The APL Orchard

nathan rogers

00:01

I don't know what that means

of course if you constantly run ⎕FIX on a file it'll get the lastest one

because that's what the file is now

but can I make edits arbitrarily in both directions?

in real time?

while the program is running?

ngn

@nathanrogers you seem to be coming from a "session" mindset

nathan rogers

not really, dyalog APL does

rak1507

don't worry, dyalog -script will be a thing by the next century!!!

nathan rogers

yeah... I wouldn't bet on it

ngn

@nathanrogers well, what happens to a running python program when you modify its source? nothing, right? why should apl be different

nathan rogers

00:05

ummm... because, like python, or slime, or geizer or cider, or whatever, I like to have my definitions in a place, play with them, modify them, run them again was rinse repeat

often times, in particular when dealing with large amounts of data, I want to cache that object in memory so that I can test operations on it without having to read in all that data, either over the network or off disk

Bubbler

Well, everyone has different workflow

but yeah, not having to load large data again is a good thing

nathan rogers

right, but the interactive model is the model as used by lispers/schemers/clojure...ers/aplers, and many oth-ers

Bubbler

It's probably one reason why Jupyter is so popular in the data science world

nathan rogers

its a useful model when you're not doing, say logic. I prefer if doing logic to write, but I prefer when transforming data to evaluate

when dealing with streams and networked events, often times the setup takes forever if the project is significant enough in size, and I really don't want to tear everything down and start over

its much easier just to have modifications read straight into the repl or in dyalog's case, the session

I personally can't really tolerate the "change 1 line: compile: execute" loop

ngn

so, the cost of setting things up is what keeps you in a long-running session

nathan rogers

00:10

time cost, yes

and interactivity

those are the 2 primary reasons I use repl's and the dyalog session to program

ngn

my argument against developing in the session is: it's too easy to get into some state for which you don't remember how you got there, and next time your restart the interpreter it will behave differently

nathan rogers

so don't declare a global ever

dzaima

@ngn I wish other langs had ways to mix both a REPL & file-editing. A file is the best place to program, and a REPL is thr best place to test/play around with it

nathan rogers

everything should be initialized by an entry point

everything should be independent

ngn

should

nathan rogers

00:13

when loading asdf modules and it doesn't load by itself, I create a new module

if I can't highlight an entire file and C-c C-c into slime, there's a dependency, and the part that can't compile needs to be elsewhere

and then there's quickload, which will load an entire project in dependency order

I get what you're saying, but I guess I have created borders around my programming style that don't get me to those points

perhaps its because I'm only ever working on problems in isolation. I'm writing class methods around encapsulating a large collection of data from an API. I define an instance of that class as global so I don't need to constantly load in the API data, at which point I write functions around that data

with that workflow I never really get to a place where the repl is in a state that I can't replicate

can't just go willy nilly delcaring variables everywhere

ngn

my workflow is: vim file.k and ./file.k (the latter being on a shortcut in vim). that's all i need.

nathan rogers

you neglect to mention the "try to comprehend why it crashed" and "figure out where it went wrong on your mental stack"

ngn

i don't work with big data. if i had to, i'd take a small sample so things start quickly.

nathan rogers

working with small samples is a good starting point, but often direct solutions are slow for unexpected reasons

rak1507

idk I really like working in the repl, I wish integration with files wasn't so painful

ngn

00:21

@nathanrogers printf() debugging works in any language :)

nathan rogers

that it does

rak1507

the easiest solution for me is to work out stuff in the repl and copy paste the solutions into a file or something

nathan rogers

and that's what I use when a repl isn't handy

rak1507

that's my most productive way of doing things

nathan rogers

but the repl is integrated printf around any arbitrary data

dzaima

00:22

@rak1507 +1

nathan rogers

which is why I coudln't understand the situation with tables earlier

ngn

@rak1507 well, it's almost the same as mine except i don't have to copy paste

nathan rogers

@rak1507 I skip the copy past and learned to use load/reload in whatever fashion is available for a language

I can't stand coding directly in the repl because its too easy to make typos

so slime lets me send code to the repl properly formed, or I edit the .py file and use importlib.reload to reload the source for a module

or us ]link.create # 'mydir' -source="both"

rak1507

@nathanrogers that's the thing, if you code in the repl and make a typo, it's literally 0 seconds to change

whereas in a file, it's slightly more than 0 seconds

nathan rogers

not so

ngn

00:24

@rak1507 rougly 0

nathan rogers

in python you have to re evaluate each line of a def

and in slime too

rak1507

oh well if you mean literal code execution, yeah

personally I just never write code that isn't instant

dzaima

@ngn but you lose the ability to iteratively define test inputs and temporarily once call things as needed

nathan rogers

like i need to load a large defun that's obnoxious to type directly into sbcl even with rlwrap or something

ngn

@dzaima add something, run, add another thing, run, etc. why wouldn't it work "iteratively"?

rak1507

00:25

it's just ever so slightly clunkier

Bubbler

It's all personal tastes (and personal experience), please don't fight about it

nathan rogers

@rak1507 idk, dyalog, slime, or vim-slime with insert repl are the best coding environs I've had the pleasure of working with.

ngn

@rak1507 was that a reply to me?

rak1507

@ngn yes

@nathanrogers yes, I'm agreeing with you :)

dzaima

@ngn in a REPL, you add, run, get the result. In a file, you have to clean up afterwards, removing things you might want later (or have a gazillion commented test temp lines)

ngn

00:27

@rak1507 no clunkier than typing ]some.ucmd.i.cant.even.remember

nathan rogers

@rak1507 vim slime means I have vertical split pane, one with vim one with a repl in the other tmux pane, and I "send" properly formed bodies of code over to be evaluated

Bubbler

@dzaima You have a separate test script for that.

rak1507

or say you have something like
foo←slow thing
bar←typofunction foo

in the repl you can evaluate foo once and then start working on it

nathan rogers

yes, but then like you said you have to copy paste it over

rak1507

@ngn I agree, the current state of link/other things is basically unusable for me IMO

but working in the repl + copy paste into notepad + save in files works fine for me

nathan rogers

00:28

but if you're already editing it in vim presently, then you already have it properly formed, you don't need to copy it anywhere

dzaima

@Bubbler how does that change anything, other than now there being another tab to worry about?

nathan rogers

I don't think you are getting the workflow I'm describing. Have you used slime?

ngn

@nathanrogers whom are you asking?

nathan rogers

rak

rak1507

I'm confused now :D I think we're all disagreeing with each other slightly or something

nathan rogers

00:32

pane of text with vim | pane with repl running

I can put my cursor over text, that is already saved in a file, and it "sends" to the repl pane and evaluates

Bubbler

6 mins ago, by Bubbler

It's all personal tastes (and personal experience), please don't fight about it

rak1507

yeah that sounds pretty good to me

nathan rogers

so it eliminates the distinction from code file and session

rak1507

@Bubbler I wouldn't say it's a fight, just a discussion, for me it's certainly interesting to get other people's points of view

I've already learnt about slime, which sounds good, so for me this has been a net positive

nathan rogers

your code is already saved to a text file, nothing to copy paste

and typos are no issue, because your cursor is still in the text file

ngn

00:34

@Bubbler is right. let's fight about something more divisive.

nathan rogers

@rak1507 vimslime with tmux allows you to use this process for anything that has a repl

@ngn like ÷0

that's pretty... divisive

dzaima

as an example, today i was playing around with simhash in dzaima/APL. There I had a single function, sometimes tweaked a bit, a couple persistent test strings, and a whole lot of temporary test strings. Outputting just the hash of a single string, comparing similarity of previous strings, and defining more strings to play around with, were all randomly interleaved, in parallel to editing the function itself

rak1507

@ngn ⎕IO←0 or ELSE!!!!

nathan rogers

÷0 is...
(•_•)
( •_•)>⌐■-■
(⌐■_■)
divisive

@rak1507 what does :Else have to do with it

leave control structures out of it

rak1507

lol

nathan rogers

00:36

how about proper ternary's in APL

rak1507

I mentioned APL to one of my friends, and then a few days later they asked me for help with some code... I was so excited until they showed me something like
∇foo
:For i :In ⍳10
...
:EndFor
∇

nathan rogers

megaOof.mp3

Bubbler

Ooh, Python

rak1507

@nathanrogers unneeded if programming in an 'array oriented' way whatever that means

Bubbler

or C++ or Rust or...

dzaima

00:38

@dzaima the worst part about a REPL is scrolling to the line you want to edit. The function is edited rarely, so having it in a separate tab (or auto-loaded file) would be helpful to remove most cases of editing older lines

rak1507

yep the whole thing was basically a python translation

@dzaima )ed ftw!

nathan rogers

@rak1507 "unneeded" but so are for and while and switch, and whatever else

but its nice to have

rak1507

Disagree if all control structures were removed from the face of the earth it would be a much happier place

dzaima

@rak1507 unfortunately, not in dzaima/APL on the commandline

rak1507

:for -> ¨, :while -> ⍣, :switch -> ⊃ ⌷ [] ...

nathan rogers

00:39

a← bool(arg) ? 1 : 0 is nicer than a←{bool ⍵: 1 : 0} arg

Bubbler

Time to introduce {...}⎕IfThenElse{...} boolean

nathan rogers

⍣ is "until" c'mon

rak1507

@nathanrogers and a←arg≠0 is nicer than both

Wezl

@rak1507 please explain libraries to me

user

@rak1507 Unfortunately, those are sometimes necessary (unless you're talking just about APL)

nathan rogers

00:40

@rak1507 I'm using a trivial example... obviously, but I'm saying that conditional assignment

rak1507

@Wezl well at least APL has some stuff... K on the other hand appears (and this is just an appearance, it could be wrong) to have nothing

Wezl

@nathanrogers clojurians

nathan rogers

is nice to have

rak1507

@user just APL

@nathanrogers and any non trivial example will not be idiomatic APL

ngn

@rak1507 k proper has ffi, of course

nathan rogers

00:40

I get it

rak1507

If you're doing manual if else checks on individual scalars, you're probably doing something horribly wrong

nathan rogers

but sometimes, conditional assignment is just nice to have

like ¨ is nice to have

rak1507

@ngn yeah

nathan rogers

so is ⍣ and ⊂

like, you know early APL had no such thing as nested arrays right?

but they're pretty nice to have

rak1507

I feel like control structures/classes/etc are so far outside the language

if there's like a venn diagram or something, core APL is in the middle, stuff added that's not really core to the language but makes things a bit nicer is just outside, and then :For is in another galaxy

nathan rogers

00:43

I'm not arguing for them

I write exclusively dfns unless some kind of mutation requires that it be a tradfn

dzaima

@rak1507 they're pretty non-array-y, but sometimes things aren't all array-y

nathan rogers

right

rak1507

yeah

imo it's better to force APL to do things it's really not meant to than to introduce new ways of letting it do them 'nicely'

nathan rogers

like the .NET zip library for APL2XL

github.com/the-carlisle-group/APL2XL/tree/master/APLSource/Main/…

rak1507

if that requires 47 ¨s so be it

nathan rogers

00:44

or true error handling

writing to the filesystem

sockets

lots of things that aren't array-y that you can't do anything meaningful without

rak1507

true

Wezl

@nathanrogers mutation is never required >:| (hopefully leads onto another divisive subject)

rak1507

lol

nathan rogers

if you look at the rest of that project, not a tradfn to be found outside of the "build" directory @rak1507

ngn

@Wezl +1 (non-mutating incrementation)

rak1507

00:47

the day someone makes a pure functional array language with a Maybe type for error handling I will give them all my money

nathan rogers

@Wezl well, until linux supports file monads, and stream monads, and the real world stops changing its own state, there will always be a need for mutation somehwere. it should just be isolated, like in github.com/the-carlisle-group/APL2XL/tree/master/APLSource/Main/…

all the true mutation happens inside the build directory

everything else is functional

ngn

@rak1507 haskell?

Wezl

exactly

rak1507

haskell with a good array processing library would be the only language I would ever use again

Wezl

"array language" -- not objective

nathan rogers

00:48

@rak1507 gotta learn scheme

rak1507

tried it, not a big fan, it's ok though

Wezl

except for I will roll out my own functional language and push haskell out of the window

nathan rogers

define "tried"

Wezl

@Wezl actually, I hope to push some imperative language out instead

rak1507

@nathanrogers played with it for about 10 minutes lol

nathan rogers

00:49

yeah, def not enough

if you're still in the "wahwahaw parenthesis wahawhwa" stage, you haven't "tried it"

rak1507

nah I don't mind the parentheses

Wezl

I'm really loving scheme right now as I translate some CL code to it, but my code hasn't started running yet so maybe my contentedness is getting ahead of me

nathan rogers

I genuinely honestly think that "array language" or not is not the thing. "A generalized universal notation for computing" I think is the thing, I just think that since arrays are a parallel to linear algebra, it makes defining a generalized universal notation for computing on arrays more achievable since there is already a lot of notation for and research around the subject.

Bubbler

Except that it is not well researched as you might think

nathan rogers

I don't think APL takes it far enough. File system interaction, stream interaction, server spawning and configuration, all of this should be notationalized

@Bubbler you're assuming how much I think its researched

Wezl

00:52

unrelated:

nathan rogers

I'm just saying that there hasn't really been a lot of research around notation for other forms of computing

Wezl

CMC: do my conjugation homework for me

rak1507

lol

nathan rogers

not until you pay CMC a conjugal visit

rak1507

what language? the only language I have any reasonable chance of being able to do it in is latin

ngn

00:53

@Wezl apl, 1 byte: + ("complex conjugate")

Wezl

@rak1507 italian

nathan rogers

italianicized

Wezl

I prefer scheme to APL, and APL to italian

rak1507

lol

nathan rogers

I prefer APL to life, and scheme to anything that isn't APL

I just wish I had scheme powers in APL

rak1507

00:54

what do you like in scheme that you can't do in APL?

nathan rogers

then I wouldn't have to worry about features added or not in the core language

Wezl

@rak1507 Imna write an awk script some time to count my lols

nathan rogers

@Wezl for the lulz?

Wezl

chat.stackexchange.com/search?q=lol&room=52405

Dyalog APL

<shapr> is the bot written in APL?

rak1507

00:56

@Wezl I have a 3rd of them

@DyalogAPL no

Dyalog APL

<Wezl> <rak1507> no

<shapr> aw

<Wezl> IIRC it's written in raku

dzaima

github.com/moon-chilled/SEProxy

Dyalog APL

<shapr> Are there network libraries in APL?

Bubbler

@nathanrogers Dyalog has .NET power instead, which may or may not work for you

Dyalog APL

<Wezl> shapr: <dzaima> github.com/moon-chilled/SEProxy

<shapr> Actually, I want to know the recommended setup for APL in emacs

<shapr> that's nifty

<shapr> My first Haskell project was an IRC bot

Bubbler

00:59

So in theory you can use anything that is supported in .NET

Wezl

@rak1507 macros, closures, everything functional

Bubbler

How clunky it is is another issue

nathan rogers

@DyalogAPL DYALOG APL in emacs is tough, because the only package I can find is the gnu-apl-mode which supports the incorrect DYALOG unicode letters

I had a rough time debugging something because gnu-apl-mode uses the wrong ⋄

Wezl

btw shapr can't hear you

Dyalog APL

<shapr> Any APL + emacs users here?

nathan rogers

01:00

@Bubbler no that's not what I'm talking about

rak1507

@Wezl well sure if you're missing scheme-y features in APL not much you can do, but it's like saying you miss for loops in haskell

nathan rogers

I can't add new glyphs, I can't extend syntax, etc

there are no "lisp macros" in dyalog

Dyalog APL

<Wezl> shapr: <nathan-rogers> DYALOG APL in emacs is tough, because the only package I can find is the gnu-apl-mode which supports the incorrect DYALOG unicode letters

<shapr> hm, maybe I should stick with J then

<Wezl> I'd suggest going to the SE chat at chat.stackexchange.com/rooms/52405/the-apl-orchard

nathan rogers

I can generate code that generates code well enough but it isn't the same as modifying the AST IRT

the way I see it, I should at least, at build time have the ability to define my own glyphs, if it can't be supported in the language

2

the only way to experiment with notation is to define my own language

Bubbler

@nathanrogers Well then APL is simply not the right language

Wezl

01:02

@nathanrogers definitely, and please people take notes

nathan rogers

I think APL's core IS the right language, but provided a few "programming features" would make APL the generalized notation for computing that will take us to the stars

Bubbler

though I do agree with defining own glyphs

rak1507

I like your optimism

dzaima

@nathanrogers then forward compatibility becomes a problem - the language becomes literally unabls to define any new syntax

nathan rogers

for me it isn't optimism. its dire frustration

Wezl

01:04

@dzaima the new extensions should be available as libraries then

nathan rogers

@dzaima I'm of the clojure mindset. A strictly controlled core language, with support of hygienic macros that allow for user extension, and user definition of domain specific notation

Bubbler

define "hygienic macros"

rak1507

I don't think that's particularly suited for the things APL should be suited for

Wezl

@Wezl bc now all the special glyphs are built-in

nathan rogers

Hygienic macro

Hygienic macros are macros whose expansion is guaranteed not to cause the accidental capture of identifiers. They are a feature of programming languages such as Scheme, Dylan, Rust, and Julia. The general problem of accidental capture was well known within the Lisp community prior to the introduction of hygienic macros. Macro writers would use language features that would generate unique identifiers (e.g., gensym) or use obfuscated identifiers in order to avoid the problem. Hygienic macros are a programmatic solution to the capture problem that is integrated into the macro expander itself. The...

scheme vs common lisp macros

dzaima

01:05

@Wezl I mean things like adding a new function form, new matched brackets, multiline comments, etc

ngn

@nathanrogers scheme is the original

Wezl

@dzaima that's also fixed in lisp so just switch to scheme basically ::), with code as data and reader macros

nathan rogers

yes I know... I've been talking about scheme and lisp for the past, idk 2 hours

ngn

@nathanrogers you mentioned clojure, but i think credit should go to scheme, even if that makes it more than 2 hours :)

rak1507

poor shapr

nathan rogers

01:08

I mentioned clojure because the implementor believes as I said in that comment, a tightly controlled core language that won't change much over time, but allows the language to be extended by users and libraries

dzaima

@Wezl so you'd need to... import multiline comments?!?

nathan rogers

the clojure philosophy is stated in almost direct opposition to the CL mindset, which is "kitchen sink"

Bubbler

@dzaima Actually multiple languages do that (even outside the lisp world).

Wezl

@dzaima you'd have I-beams otherwise, so sure

nathan rogers

@ngn I was referring to @dzaima s comment that it detrimentally effects the languages future compatibility, and I agree, it should

@ngn as is the case with: clojure

Bubbler

01:10

I have never fully understood how the AST-level macros actually work, so I have no idea how it can fit into APL

(or, more specifically, the syntax for defining such macros)

nathan rogers

@Bubbler that's why you need to learn a lisp or a scheme. APL dialects and Lisp dialects are at the top of the language hierarchy of expressivity

and both have strengths over the othre

but for me, they share the throne of language design

ngn

and k is in the middle

dzaima

@Bubbler and people complain about java being verbose?

rak1507

ooh @shapr 'Languages: english, swedish', how well do you know APL again?

nathan rogers

and both need to learn from the other. DFNS and lexical scope learned from Lisp, but there's still so much more

ngn

01:12

kparc.com/lisp.txt

nathan rogers

@ngn I'm lumping K with "apl dialects"

Wezl

ngn

@nathanrogers i think it's gone too far from apl already

nathan rogers

but I get that K shares a lot of concepts with lisp

Wezl

^^^ scheme extending the syntax

ngn

01:13

@nathanrogers see the link

nathan rogers

I saw

I think K deviates too much from the concept of notation, sacrificing notation for terseness

I think terseness isn't the correct goal, but brevity is

rak1507

what's the difference?

nathan rogers

terseness: sparing in the use of words; abrupt.
brevity: concise and exact use of words in writing or speech.

brevity is the soul of wit
tersity is the soul of shortness for shortness' sake

one is elegant, the other is short

short can be elegant, but not all things that can be described shortly are elegant

brevity is the balance of clarity and shortness

which leads to elegance

ngn

@nathanrogers how does it "deviate from the concept of notation"?

nathan rogers

there is a different kind of beauty in tersity, but the pursuit of the shortest often obfuscates meaning

K deviates from the concept of notation by hyper-overloading symbols to the point of absurdity, due to the restriction that only ascii characters may be used

and it also seems to emphasize functional abstraction over data abstraction, but that might be my lack of experience with it

Wezl

01:19

I believe in making DSLs that can do things tersely, not a condensed language that can do everything tersely-- up to a point

nathan rogers

@Wezl I'll upload my GPL, GLFW parser

there's a dsl there, but it isn't terse, its brief

Bubbler

I... am starting to think APL can have user-defined macros

Wezl

e.g. April has APL for when you need it, but it's built on a more powerful language IMO

Bubbler

I'm feeling more and more need to implement my own APL to experiment with various ideas

Wezl

@nathanrogers GLFW??? that's terse language that I don't understand

ngn

01:21

@nathanrogers overloads - that's a common complaint, but being able to type on any machine is so nice..

nathan rogers

@Bubbler what I'd really like to see is a language level config file that contains a map, where you can map any "symbol" in the symbol sense to any procedure

@Wezl GLFW as in opengl

I have an openGL parser that loads in opengl bindings

Wezl

ah ok

nathan rogers

remind me later if you're interested no time now

and I need to clean it up, haven't touched it for a solid 2 years

need to format and comment, and document a bit

but I'm surprised at its cleanliness, and it was written by me :P

but there's also some obvious lack of familiarity with dyalog in there as well

so I want to refactor too before I get it up in github

but the DSL is super tight and clean for parsing the various GLFW/GLEW files

Wezl

something in me is addicted to rewriting code. I got half of something working, and a month later I haven't worked on that at all but I've reimplimented it in another language

nathan rogers

@Wezl this is a portion of the DSL

agltypes is a relational array to gltypes mapping library and c types to APL types of ⎕NA

rak1507

01:28

includes←∨/∊

nathan rogers

it is a dsl after all

and you'll notice there's a lot of operators so these combine very nicely

without a lot of trains everywhere

rak1507

I see a lot of ¨s :(

nathan rogers

there's a lot of nested vectors

like I said this was a project I wrote a long time ago

rak1507

ah

fair enough

nathan rogers

but the actual parsing sections are quite clean

Wezl

01:30

@rak1507 if you pressure the author of this enough, you might get what you want :)

nathan rogers

@rak1507 this reads super nice

rak1507

@Wezl doesn't look very array-y to me

also no examples

Wezl

@rak1507 but it's early in development so what I'm saying is you can change that

rak1507

idk what 'based on graph reduction' means but it sounds ... interesting

Wezl

btw warning: promotion that repo was put up like 2 weeks ago and it has two stars already, you should use codeberg.org

(no affiliation)

rak1507

01:34

advantage over gh?

Wezl

open source

dark theme

(I know gh has one, but it started with a dark theme so it's objectively better :)

rak1507

gh has a dark theme now

lol fair enough

Wezl

also xpqz's there codeberg.org/xpqz

(I think you need an account to use dark theme though)

rak1507

'⍝⎕∨∊∩⊤ ○⌈ ⊂○⎕∊ ⍳∩ ⍝⍴⌊', nice

Wezl

btw thanks for the Outspoken badge ppl

:( there's no gold chat badge

or maybe it's better if there's less SE manipulation?

Razetime

01:53

@Wezl gratz

@nathanrogers i see you use material theme as well

@Bubbler and so it begins

nathan rogers

@Razetime well if someone would make a better one

@rak1507 so... I've not found an effective strategy for handling strings without ¨

especially when those strings are already nicely read in as lines from a file, as a nested character vector

rak1507

strings are a pain I agree

nathan rogers

if you know of any nice strats for strings that don't involve ¨, I'd love to learn

rak1507

depends what you're doing with them I guess

nathan rogers

I find that many of my problems take the form of nested vectors, and ¨, but that's probably a limitation in my skillset

@rak1507 well in the example images, for example, I'm searching for lines that include or begin with certain text, and then formatting them according to the "type" of lines based on the pattern matching

if there's an array operation for that, that'd be slick

but usually, what winds up happening is I might identify lines where that's true, compress those lines, then I parse each one

rak1507

02:02

you could probably do something with the rank operator

not sure if it'd be any faster or better than ¨ though

nathan rogers

right, but I've been informed that for performance rank and each are pretty similar

Bubbler

but then you need to make the text rectangular, which may be an issue

nathan rogers

that too

lots of extra padding

Marshall

@nathanrogers I don't think array style is ever going to be as easy as nesting here, but you might find my markdown processor to be interesting reading. I did it in an array style as much as was convenient.

nathan rogers

looks cool

but I can't read bqn

Marshall

02:05

The BQN compiler is much more strict about array style, but it's solving different problems.

nathan rogers

i can fry it though

Marshall

@nathanrogers Just the comments have a fair amount of good information.

And there's a BQN-Dyalog dictionary that might make reading faster.

Razetime

@nathanrogers there is no better colour palette

nathan rogers

03:17

@Marshall @Adám That's actually a great question. I find that most of my solutions begin with data in the form of nested lists, whether reading or writing to files, generating or parsing text. I tend to find an intermediate representation that I know how to work with, or some kind of intermediate expression that gives me a launching point for the next step in a procedure, but I rarely find myself doing array operations unless they're simple arrays, and I rarely have simple arrays.

What are some ways that I can begin thinking differently about representation of data so I can begin approaching solutions differently. I feel like so much of my hang ups when writing stem from getting caught in the intermediate representation or some intermediate transformation when there is a bigger picture representation or transformation that does what I want more idiomatically.

Marshall

@nathanrogers I think my partition representations article is pretty useful. It ties into the sparse (i.e. Indices) representation but I don't know of a nice writeup on that. An approach I often take is to identify a representation and try to figure out various concrete objects it might represent. Usually there are quite a few.

Bob Smith's original paper on partitioning, included in this pdf, is also a very nice resource because it shows how to do a few things on partitioned data with completely flat arrays.

Another good tool for data with repetition is decomposing into unique values and indices into them, which can't remove nesting but can reduce its impact. It does make operations that only care about which values are equal, but not their actual values, flat.

Russtopia

03:36

Hi all, am I crazy or does the Pascal's Triangle example on Wikipedia have syntax errors? I tried it on both GNU APL and Dyalog, both reject it. See

https://en.wikipedia.org/wiki/Talk:APL_syntax_and_symbols#Solving_Puzzles_:_Pascal's_Triangle_example_has_errors.

Marshall

@Russtopia How about if you run it in Dyalog with ⎕ML←3? That ⊃ looks suspicious.

Russtopia

@Marshall Aha, that works in Dyalog, at least.

I almost think GNU APL is hitting a bug on it, since the parse line it spits out has a garbage char at the end.

Marshall

I don't know GNU well enough to say. Given the use of a dfn, it seems pretty likely it was written for Dyalog, so I think it would be better to use ↑ rather than ⊃ as ⎕ML←1 is the default.

Russtopia

OK thanks. Yeah I read that GNU APl's support of {} is a bit limited (no guards or multiple statements ...}

Marshall

@Russtopia They're more of an alternate syntax for tradfns than an attempt to support dfns.

Russtopia

03:48

@Marshall thanks. I'll re-read the ⎕ML help, good to know

Dyalog APL

04:48

<abrudz[m]> shapr: See apl.wiki/Text_editors for two utils that should make using Dyalog easy with emacs.

Adám

09:36

@Wezl Do your other fonts also support APL? If so, any way to get them?

Transcript for