result every time the function is called. E.g. 
      (+⌿tc÷tc≢tc)3 1 4 1 5
 ≢  3 1 4 1 5  =>  5
 +⌿  3 1 4 1 5  =>  14
14  ÷  5  =>  2.8
2.8
But you still need to know the syntactic class of every glyph; function/monadic operator/dyadic operator.
@thejonymyster Operators have long left scope, so for 10∘⊥⍣¯1 the ⍣ "looks" to its left for an operand, and finds ⊥. Does ⊥ have a companion left argument? Let's look further left. Oh, ∘; that'll bind ⊥, so we need to look further left. 10, OK, that works, but maybe it is bound but something further left. Nope, there's a ( so we're done.
 
Adám
Adám
8:59 AM
@luserdroog I don't think so. Those classes kind of become in-depth BQN development discussions.
@LdBeth Or better: (?10 10⍴1000)∘<¨
 
 
2 hours later…
xpqz
xpqz
10:40 AM
@Adám I don't see how that can be, unless I'm missing something (entirely possible), or there are a further number of hoops to jump through. Let's say I have the pattern 'ab' and the replacement string 'ba' and the text 'ab ab ab'. I want every possible replacement, so that's: 
'ba ab ab' 'ab ba ab' 'ab ab ba' 'ba ba ab' 'ba ab ba' 'ba ba ba'
If I run this with ⎕S with a function operand of something like this: 
reg←{
    s←⊃⍵.Offsets
    pre←s↑⍵.Block
    post←⍵.Block↓⍨s+⊃⍵.Lengths
    pre,'abc',post
}
I only get:
 
Adám
Adám
@xpqz I don't understand what you mean by "every possible".
 
xpqz
xpqz
      'ab'⎕S reg⊢'ab ab ab'
┌→────────────────────────────────────┐
│ ┌→────────┐ ┌→────────┐ ┌→────────┐ │
│ │abc ab ab│ │ab abc ab│ │ab ab abc│ │
│ └─────────┘ └─────────┘ └─────────┘ │
└∊────────────────────────────────────┘
 
Kamila Szewczyk
Kamila Szewczyk
not sure if this is the right place to post this
but i'm willing to give +100 to anyone who posts a golfed idiomatic APL solution to codegolf.stackexchange.com/questions/205809/…
 
Adám
Adám
@KamilaSzewczyk ? Oh, I thought you were reacting to xpqz's posts.
 
Kamila Szewczyk
Kamila Szewczyk
ah, no, it's about the bounty.
i thought that posting it here might attract some APL programmers
not sure how i'd approach this problem myself so i'd like to see how someone solves this
 
Adám
Adám
10:43 AM
@xpqz Ah, so you don't want a global (or even limited) replacement; you want all subsets of replacements (except the empty set).
 
xpqz
xpqz
That's it, yes. All the different ways you can successfully replace the pattern in the text.
@Adám :) me, too. Was furiously reading back my words looking for something offensive.
 
Adám
Adám
@xpqz Missing 'ab ba ba'?
 
xpqz
xpqz
Eh yes, looks like it. I was generating those whilst typing... missed it.
I guess one could do ⎕R inside the operand, up to Offsets+Lengths, but that seems hideous.
 
Adám
Adám
@xpqz Something like this: 
⋄  ⎕IO←0 ⋄ 'ab'{⍵⍵{⍺⍺⍴⍨∘≢@⍵⊢⍵⍵}⍵¨∊¨+∘⍳/¨¨((l⍴2)∘⊤¨1↓⍳2*l←≢s)/¨⊂s←⍺⍺ ⎕S 0 1⊢⍵}'ba'⊢'ab ab ab'
 
TryAPL
TryAPL
@Adám
┌────────┬────────┬────────┬────────┬────────┬────────┬────────┐
│ab ab ba│ab ba ab│ab ba ba│ba ab ab│ba ab ba│ba ba ab│ba ba ba│
└────────┴────────┴────────┴────────┴────────┴────────┴────────┘
 
Adám
Adám
10:56 AM
However, this assumes the replacement has the same length as the replacee.
Ah, I have another idea.
 
xpqz
xpqz
:) I think that validates my solving this without regex...
 
Adám
Adám
11:13 AM
@xpqz OK, here's a general regex solution: 
⋄ ⎕IO←0 ⋄ 'ab'{((l⍴2)∘⊤¨1↓⍳2*l←≢⍺⍺ ⎕S 0⊢⍵)(⍺⍺{⍺⍺ ⎕R(⍺{0=⍵.⎕NC'n':⍺⍺ ∇ ⍵⊣⍵.n←0 ⋄ (⍵.n+←1)⊢⍵.Match ⍵⍵⊃⍨⍵.n⊃⍺⍺}⍵⍵)⊢⍵}⍵⍵)¨⊂⍵}'ba'⊢'ab ab ab'
 
TryAPL
TryAPL
@Adám
┌────────┬────────┬────────┬────────┬────────┬────────┬────────┐
│ab ab ba│ab ba ab│ab ba ba│ba ab ab│ba ab ba│ba ba ab│ba ba ba│
└────────┴────────┴────────┴────────┴────────┴────────┴────────┘
 
Adám
Adám
Despite the looks, it is actually quite simple. We just iterate over the masks to see if each match should be replaced or kept as-is.
 
xpqz
xpqz
Wow. Fair play.
 
Adám
Adám
I think a double pass (as indicated by using both ⎕S and ⎕R) is necessary, as the number of results increase twofold each time we hit a match.
 
Razetime
Razetime
11:33 AM
@KamilaSzewczyk does regex count as idiomatic APL
cause Regex would do quite well here
 
Kamila Szewczyk
Kamila Szewczyk
@Razetime nope
i know that regex does well here :)
 
 
2 hours later…
Adám
Adám
1:07 PM
@KamilaSzewczyk I'll match that, and then double that sum if the answer is also well-explained.
 
thejonymyster
thejonymyster
in re: these three pings:
1. yeah, im not sure why theyre differently labelled but im sure its not a huge deal
2. aha, this will be very helpful though thank you, thats exactly what i was hoping to achieve
3. makes sense, im just surprised it can get the inverse of longer functions like that. very cool
:D
 
Adám
Adám
3. is Dyalog's magic sauce ftw.
 
Kamila Szewczyk
Kamila Szewczyk
@Adám hahaha, so +300 for someone in total :)?
 
Adám
Adám
Yes, I'll add 300 to your bounty.
 
Kamila Szewczyk
Kamila Szewczyk
faaancy.
can you cross one of my bounties off your list and gift this rep to the person who posts a solution?
it might be easier this way.
 
Adám
Adám
1:18 PM
I've already fulfilled all your requested bounties.
 
Kamila Szewczyk
Kamila Szewczyk
oh oops. nevermind then
 
Adám
Adám
1:48 PM
@KamilaSzewczyk does this do the trick? 
{∊{0=≡⍵:⍵ ⋄ '(',(∇¨⍵),')'} {(0=≡)tc⍵:⍵ ⋄ ('I'≡⊃)tc⍵:⊃⌽⍵ ⋄ (2=|≡⍵)∧('K'≡⊃∘⊃)tc⍵:⊃⌽⍵ ⋄ (3=|≡⍵)∧('S'≡⊃∘⊃∘⊃)tc ⍵:(x z)(y z)⊣(((s x)y)z)←⍵ ⋄ ∇¨⍵}⍣5⍎⍕'\w'⎕R'''&'' '⊢⍵}
 
Kamila Szewczyk
Kamila Szewczyk
beautiful
 
Razetime
Razetime
Anyone got some testcases I can test an SKi simplifier on?
 
Kamila Szewczyk
Kamila Szewczyk
I'll test it later
 
Adám
Adám
Not golfed yet, and I have tcs in there too.
 
Razetime
Razetime
{'I'≡⊃⍵:⊃⌽⍵⋄(2≡≢⍵)∧(2≤|≡⍵)∧'K'≡⊃⊃⍵:2⊃⊃⍵⋄(2≡≢⍵)∧(2≡≢⊃⍵)∧(3≤|≡⍵)∧'S'≡⊃⊃⊃⍵:(2⊃⊃⊃⍵)(2⊃⊃⍵),¨⊃⌽⍵⋄∇¨⍵}⍣≡{' '~⍨,⍵,⍨⍪' '''[1+0,2≠/'()'∊⍨⍵]} is what i've got so far.
works on the single test case given. that question could definitely use comprehensive tests.
@Adám hm did not think about pattern matching
do you plan on using an error guard
 
Adám
Adám
1:52 PM
How? Why?
 
Razetime
Razetime
well if the pattern match assignment fails you can try a different pattern.
and then finally if all the patterns fail just return ⍵
 
Adám
Adám
Afaict, I check if the depth is 3, so the assignment must succeed, I think.
 
Razetime
Razetime
oh ok
 
Adám
Adám
@KamilaSzewczyk Are we allowed to use regex to insert quotes or no regex at all?
 
Kamila Szewczyk
Kamila Szewczyk
2:15 PM
@Adám i'd rather you not use regex
 
Adám
Adám
No worries. It is like 2 bytes longer without.
 
Kamila Szewczyk
Kamila Szewczyk
oh, very cool
i wonder whom i should award the bounty to in the end, provided both of you post
 
Adám
Adám
@KamilaSzewczyk What do we do about loops? My solution alternates between ((KI)(((SI)I)((SI)I))) and ((KI)((I((SI)I))(I((SI)I))))
 
Kamila Szewczyk
Kamila Szewczyk
@Adám assume looping input as illegal input, i think that's what i did. 
'I'≡⊃⍵:⊃⌽⍵
(2=d)∧'K'≡⊃⊃⍵:⊃⌽⍵
can't this be golfed to a single guard?
i guess that no short circuiting might be a problem maybe?
 
Adám
Adám
That shouldn't be a problem. Thanks.
 
Kamila Szewczyk
Kamila Szewczyk
2:28 PM
i feel so stupid, it took me a few lines of code to write what you expressed as ⍎⍕{3⍴''''⍵}¨@{⍵∊⎕A}⍞
i guess one good thing about my solution is that it was flat and didn't use eval... but that's not what you care about in code golf :)
 
Adám
Adám
Well, you asked for idiomatic code. Although not sure what "golfed idiomatic" means. Oxymoron much?
 
Kamila Szewczyk
Kamila Szewczyk
oh - i just didn't want regexes and that sort of stuff
 
Razetime
Razetime
well there's probably a way with the depth array from -⌿'()'∘.=⍵
 
Kamila Szewczyk
Kamila Szewczyk
(your solution is good i like it)
by the way - @Adám shouldn't ((K(SS))S) simplify to (SS)?
 
Adám
Adám
I have no idea. I don't know anything about this; I'm just happily implementing whatever the OP says.
 
Kamila Szewczyk
Kamila Szewczyk
2:33 PM
SKI combinator calculus
The SKI combinator calculus is a combinatory logic system and a computational system. It can be thought of as a computer programming language, though it is not convenient for writing software. Instead, it is important in the mathematical theory of algorithms because it is an extremely simple Turing complete language. It can be likened to a reduced version of the untyped lambda calculus. It was introduced by Moses Schönfinkel and Haskell Curry.All operations in lambda calculus can be encoded via abstraction elimination into the SKI calculus as binary trees whose leaves are one of the three symbols...
using the rules here
((Kx)y) = x
your implementation gives ((SS)S) which seems like an easily fixable problem
(SS) = x, S = y, for the record
 
Adám
Adám
Right, I see that.
Huh, I fixed that, but broke OP's test case ⍨
@KamilaSzewczyk I had a typo there, making it seem like they can be combined.
 
Kamila Szewczyk
Kamila Szewczyk
you can use this for testing expressions
 
Adám
Adám
@KamilaSzewczyk I wouldn't really know what to put there :-) but I've added that test case. Have a look.
 
Kamila Szewczyk
Kamila Szewczyk
very nice
 
Adám
Adám
Now I so want monadic ⊇ to be ⊃⌽, (which is a Dyalog optimised idiom, btw, so a great candidate for a primitive).
I wonder if (x z)(y z)⊣(((s x)y)z)←⍵ could be written better.
x y,⍥⊂¨⊃⌽(((s x)y)z)←⍵
 
xpqz
xpqz
3:05 PM
@Adám the TIO link to the database keyed property in chat.stackexchange.com/rooms/52405/conversation/… is a bit truncated; any chance you recall what the statements were that you ran in that demonstration? I see: 
i←⎕NEW DataBase
i.K[⊂'Dave';'name' 'age']←'David' 31
i.K[⊂'Ernie';'name' 'age']←'Ern
 
Kamila Szewczyk
Kamila Szewczyk
3:29 PM
      v←?⍨10
      (⌊/v)←6
SYNTAX ERROR
      (⌊/v)←6
           ∧
@Adám :(
 
Adám
Adám
⌊/v is always ⎕IO here.
But you can't assign to a value, only to an element of v
 
Kamila Szewczyk
Kamila Szewczyk
oh, hmm
could one make this work?
 
Adám
Adám
((⊃⍋v)⊃v)←6 works
 
Kamila Szewczyk
Kamila Szewczyk
oh. true
 
Adám
Adám
((v⍳⌊/v)⊃v)←6 too.
 
Kamila Szewczyk
Kamila Szewczyk
3:32 PM
so [] and ⊃ return elements that i can assign to
is modified assignment documented strictly anywhere?
 
Adám
Adám
Yes, and ⌷ and the function / etc.
I was about to refer you to APLcart, but it points at the wrong place.
However, this is actually selective/indexed assignment.
 
Kamila Szewczyk
Kamila Szewczyk
hm, thanks
assignment in APL is so cool
reminds me of elixir style pattern matching but much better
 
Adám
Adám
@KamilaSzewczyk … except in dfns. Dfns are broken.
 
Kamila Szewczyk
Kamila Szewczyk
how so?
 
Razetime
Razetime
Dec 7 '20 at 17:32, by Adám
@voidhawk … ⊢{a←1 ⋄ }0 giving a value error; can't trace into a one-liner dfn; →0 in the session quits instead of resuming execution; plus←+ ⋄ a plus← and plus←+ ⋄ 1 plus a←2 overwriting plus; :: failing to catch ⎕SIGNAL etc.
 
Kamila Szewczyk
Kamila Szewczyk
3:39 PM
oh some of these issues i knew about
will these be fixed one day?
 
Adám
Adám
The first one could be fixed, the second is so-so because it is ambiguous, the rest cannot be fixed without braking lots of existing code.
 
Kamila Szewczyk
Kamila Szewczyk
> Can't. The curse of a user base with existing code.
aaaaaaaaaaaaaaaaaaaa
 
Adám
Adám
We are contemplating adding a variant of tradfns which would provide auto-localisation and anonymous definition, with the assumption that those are actually the main reasons people like dfns.
 
Razetime
Razetime
adding new syntax that is similar to an existing thing is giving me C++ flashbacks
 
Kamila Szewczyk
Kamila Szewczyk
APL++!
 
Adám
Adám
3:44 PM
What else can we do?
 
Kamila Szewczyk
Kamila Szewczyk
~~we need a new ML~~
 
Razetime
Razetime
@KamilaSzewczyk Dark Souls boss music plays
 
Kamila Szewczyk
Kamila Szewczyk
@Razetime are you planning to post your solution, btw
 
Adám
Adám
@KamilaSzewczyk Yeah, but changing syntax with "⎕ML" is more severe than just changing semantics.
 
Razetime
Razetime
no, you can award the bounty to Adám
 
Kamila Szewczyk
Kamila Szewczyk
3:45 PM
@Adám we could have a toggleable "sane apl" vs "traditional apl" thing
or something
fair
@Adám feel free to chip yourself +300 while i set up +100 now
 
Adám
Adám
Uh…
 
Kamila Szewczyk
Kamila Szewczyk
(joke)
done
 
Adám
Adám
I'll just sit here basking in my own rep.
 
Kamila Szewczyk
Kamila Szewczyk
haha
i can't award it now but i'll see if i can tomorrow/in 2 days due to this dumb restriction. it's up anyway.
 
Adám
Adám
No rush. Feel free to leave it for a week for maximum exposure :-)
 
taronish
taronish
4:21 PM
Can I use rank (⍤) to iterate over the columns of a matrix, without like transposing the matrix first?
 
Adám
Adám
Depend on what you're doing.
 
Brian BED
Brian BED
4:32 PM
is there an installation guide on py'n'apl? i am getting frustrated trying to download it because i can't get anything to work
 
taronish
taronish
I'm working 2016 practice problem 1, where we compute the mean, which I thought would be trivial. I'm not sure how to make it work nicely with an empty array other than a guard (I routinely run into this problem - either mapping or reducing across an empty thing and not getting the value I want). Also, I made it work for the multidimensional case, but I'm not sure if there's a superior way to do so.
{⍵≡⍬:0⋄((+/÷≢)⍤1)⍉⍵}
 
Adám
Adám
@taronish Have you tried +⌿ instead of +/?
 
taronish
taronish
@Adám I did, in an earlier version, but couldn't quite make it work since the reduction was happening along the columns but the rest of the train didn't operate on the values I needed (I think, anyway) and I couldn't quite figure out how to fix it.
I can start on another attempt with +⌿ and see where it leads me!
 
Adám
Adám
@taronish In general, for maximum versatility, write your code using leading-axis orientation (⌿ instead of / and ⍪ instead of , etc.) as that will make application using ⍤ much neater.
 
RikedyP
RikedyP
@BrianBED Is there a specific issue? I download the project as a zip and unzip it, then 
]load /path/to/pynapl/pynapl/Py.dyalog
 
Brian BED
Brian BED
4:39 PM
:O imma try that
 
RikedyP
RikedyP
@BrianBED that's "technically" what the first part of the README says, but I can now see that it's not as explicit as it could be
 
taronish
taronish
@Adám I definitely don't have a good grasp on concepts like leading-axis yet (and I'm not sure that the page on the APL wiki is going to be of much help). I got this working, but I must confess I used the tried-and-true method of monkeying around until things worked, and I don't understand why it works with the "2".
{⍵≡⍬:0⋄((+⌿÷≢)⍤2)⍵}
 
Brian BED
Brian BED
@RikedyP yea dude i was basically trying ]load py and every time it gave error i moved the file location.... i did this like 10 times :|
 
RikedyP
RikedyP
@BrianBED Also note: ]cd tells you the current directory
you might be in /win/system32
 
Adám
Adám
@taronish I believe that would fail on say 0 2 3⍴0 but try (+⌿÷≢) ans see if you can fix it.
 
Brian BED
Brian BED
4:44 PM
@RikedyP ahhh yes i am
 
RikedyP
RikedyP
@BrianBED I don't know why Dyalog starts there, but it's easy enough to work around so I guess changing that hasn't been prioritised
 
taronish
taronish
@Adám It passed all the tests on the site ... but I will start fresh with:
(+⌿÷≢) 0 2 3 ⍴ 0
and see where it goes!
Thank you.
 
Adám
Adám
@taronish I'll look at updating the site for more comprehensive tests.
 
Brian BED
Brian BED
@RikedyP imgur.com/1UNVQft uhhh i must be doing something wrong but idk what
 
Adám
Adám
> Persinal
 
RikedyP
RikedyP
4:47 PM
@BrianBED you need to specify the full path down to the Py.dyalog file
 
Brian BED
Brian BED
ah cool one sec
 
RikedyP
RikedyP
e.g. ]load "C:\Users\rpark\Downloads\pynapl-master\pynapl-master\pynapl\Py.dyalog"
Then you should be able to:
py ← ⎕NEW Py.Py
'1+⎕' py.Eval 3
 
Adám
Adám
@RikedyP You don't. It can be relative to the current dir, but if you begin with / it'll be relative to the current drive.
 
Brian BED
Brian BED
@Adám lmao that directory has too many python scripts that have a specified path
 
RikedyP
RikedyP
@Adám ok but I'm trying to give advice that will work, not explain the whole Windows file system,
 
Brian BED
Brian BED
4:52 PM
still error
wait im dumb
 
RikedyP
RikedyP
@BrianBED you have to change the thing
 
Brian BED
Brian BED
yess!! it now said #.py which is indeed not an error :D
 
RikedyP
RikedyP
@BrianBED nicely done :)
 
xpqz
xpqz
Still amuses me: APL in python in APL in python....
 
Paul Mansour
Paul Mansour
5:19 PM
@Adám Seems like failing to catch ⎕SIGNAL could be fixed in a major release. Wasn't the whole scoping rules for error guards changed in not so distance release?
 
Adám
Adám
@PaulMansour You can't change that, as people can rely on a dfn used inside a tradfn to leave.
 
Paul Mansour
Paul Mansour
Dyalog can and has made breaking changes, right?
 
Adám
Adám
Only when very convinced it was a real bug, and even then it was regretted. E.g. ⍳⍬
 
Paul Mansour
Paul Mansour
The way error guards were handled was changed
I don't think it was regretted.
 
Adám
Adám
I'm not sure what it is you have in mind. We're currently fighting with fall-out from the fix to the famous Mantis 17.
But that's dfn scoping in general, not error guard scoping.
 
Paul Mansour
Paul Mansour
5:25 PM
Error guards did not catch errors thrown in dfns in a different capsule
This was changed not that long ago
 
Adám
Adám
Yup, I see that: 
Dyalog APL/W-64 Version 12.1.1
Serial No : 314159
Unicode Edition
Mon Jan 31 17:26:45 2022
clear ws
      foo←{0::'caught' ⋄ bar ⍬}
      bar←{∘∘∘}
      foo 2
SYNTAX ERROR
bar[0] bar←{∘∘∘}
             ∧
 
Paul Mansour
Paul Mansour
This was not a bug, but an original design decision to be more "functional". Scholes and Foad changed it.
 
Adám
Adám
Not sure how that's not dysfunctional.
 
Paul Mansour
Paul Mansour
Static scoping vs dynamic scoping maybe?
 
Adám
Adám
Maybe. Error guards are dynamically scoped, afaict.
 
Paul Mansour
Paul Mansour
5:31 PM
Anyway, the issue with ⎕SIGNAL is only if its in the same capsule, and if its in the same dfn, its a little silly anyway, as one could just use a regular guard. So I wouldn't advocate changing it.
 
Adám
Adám
Can you explain how one could replace an error guard with a regular guard?
 
Paul Mansour
Paul Mansour
Why would use ⎕SIGNAL in a dfn to be caught by an error guard in the same dfn?
 
Adám
Adám
E.g. here I use !# instead of ⎕SIGNAL because the latter would leave.
I don't want to repeat A⊣((⍺⍺)A)←⍵ and having to dfn-wrap and name that would complicate the code a lot.
 
Paul Mansour
Paul Mansour
Right, i said that above, a nested dfn is really the only place this bug hurts.
In a single standalone dfn, ⎕SIGNAL to an error guard doesn't seem to be to useful, I doubt anyone does it.
 
Adám
Adám
The current behaviour does fit the name "signal" as it sends a signal out. You never signal yourself, however you can "throw" something to yourself. Maybe we could simply add ⎕ERROR or similar.
Or maybe just an option on ⎕SIGNAL to stay in capsule.
 
xpqz
xpqz
5:54 PM
 
Paul Mansour
Paul Mansour
6:10 PM
@Adám Good ideas. But regarding non-breaking changes/enhancements to Dyalog, does not almost every enhancement have the potential to break existing code? The only questions is how much code. I'd bet changing ⎕SIGNAL in dfns would break less code than say extending the domain of some primitive.
Or maybe not, but the point remains.
 
Adám
Adám
@PaulMansour As I understand it, we reserve the right to make errors into non-errors. (Took me a while to get used to writing ∘∘∘ instead of *** when I came to Dyalog from APL+)
 
Paul Mansour
Paul Mansour
Right, but you are not going to make 2 3 + 1 2 3 work.
Because the real question is how much code you are going to break.
 
Brian BED
Brian BED
@RikedyP Ok so i hope I'm stupid again but honestly i can't figure out why it isn't importing correctly. i've been trying for like an hour now trying to import APL (which does work somehow) but it doesn't work at all because it says undefined when i use it for eval or anything. imgur.com/IHaG3Le
 
Adám
Adám
@BrianBED Did you do the apl = APL.APL() step?
 
 
3 hours later…
LdBeth
LdBeth
9:39 PM
Can I make an APL function optionally take 0 or 1 argument?
 
LdBeth
LdBeth
10:21 PM
hmm, figured that's impossible, it seems the only one possible to be optional is left argument.
 
Adám
Adám
10:38 PM
@LdBeth Let's say you wanted to create the vector 1 2 but instead of 1 you want your function's return value when called niladically. Easy: myFun 2. Let's say you wanted to call your function monadically with the argument 2. Easy: myFun 2. You see the problem?
 
Adám
Adám
11:24 PM
@xpqz Click on a message to jump to the transcript where you can see the whole thing.
@xpqz Maybe add a note and link that if you then hash your lookup table, your sparse array can be very performant.
 

« first day (1832 days earlier)      last day (822 days later) »