Yeah, HPB is also selling it via Amazon... and someone (GlassFrogBooks) thinks they can actually get $132 for it!

@Adám Ah Ok, so the idea is to able to click on some file, it starts the interpreter, fixes all the code and runs something similar to quadLX

@Salim Yup. 18.0 will have such a pathway though.

Excellent. Is the design of this somewhat settled?

@Salim It'll be through a JSON/JSON5 configuration file which will contain instructions to the interpreter on what to do at startup.

OK. Is the configuration file an independent entity? Or is it bundled with a folder full text files full of code?

@Salim The file is a separate thing, and is mainly used as a cross-platform way to set settings for the interpreter, but also has the option to start or load arbitrary things.

Ok, cool, thanks.

Does there exist a nonquadratic idiom for `\`

@AntonDyudin a linear scan can be implemented through each (accumulating with ,←) or reduce or recursion

example with reduce

example with recursion

example with each

(doesn't work well with empties)

yeah, it needs to be special-cased if you expect \ not to call the operand for an empty ⍵

¨ calls it once to determine the prototype of the result

Is the prototype that important when we have mixed arrays...?

And is there really existing code that relies on the behavior of ¨ on empty arrays? That sounds pretty fragile.

IMHO ¨ should stop it and produce a "prototype-less" empty array, which will error when a prototype is requested, e.g. 1↑x or ⊃x.

@Adám I have a question that is bothering me. I finished the second problem set on "language processing" of last year's competition. I wrote an UbbiDubbi translator, a Rovarsprket or Rovarsprekt - or whatever it is called :S - translator and a PigLatin translator.

@RGS OK?

but I don't know if they handle all the possible types of inputs because I don't understand the input spec. I understand that the input spec says I only have to handle characters in [a-zA-Z ]

But then the "type" of the input is something I don't understand

Quoting the spec, "the argument to the functions you will write will be a character array which could be a scalar (single character) or a vector of length 0 or more. If it's a vector, it might be a single word (no spaces), or a set of words separated by spaces."

So I assume my function can be called with f 'c', f 'this is a sentence' or f 'word'

@RGS Right, but also f ''

are these the only "types" of inputs? Or should I handle things like f 'a sentence' 'c' 'word' 'another cool sentence here'?

@RGS No, those are not included. Otherwise it'd say a "vector of character vectors or scalars"

@Adám ok

Then why is the 'c' case not included in the description "a vector of length 0 or more"?

Because "0 or more" includes "1", and I thought a vector of length 1 would be ~ a scalar, but the way the two are mentioned separately, made me feel like a vector of length 1 was something completely different... :/

Btw, @all happy Easter if you celebrate it!

@RGS They are indeed different. 'a' is a scalar and ,'a' is a vector. Try 'a'≡,'a'

oh...

but '' is the 0 length vector, right?

'' ≡ 0⍴'asfadsf' gives 1

@RGS Correct.

alright

then let me see which of my translators don't handle all types of inputs

There is no direct notation for a 1-element vector (yet — coming soon…)

@Adám :-D

@Adám so this turned out to be needed again

and ⎕SE.Dyalog.Utils.repObj myWeirdObj gave 0⍴⊂''

what is that..?

@RGS I need it so often that I've put it on F11

@RGS An empty vector of empty character vectors.

I can tell you how I produced it, if that is relevant

@Adám hmmm

@RGS I while back I started writing a function that would articulate a description of an array in plain English. I wonder if I still have it around.

if the first dimension is empty, then would f¨ myWeirdObj leave it untouched?

@RGS No, it'd do (⍴myWeirdObj)⍴f⊃myWeirdObj

Oh that is unpleasant...

For "pure" algorithms, it tends to "just work" out. For heavy application code, I tend to write my own cover: ForEach←{0∊⍴⍵:⍵ ⋄ ×⎕NC'⍺':⍺ ⍺⍺¨⍵ ⋄ ⍺⍺¨⍵}

@RGS indeed it is. i made the decision to throw that out as it's just annoying.

… but even my cover isn't perfect, as it makes assumptions about the roles of the arguments.

@Adám what's that : ?

@RGS A dfn guard.

ah I see

similar to Haskell's pattern matching?

or an if-else condition?

@RGS Just a simple if-then-return-that-else-continue-with…

@Adám is 0⍴⊂' ' also that?

@RGS No, that's an empty vector of 2-element character vectors :-)

how can that even make sense xd

heh, try: (0⍴⊂' ')≡(0⍴⊂'')

I'm not saying I don't believe you... it just looks weird...

Think about an empty array as the same array but with all 0s in the shape changed to a special "empty-1".

how so?

If we denote the "empty-1" as ¹ it should be clear that (¹⍴⊂' ')≢(¹⍴⊂''), no?

I mean, it is not that weird, really...

(This is in fact how empty arrays are represented internally. Their shape has 0s, but the data has a single element.)

Being "empty" is a property of things that is easy to understand. And I can live with the fact that there are two different bags, one bag expecting "strings" of length 0 and the other "strings" of length 2

and those two bags are empty

@RGS Is an empty pallet of egg boxes the same as an empty popcorn bag?

@Adám refer to my most recent comment :)

@RGS Yeah, sorry, I was too slow.

"ninja'd"? :D

Pre-empted.

what sort of operations raise a DOMAIN ERROR?

my PigLatin translator doesn't work for '' and I am trying to understand what operation fails... Alas it is the very last one.

As far as my experiments tell, I can use f/ arr on an empty array arr, correct?

@RGS E.g. attempting arithmetic on non-numbers, or giving nested arrays to a primitive that doesn't expect one.

@RGS Only if f has a known identity element.

AH!

Can I specify my own identity?

@RGS Not in Dyalog, only in NARS2000. (See also APL Wiki on the subject.)

@Adám what syntax does NARS2000 use for that?

(i cant type..)

@dzaima A system label in a tradfn: ⎕ID: — Execution starts there if an identity element is needed.

@Adám ah, i had a similar thought of allowing ⎕ID←… but right, that'd still require executing the function (short-circuiting is an option, but what arguments to give?)

another idea - implement ToE/metadot, and be able to set stuff with that?

Correction: apparently the label works in dfns too… somehow.

Alright, if my PigLatin translator receives '', would it acceptable to return 0⍴' ' or 0⍴' ' or something like that? (the amount of blank spaces changed)

@RGS Remind me, what is the normal output? A simple character vector?

Yes

@dzaima Full treatment

sentence in, sentence out :P

@RGS Then I think you should always return a simple character vector, even when the input is empty. It is as simple as applying monadic ∊ to the result before returning it.

@Adám ok thanks

Well

What do you suggest? I have 6 solutions that I would like feedback on :D

@Adám oh huh. seems ⍺ doesn't get set and ⍵ is set to the prototype. Makes sense.

(the first two easy problem sets)

@RGS Suitable for CR, maybe?

all of them?

I really don't mind

But not sure if I'd post the 6 in a row, or 1 per day, or whatever

because then it'd also take some effort for you or Bubbler or someone else to reply decently

Here in chat you can just give a couple of tips/hints/comments; over at CR you have to structure a proper answer

Either way. Maybe post here the ones that probably don't need a lot of feedback, while ones you are unsatisfied with can go on CR?

Seems sensible. Are you up for a quick review then? :P

@RGS I'll try. I might be interrupted by my kids soon.

ahaha sure :)

(how many kids do you have?)

@dzaima I have aired the idea of dfns allowing ~∊0 1 guards for these things, but as proper guards, so you can do various preparatory work.

@RGS 2, aged 6 and 8

For the UbbiDubbi and the Rovarspraket translators I followed pretty much the same structure, so if one is good then the other is probably good as well...

UbbiDubbi ← {
  ⍝ Monadic function expecting a character vector and returning a character vector.
  ⍝ Translates an English sentence into Ubbi Dubbi.
  ⍝ e.g. 'Hello' gives 'Hubellubo'

  ∊ ('ub'∘,)¨@(∊∘'aeiouAEIOU') ⍵
}
Rovarspraket ← {
  ⍝ Monadic function expecting a character vector and returning a character vector.
  ⍝ Encodes an English sentence with the Rövarspraket Swedish game.
  ⍝ e.g. 'hello' gives 'hohelollolo'

  consonants ← ⎕UCS(32∘+,⊢)⎕UCS ⎕A~'AEIOU'
  ∊ (⊢,'o'∘,)¨@(∊∘consonants) ⍵

@Adám that's sweet :D

@RGS You can lowercase with 819⌶ (just a few more months until we get ⎕C…) so (819⌶,⊢)⎕A~'AEIOU'

@Adám someone showed me something like that but I don't really like 819⌶ too much... it gives me no clue as to what is happening :P even though you said somewhere in this chat that 819 kind of resembles the word "big"

@RGS Yeah, BIg :-D ⎕C is for Case

@Adám I prefer ⎕C over 819⌶!

@RGS 'ub'∘, doesn't need parenthesising.

@RGS (⊢,'o'∘,) might be clearer as {⍵'0'⍵}

@RGS You can always do ∆C←819⌶

@Adám doesn't need or shouldn't have? Because parenthesizing the tacit function creates that little section

@Adám this is 110% true

@RGS Just my style.

@Adám (no one is looking; is your style the go-to style for the competition or does my style look acceptable as well?)

@RGS Yours is perfect. At least one judge disagrees with my opinion on parentheses.

@Adám :') fair enough ahaha

I'll be sure to get that judge's favour

Any more comments on the two translators above?

@RGS Being that these are structurally identical, you could write:

_Transform_←{∊⍺⍺¨@(∊∘⍵⍵)⍵}
UbbiDubbi←'ub'∘,_Transform_'aeiouAEIOU'
Rovarspraket←{⍵'o'⍵}_Transform_((819⌶,⊢)⎕A~'AEIOU')

Oh how come I didn't think of that

That is really elegant

and that is the sort of thing I usually do

at least with languages I'm really comfortable with

(but I'd go with "_Translate_" maybe)

Thanks for this! I have to go now :D

Gotta go too, my kids are literally crawling on me:

:D have a nice Sunday

@Adám :) they look a lot like you :D

just noticed my scalar function handling for dyadic functions 'forgot' any bit specialization after 1 depth, so while 1 0 0∨0 1 0 used the bit-booleans, (1 0 0)(0 1)∨(0 1 0)(1 0) resorted to converting everything to doubles. :|

@dzaima don't take me wrong: I notice you write thoughts like that in here with some frequency. do you write those for anyone in particular or just as self-notes?

Because I noticed you also reply to your own messages with some frequency :P

@RGS - Does it really matter? I've noticed that if someone has a thought prompted by one of @dzaima 's notes, he (dzaima) will readily engage in discussion about it, explaining his thought processes if asked, or listening to alternative ideas if offered.

@RGS while this message is of the more pointless ones (not even causing any unexpected/buggy behavior), usually i hope to bring up some discussion about something, much as ^ noted

alright :)

fair enough, I was just curious :)

@Adám Is there a setting to persist the scrollback log? My macbook is getting on in years, just now the battery cut out at "50%", and upon reopening Dyalog had lost considerable context; definitely back past the last )save and I think even past the last "graceful" close and reopen (at which point it had shown up the existing log just fine), though ideally I'd have it on something like a 1-second timer(if modified).

Not the end of the world since I had in fact )save'd most important things but still

@AntonDyudin Do you know about APLcart?

@AntonDyudin You could auto-save in the background using a Timer object.

nod

@Adám - @AntonDyudin has a good, if minor, point - a session isn't always checkpointed, and if the terp is ungracefully terminated - on my Surface Pro 3, there's something going on with Windows 10 so that it ignores my Windows Update settings, and forces reboots when it wants, rather than when I want - one might lose unsaved work in the session, or the session log. Some way of forcing specific and/or periodic checkpointing of the session might be a Good Thing.

I mean yeah I come from a background of urbit, where "program requires user-initiated termination and data loss occurs upon kill-9 or otherwise being run over" is something that is Not Done - though I recognize even like chrome could be better in this regard!

@JeffZeitlin what do you mean by "my Windows Update settings"? is there even such a thing? :p

@dzaima Yeah. Microsoft has gracefully re-allowed us to decide when Windows updates (something that used to be a given back in the good old 7 days) — within limits.

@AntonDyudin @JeffZeitlin How about something like:

SaveLog←{'/tmp/DyalogLog.txt'1⎕NPUT⍨⊂⎕SE⎕WG'Log'}
'AutoSave'⎕WC'Timer' 1000 1 ('Timer' 'SaveLog')

Of course, this makes an unnecessary file write every second. A better event to attach this to might be the SessionPrint event, but it is likely to be taken by the handler for boxing etc., so you'd have to factor that in. Possible enhancements might be to read the file and check if it is identical to what we're about to write, and also limit the number of lines to the last N.

@dzaima - Yes, things like "active hours", and if you know where to find them, some additional registry hacks that ostensibly do things like force Windows not to reboot without being explicitly told to...

Check the size of the file, append anything past that…

@AntonDyudin @JeffZeitlin Of course, if you use Link and the IDE's editor, all your functions will automatically be saved.

Re: factor that in, you'd grab the current handler bound to the event and call it first, right?

@AntonDyudin Yes.

@Adám - Which is good for 'fixed' functions, but not for the stuff that's not ready for fixing - basically, the sort of thing where I'm still experimenting to determine whether my approach is even valid. That ends up in the session log, so that when I have to run away for a bit, I can come back to it and see what I'd already looked at.

My functions are saved it's more a matter of 20 iterations or calling them with different parameters getting various results

exactly

Right, I understand. It makes perfect sense to want one's log saved. And of course, if APL shuts down normally, it will save the log, but you're not alone in seeing the occasional odd closure.

@Adám - Exactly. It doesn't happen often - but when it does happen, it always seems to be at the most annoying possible time... :)

^^ file I will not be able to precisely recreate (well it uses ? but even beyond that), though I do have an earlier iteration in a "demo" thunk (brought on by similar experiences in a previous crash)

@JeffZeitlin Murphy dictates that, just like he dictates my boiler breaking when under lockdown and repairmen are prohibited from visiting…

@Adám come to think of it, can I do graphical stuff with APL? I'm pretty sure I can... how difficult is it to create an image?

Btw, something that bothers me is that all my APL instances share the same log. I often run many APL processes in parallel. An enhancement to the autosave thing would be to include the process ID.

@Adám the boiler with which you make tea, or the boiler that heats the water for the baths??

@RGS House heating and warm water.

(which, for the curious, sounds like dropbox.com/s/0ohjab84121spym/wav.wav?dl=0)

That is unfortunate :(

@RGS Did you try APLcart?

@Adám I'm sorry... I didn't ^^' I'm just too impulsive and ask questions right away

I've jut been bitbanging raw .wav PCM sample and .bmp pixel bytes into a data uri into 3500⌶ 😅

Still trying to get my head around the ⎕WC object system, see: my first question on joining this channel wrt how do websockets and events

@RGS I'm not annoyed or anything, I'm just trying to teach you to fish.

@Adám - Oh, yes, I'm a firm believer that Applied Murphology is a science. You may be interested in Zeitlin's Reformulation of Murphy's Law: "Anything that can go wrong has already done so - you will simply not notice it until the least opportune moment."

@Adám I live by that quote but I'm used to being the one teaching how to fish. When the roles are reversed, I tend to forget that knowing how to fish is way better than asking you for fish :upsidedownface:

And yes, being able to identify which session of several running ones would be useful.

@JeffZeitlin Nice one. Also: Any amount of anticipating Murphy's law, will exclude the worst event bound to happen according to Murphy's law.

@AntonDyudin How did you find it?

APLCart ^-^

:-D

@Adám - /me scribbles that one into his notebook...

@Adám @JeffZeitlin Guys, allow me to simplify it for you

I had been looking for a srand earlier but in other places, and it occurred to me again after boldly stating "well it uses ? so it's not reproducible reproducible" 20min ago…

"sh1t happens"

Excuse me the stronger language... but I believe it captures well what you were discussing.

@AntonDyudin Thanks, I'll add "srand reproducible reproduce" as search terms to ⎕RL. Any more relevant terms to add there (or anywhere else)?

"deterministic" perhaps

@Adám "seeded"

and "scroll-back" was something I think I had searched wrt saving the session log

@RGS - Murphology is about how $#!+ happens, but you can't anticipate it. Yeah, $#!+ happens, but sometimes you can anticipate it, and mitigate it.

@JeffZeitlin well alright :P

@JeffZeitlin Heh, that's a fun use of those symbols. I've long feared the day I needed the enclosed intersection of the union and the encode… as a train.

@dzaima @AntonDyudin Thanks for the synonyms. Added now.

@Adám - /me snickers.

@JeffZeitlin You've seen the APL puns, right?

There's a clothing company that sells in Macy*s here, called "French Connection UK". They tend to use their initials on the labels and logos...

@Adám - Actually, no. Got a pointer?

@JeffZeitlin APL puns.

@Adám - Thank you, sir!

@AntonDyudin Yeah, I noticed earlier. I'll see if I can get anyone to kick it into submission.

oh I think I figured out a callback thing

(which I suppose is unidiomatic and what operators are for?)

@AntonDyudin Funny you mention it. I actually just saw an internal proposal for implementing proper closures.

y'all have proper closures ∘ is the "make function pointer into closure" operator :P

Given the current behavior I expect that would be an extremely breaking change? but I suppose you could have an operator or quadfn or sth for "as of the evaluation of this operator, bind all visible names into a namespace and attach it to the ⍺⍺ as a this"

Function::bind etc

Context for this is :54039319 fwiw

> (a:1 ⋄ f:{a}).f⍬ returning 1 would be cool and not too impossible otherwise

@AntonDyudin That's my namespace notation I presume?

(being able to edit an external constant as in e.g. youtu.be/a9xAKttWgP4 seems at least at risk of breaking if not directly affected)

not-good ways I've generally ended up using: 4 3⍴ 1 2 3 4 5 6 7 8 9 0 1 2,

↑(1 2 3)(4 5 6)(7 8 9)(0 1 2)

@AntonDyudin Works by me:

      ⎕SE.Link.Deserialise '(a:1 ⋄ f:{a}).f⍬'
1

@AntonDyudin You can just download 17.1 (no registration required)

@Adám - just a random comment... The stylized D-in-hexagon logo that the current Dyalog APL uses as the icon in Windows is much nicer than the 3-d block letters "APL" that was the icon for the free versions of 13 and 14...

@JeffZeitlin Certainly more modern, but it has always bothered me that it is physically impossible without trick photography: The D's shadow can't fall like that on a cube (it isn't a hexagon). Another thing is that I suspect the change was a delayed reaction to moving from "APL" to "Dyalog" because "APL" had a bad rep. Ironically, APL is becoming hip again, so now I think we should emphasise that Dyalog [APL] is just a progressive APL implementation, not a separate language.

@Adám yep. I wanted to say that that is now working in dzaima/APL, but, as-is, it isn't, as i still haven't fixed my dot-notation parsing (it's been a problem for a long, long time), and it only fully (ish?) works when the first object is a variable name.. (((a:1 ⋄ f:{a}).f)⍬ and m←(a:1 ⋄ f:{a}) ⋄ m.f⍬ both work)

@AntonDyudin I don't think so. Rather it is an effect of management taking necessary steps to modernise.

("no good" here a phrase meaning "disruptive to the very particular thing I'm hacking together" - though probably id someone else 3500⌶s an autoplaying element they'll also want to hear it)

@AntonDyudin 3500⌶ renders via Electron, so I think it is RIDE that's at "fault" instead. Have you tried the real CEF, ⎕WC'HTMLRenderer'?

I have not; this looks much more involved but I'll give it a shot

ran docs.dyalog.com/17.1/… to no obvious effect, should this work on macOS sierra?

@AntonDyudin It should. What does ⎕NEW⊂'HTMLRenderer' return?

@Adám +100000 for quoting me saying "APL s fun" :-D

The advice on using regexes is quite sensible and I'll finish digesting it tomorrow, thanks! For now, a lingering doubt must be expressed: isn't using regexes """"cheating"""" (notice the quadruple ") in the sense that I'm not using much APL?

#.[HTMLRenderer]

@Adám - Is APL becoming 'hip' again outside of things like CGCC? Certainly, I've believed for many years that it's severely underrated, and that C-like languages were becoming overrated and over-used, to the point where people can't handle more restrictive languages like Pascal... or APL.

@RGS From a purity point, sure. But as an integral part of the Dyalog language, you're simply using the tools at hand. And the multiple-patterns feature is actually unique to Dyalog.

@Adám so it is perfectly acceptable to use in an APL competition?

@AntonDyudin That seems fine. Then you should be able to create and keep it simply by assigning the result to a variable. I have no idea where a new window is shown on macOS, so you may have to dig around a bit.

@RGS Absolutely.

Alrighty then! I'm off to bed; see you guys tomorrow.

○/

XD APL'd

\o

it is not at any rate in the window list

@Adám they had the right idea though, with the shadow becoming "longer" further down the darker side. Anything more realistic looks awful however

@AntonDyudin (⎕NEW⊂'HTMLRenderer').Visible?

I have figured out a workaround, thank you for pointing me towards Electron; not sure what the protocol is for submitting patches but uh ix.io/2hN3

      (⎕NEW⊂'HTMLRenderer').Visible
1

—I beg to differ XD

⍨

@AntonDyudin Can you email support@dyalog.com about this so appropriate parties can investigate?

Sure done, not sure what information to include precisely but I linked this conversation + screenshots + "I tried to open a window and it didn't work help"

⎕WC objects don't persist with the workspace so I'd have to stash the AutoSave initialization inside some sort of startup script, right?

@AntonDyudin Uh, they do persist, but you probably don't want to always use the same workspace. A thing like this belongs somewhere in ⎕SE methinks.

(relatedly, )continue only closes the main RIDE and not its attached 3500⌶ popup, making it impossible to reopen Dyalog without closing the popup first)

@AntonDyudin I'd log that as an issue against RIDE.

Oh I suppose that also answers the PRs question!

Dyalog/ride#570 filed