@Razetime No, but I guess I could make one. Also, I only figured out lately how to "recursively" extend glyphs, while working on TryAPL 3, and as you know, TIO has gone stale.
[Made an answer for the interval notation question](https://tio.run/##dVCxSgNBEO39inS7S/b09jYnWttoYZPYXa4IhmgR0MJGYkAQjhjYoEiIjY2iSBCuUBvLy5/Mj5yze8eYO7DYnTdv3rydnd750Otf9oZnJzmYJyZZg3Uje8UMZvM2g@s3zDgewWD6nI/AzKjoKKt3GXeZLJIRB/MNt69gfsQWXhymLxKPB5NHLH1iuFotVgvhCRSOCyPe7Tdd@wjNJJi5auJzCpIH1Gwe9i6OT8cMNV1RqoRVvVuh5@IsS6tq650PILnDGkxvwHyA@cpSDck9OnTae3gf7R908kGDRb4MY7aBiBOynCCuQNGuVLosZ2ko8c1qVjZUarxSQ6WWQbyGqYd4vsbjGD6N5tNoPo1WokiREyHn7wuYLFGn4xoX1bj4Py6glShEBdOqbcGyWbqNOIj/5im02kUtWy62aJmBLHdiBwylop9oqXaslv0C "APL (Dyalog Unicode) – Try It Online")
I'm not sure why, but the ⎕R acts weird when I group everything into a single call
@Razetime multi-regex ⎕R requires that both left and right args are precisely vectors of vectors of characters. So 'a' 'bc' doesn't work, you need ,¨'a' 'bc' or (,'a')'bc'
⍝ Examples assuming parentheses are required for both strands and modified assignments:
f x ← a ⍝ use of pass-through value
(x y)← a ⍝ strand assignment
x(f ←)a ⍝ modified assignment
x f y ← a ⍝ use of pass-through value
(f g) x ← a ⍝ use of pass-through value
f(x y)← a ⍝ use of pass-through value from strand assignment
f x (g ←)a ⍝ use of pass-through value from modified assignment
(x y)(g ←)a ⍝ modified strand assignment
(x y z)← a ⍝ strand assignment
x(f o ←)a ⍝ modified assignment with derived function
Sure, but that's not the point here. Any derived function (including a train) would be allowed.
The list isn't exhaustive, as there are many more possibilities with various name classes. The point is if what's happening can be determined without knowing the name class of anything.
Currently, structural modification using named structural functions isn't allowed, but you have a good point.
Imo, neither structural, nor indexed assignment (nor @) should ever have been added. Rather, we need structural under (which could have used the @ symbol).
though @ has its bonuses too. in particular, it's much better when the items you need to apply to depend on the items themselves (e.g. lowercase only vowels, reverse items with odd length)
@Adám those can't actually ever work, as (f⍢g ⍵) ≡ (g⍣¯1)f g ⍵ → (g f⍢g ⍵) ≡ f g ⍵, but, e.g. substituting the first example, ((∊∘'AEIOU'⊢⍤⌿⊢)'aabB') ≢ 1∘⎕C (∊∘'AEIOU'⊢⍤⌿⊢)'aAbB'
i guess the second can technically, maybe work as ≢ isn't relying on the actual items of ⍵, just the structure
@Adám there's no ⍣¯1 in the checking function, and if it broke a valid ⍣¯1 invariant in a non-erroring situation, either they can't live on the same character or the definition of under in general needs to not have the identity.
@dzaima I think it needs to either analyse its operand or try the structural approach first, and only if the operand clearly isn't structural/selective or if that approach fails, it should go the computational ⍣¯1 route.
@RomillyCocking I'm with Jeff. While well-written in general, I don't find it strongly compels non-APLers (or even more so, APL-haters) to explore APL.
Thanks. I'll have another think. I am deliberately trying not to sound too compelling, as that is fatal in my experience. It sounds as if I have gone too far the other way.
@Adám It's a public service broadcast. I don't want to sell APL. What I want is for people who know nothing about the language to be a little bit curious - curious enough to take the first step.
@Razetime Yup, and some of them are a real issue. Eben Upton (creator of Raspberry Pi) is one.
Cos 1) he wants people to learn languages where anyone can guess what code does, like simple Python and 2) APLers tried to sell it too hard in the past, and pissed everyone off
@ngn I mean, you don't stay excited about any language for long. APL is pretty good at what it does, but unfortunately it doesn't do everything many want
@RomillyCocking the "Maybe not." section is missing the elephant in the room - there's only one decent impl and it happens to be proprietary and commercial, blocking progress as any monopoly would do
@ngn I respect your pov but I do not share it. I believe profoundly that knowledge (including s/w) should be free but I also believe in evolution rather than revolution. Right now APL needs money behind it.
Also backward compatibility is not just an issue of commercial interest. It's also a consquence of respecting the human value of what has already been created, and still needs to work.
Ah I misinterpreted 'respecting the human value of what has already been created' of referring to the programming language itself rather than the software built in it
@RomillyCocking I wonder how these differences get resolved in a good way - how is it possible to "actually get there". (We have a similar situation in the Elm language community where some aspects of how the creator is handling the development make other folks grow impatient with it and want to create a community-led fork... but then, how do you make the fork sustainable?) Perhaps Perl folks managed to pull this off lately?
Making a breaking change (even for the better) is a tremendous undertaking, and can easily take decades, even with conversion tools available. See Python2 → Python3.
@MartinJaniczek If there are fundamental unchangeable differences between impl and users, I think the only sensible thing is to attempt to fork. Doesn't guarantee success, but is the best option
@Adám if it's going to take a while, might as well start sooner rather than later...
Also, one of the biggest issues with Python's conversion was having to port all the libraries that people used, seeing as there are very few APL libraries, that seems like it would be less of an issue
apl users are so few, dyalog could literally go through the list of customers one by one, tell them about the prospect of introducing a breaking change, and ask them to adapt
@ngn Right. So Dyalog is mostly stuck with keeping backwards compatibility, whether you, or Dyalog, or other users, like it or not. Again, best option is to have a separate impl. But there's noone willing to put actual effort into it for approximately zero gain
but i think it's ridiculous to write a paragraph about why not to use apl, and not even mention the problem (whatever one's opinion about it is)
the inconveniences of the character set are so insignificant compared to the problem of eternal backwards compatibility and the lack of good free alternatives
I disagree. I think the main issue today to using APL for production code, is lack of good interoperability. However, all of these issues are kind of irrelevant for the benefit of using APL to explore algorithms and solve problems ― often with the goal of re-implementing in another language later.
@Adám hm, what good is interop if you don't even have an interpreter you can use freely for any purpose? this may depend on the point of view. if i understand correctly, romilly's readers are likely to be young and unaffiliated with any company. learning apl would be a bad investment for them.
@ngn But you do have an interpreter that you can use freely for these purposes.
@ngn Why? If it gives them a free tool to explore algorithms, solve problems, and learn an alternative/better way to attack things, then what's not to like?
@RomillyCocking 'willimg'. interesting story of the origin of one-liners, is that how code golf 'is known to have been popular with earlier APL hackers'? (quote from en.wikipedia.org/wiki/Code_golf, but unsourced)
@user41805 Unsourced there, but my father told me the same (before Wikipedia existed), and he was there to witness (well, and participate in) it!
At IBM (he worked at their development labs '68-'72), they used to announce things like "I'll give a coffee to whomever can reduce this code by one character" (or whatever unit they used).
@RomillyCocking argument is the most efficient way to get to the truth, they say. as long as we don't make it too personal, i think we all should seek confrontation, not avoid it. it would be rather boring if we all pretended to agree all the time. i hope none of my messages was offensive.
@Adám It also has a transcript in case you prefer that format. I found it pretty insightful, and finding out the historic reasons for "why are one-liners a thing" felt similar just now :)
I have a question about Dyalog parsing rules. Assume you have an ambivalent two-arg operator: OP. The left operator arg being a function, and the right can be either a function or a value. How is the following parsed? 1+OP+3?
Is it parsed as an operator taking two functions, + and +?
Or is it assuming that +3 is the right value argument?
I give up on operators, APL's pseudo-infix is too ambiguous. I need a stack-based language with scalar extension. And static typing would be interesting too.
@Wezl The syntax isn't enough. One has to understand the memory model, pointer behaviour, etc. I have a large amount of experience with it and understand it very well. But, even I wouldn't want to write a large C program again.
But I wouldn't call C a "bad" language. I't woefully inadequate for most projects, but it dpes exactly what it sets out to do. In contrast there are very valid arguments why C++ is a quantifiably "bad" language.
@rak1507 It's hard to take a single feature of C++ and point to it and say "see? this is what makes it bad". It's the mix of all the things it does. It's the collection of seemingly nice features, all just shoved into the language with little thought of the whole. This is why you have several generations of C++, where for each generation, C++ programmers say "oh, you shouldn't use that feature, do it this way instead"
How many different ways do they have to track references to instances again?
But each and every (well, almost) feature of C++ sounds really good on its own.
Your messages were not at all offensive. I was feeling very vulnerable for reasons nothing to do with this community. No longer an issue, as a major personal problem has just been resolved.
I think I'd say 'discussion' rather than argument, but I agree that open and friendly discussion (which I see a lot of here) is a great way to get closer to the truth.
@ngn I think the problem with breaking changes is that the biggest users (who also own shares in the main APL vendor) have many senior execs who would love a reason to force their employees to move away from APL. Of course they would be mad to do so, but I've seen bigger companies try to do madder things.
For instance I know of one world famous investment bank that has tried to rewrite the software that makes most of its money, in order to move from Smalltalk to Java for 'religious' reasons. They have spent many millions over decades and falled three times. They are about to try a fourth,
@Adam I feel very welcome here, and I have got much benefit (and encouragement) from the feedback.
@RomillyCocking IIRC, the biggest customer(s) even bought in and arranged leadership for the explicit purpose of facilitating migration of their own things away from APL. But that new leadership instead managed to turn the ship around and made the customer abandon (some of) their migration plans.
@RomillyCocking i believe language design should be driven by other factors than executive politics
and, given the situation you describe, what does dyalog have to gain from not fixing old mistakes? maybe a bit of time, but also more reason for those companies to move away from apl
@ngn So do I but while we remain in a primarily capitalist economy the piper needs paying, and the payer calls the tune. Don;t get me wrong - I know that open source languages can prosper, but all those I know of have been supported by funders who had a lot of influence over their direction.
@RGS The draft blog post is going to get an update tomorrow morning- corrections and some new text. I'll be using some ANN examples, based on my blog posts of 2016 and also talking about Hopfield networks which are interesting and very well suited to implementation in APL.
@ngn I think if you/we want to develop an open source APL from the ground up you need someone with deep pockets.
My strategy is to try to help Dyalog to the point where they felt they could afford to fund that. I trust the people at Dyalog. They have been friends for many years. I have never known them do anything remotely dishonest and they have always been completely open with me. That's rare in business.
PS I can tell you what I'd love: a FOSS APL with syntax and semantics inspired by both APL and Smalltalk.
@Dogbert I like treating everything as an object (even arrays). I like the environment to include soirce code for all of the language. (In some cases that source will say 'call this bottom-level primitive here' and the byte-code compiler will need to inline that, but that's OK). I like the way that Smalltalk handles multiple method arguments. I like the fact that you can break everything by redefining + on integers :)
@RomillyCocking the lazy-pure-functional-language community managed to unite and created something as amazing as haskell. the dominant impl is free to use for any purpose (still funded development, but free for the user). tbh, i don't understand why the array community should be so different.
@Adám Smalltalk is not the only language where you can do very naughty things :)
And I like the fact that Smalltalk lets you dig in to the internal structure of objects, and create proxy objects, and (cos I am vain) the fact that there is a little code of mine in the (now defunct) Digital Smalltalk product.
@RomillyCocking "trust" - yes, for most things. "remotely dishonest" - i hate to say this, but if it matters, i can point to at least 2 examples from recent competitions :)
@RomillyCocking I'm familiar with Ruby which borrows a lot from Smalltalk and lets you redefine + on integers. I'll check out the "multiple method arguments" thing!
Are there any blog posts or anything floating around with suggestions from users on improving (or changing) the syntax and semantics of APL? any interesting derived languages?
ALso Smalltalk supports three kinds of message: unary (suffix), binary (infix) and keyword.
So you can say stinrg length (unart ,essgage sent to a string), 2 + 3 (binary cos it's a special cjharacter, but redefinable in your code if you want) or like my bucket example.
Also if you send a message to an object that belongs to a class that doesn't understand it, the system sends the message doesNotUndertand: message to it. If your class reimplements that method, you can do anyhing you like.
It also opens the opportuninty for the follwing error message which I once saw:
<moon-child> k has certainly deviated the most from trad. apl. (Where j and bqn, as well as more modern apl, are more of a direct evolution.) There are also a bunch of less prominent derivatives: xs, klong, o, nial, a couple of others I forgeet
@EliasMårtenson How about making the header part of the code? This will allow you to create anonymous and/or inline functions with a header. Also, do you not want to be able to specify the return variable name? E.g. ƒ foo {(r0 r1 r2)←(a0 a1 a2)∇(b0 b1) → a1+a1+a2+b0+b1} or how about foo←{(r0 r1 r2)←(a0 a1 a2)∇(b0 b1) → a1+a1+a2+b0+b1} ? If the return variable isn't specified, just use the last statement's value. The terse form would be equivalent to the header ⍺∇⍵ →
I seem to recall you complained about something, but I can't remember what. As we are currently reviewing the site content for the 2021 round, it is a good time to fix issues.
@Adám [0] the 2017 coding challenge (not: student competition, not: optima codegolf challenge). i was the sole winner. dyalog violated 2 of their their own rules to put someone else on first place too. i was complaining for months, but was just told to wait some more. (@Adám see your email, you were on cc then, i bumped the thread)
@ngn You did not, and have afaik never gotten negative points for lack of comments. You did, however, miss out on potential extra points you could have gotten for comments. And to be fair, we did state that one should comment the code and that the code should be clear.
@ngn the first isn't really "dishonest" since it did say that comments are judged (not whether they're positive or negative points, but providing a checklist of how precisely to win would be kinda stupid imo)
@rak1507 oh so that is it. I was finding it weird that ngn was the winner in every single month, there being months for which ngn was the only 1st place, and then someone else also gets the same score as ngn.
Week 9: as I wrote a year ago, "I've assumed that character scalars are not acceptable results - The Rules mention only character *vectors*. If scalars are ok, please don't count the leading glyph in the expression for 9."
Week 25: The first item of ⎕SD is not necessarily 25 - for example in the terminal I'm typing in right now ⎕SD is 39 151. The Rules say: "Expressions must not rely on local conditions, for example, the operating system, the exact date/time or specific input from the user."
Right, I see that. Having Jonas Stensiö as a shared winner seems like a mistake. It being dishonest is a stretch.
And lack of clarity in a problem spec is a well-known issue in these parts, of course. It was a rather informal competition, without prizes, and I guess Fiona just drowned in more important matters. Again, "dishonest"? I wouldn't call it that. Nobody tried to trick or deceive anyone here.
@Adám did you read the whole thread i bumped for you? i asked about these rules in advance, and reminded about correcting the mistake 3 times after the competition was over
@Adám so, is anyone going to correct the score now?
@ngn Even if we don't correct mistakes in time for it to matter, it is hardly "dishonest", is it? What's next? Is ignoring reports about software bugs also "dishonest"?
@ngn No, we're stretched thin, and unfortunately have to prioritise sometimes. Some of us even have lives outside work.
@ngn I'll ask Fiona if she can find the old data, or else, at least remove Stensiö as that seems to be a mistake. Would also like us to announce it on instagooglefacetweet, or issue a press release?
@ngn I'll be proud if I ever manage to manipulate you. Still, isn't that objectively more correct than "remotely dishonest" - i hate to say this, but if it matters, i can point to at least 2 examples from recent competitions when all you have is an example of your repeated (very valid, imo) requests remaining unaddressed after you were told:
> Sorry Nick; I got back today but have a squillion emails and other jobs to work through. I'll get back to you on this later this week.
@Adám i asked a question about the rules. i got a clarification and abided by it. at the end of the year i find that the exact same rules have been violated. complaint - no reply. reminder: "later this week". next week: "Not forgotten but bumped down the list". it's been 3 years since :) now what would you have thought in this situation?
> Not forgotten but bumped down the list due to having my laptop die and needing to get a new machine up and running. I will come back to you but it will be a while. Sorry.
@ngn Absolutely. I forget a lot, to the point where I've considered giving up working because of it. But in this case, I did apparently ask Fiona, and assumed she'd take care of it. I do not believe that you honestly think Fiona was playing tricks on you. You know her well enough for that. And what did she or Dyalog stand to gain, anyway? It isn't as if we've otherwise tried to hurt you, is it?
i forget a lot too. i can understand how in rare cases someone can forget twice. maybe even three times is possible. but four.. ? what would you assume is more likely to have happened?
@ngn If it was me, oh dear, I can forget the same thing, every day, for months. Despite daily reminders.
And it being mostly a single email thread… I just hit "unread" when it is something I need to do about, and then it sinks into the pile of unread emails.
