Vyxal

lyxal

00:04

@user merge the scala native pr when you're ready

hyper-neutrino

00:17

vyxapedia has been totally rewritten; it now has all of the elements populated from elements.yml - please let me know if you see any bugs or errors in it

6

!!/run ⟨⟨1|2|3⟩|⟨4|5|6⟩|⟨7|8|9⟩⟩ λ∑;

Vyxal Bot

@hyper-neutrino ⟨ 12 | 15 | 18 ⟩

hyper-neutrino

is this how it's supposed to behave? I thought λ∑; would push a function object

or do they get automatically executed if they are TOS at the end

!!/run ⟨⟨1|2|3⟩|⟨4|5|6⟩|⟨7|8|9⟩⟩ λ∑; 1 W

Vyxal Bot

@hyper-neutrino ⟨ ⟨ ⟨ 1 | 2 | 3 ⟩ | ⟨ 4 | 5 | 6 ⟩ | ⟨ 7 | 8 | 9 ⟩ ⟩ |

hyper-neutrino

is the hanging | at the end cuz the function can be stringified or why is that

emanresu A

Functions do weird stuff

IIRC they're executed when printed?

@hyper-neutrino Love the new theme!

hyper-neutrino

00:23

@emanresuA that would explain why , and then pushing more stuff later prints the sum instead of a function object

also glad you like it!

emanresu A

What happened to element search?

hyper-neutrino

I removed it because the only point in the old search was that you could filter by various things

but the vyxal element list is not in a nice format for that and doesn't have keywords, and I am too lazy to write them in

if elements.yml gets proper arity formatting (rn half of them are inconsistently expressed strings) and keywords I might be able to add them back

it's easier for Jelly cuz the language hasn't changed in many years :P

lyxal

@hyper-neutrino this

@hyper-neutrino smells like it's an invisible error

staging.ato.pxeger.com/…

Basically, don't wrap functions in lists at end of execution

Because printing functions executes them

@hyper-neutrino v3 will be nicer for things like this

user

00:57

@hyper-neutrino Nice! Could you make it so in the menu, the current page is bolded instead of home?

Vyxal Bot

ysthakur merged PR #1588 (Vyxal/Vyxal) (Vyxal:v3-native → Vyxal:version-3): Scala Native support

ysthakur deleted branch Vyxal/v3-native

user

Could it automatically use a light theme if that's someone's system setting?

I kinda miss the old one, it looked so soothing

The red makes me feel like I'm in danger

lyxal

I am not in danger, user, I am the danger. A guy opens vyxapedia and you think that of me? No, I am the one who code golfs.

@hyper-neutrino that's because not everything takes a single arity number of things

some things can take 1/2 things based on types, or can take a variable number of things, or operate on the whole stack or not take anything or take a certain number + a variable amount

Vyxal Bot

01:15

Lyxal created branch Vyxal/add-keywords-to-v2

lyxal

01:27

@hyper-neutrino what do you think about the keywords in ^?

Obviously it's not finished, but just as a general first impression for searchability

Is the format suitable?

user

There seems to be some repetition with the comment keywords

"comment out until newline" is repeated twice for #

Also, if you're using fuzzy searching, "comment until newline" probably isn't needed since comment out until newline already includes that

And the multiline comment keywords seem to have a lot of the single line keywords, although that may be intentional

lyxal

@user copilot

@user also copilot

user

lol

lyxal

02:26

125 elements now have keywords

That's almost half way

This might be done in time for 2.19.1 tonight

Ah nope forgot there's all the digraphs that need doing

There's at least 298 of them

hyper-neutrino

@lyxal right, but so long as the format is somewhat standardized it can be searchable

unless it is and I am just dumb

@lyxal oh this looks pretty good - I would separate them though so like ["not", "logical negate"] could be ["not", "logical", "negate"]

lyxal

@hyper-neutrino it is

hyper-neutrino

oh also another thing I forgot, tags (so like, logical not gets the "logical" and "non-vectorizing" tags)

github.com/Vyxal/vyxapedia-legacy/blob/main/src/data/…

for reference this is a sample of what the old vyxapedia format was like

how many elements are there?

lyxal

At least 554

hyper-neutrino

559. hmm... that's quite a few, I was thinking maybe I could just go through them all myself to make sure they're standardized in a format that works for me but that's a bit too many

alright I'll write up a spec for what I want the format to be and we can discuss and modify it and then maybe make that a project for the future (I assume we're prioritizing v3 element impl)

lyxal

02:38

@hyper-neutrino We are

v3 has keywords baked into the implementation because literate mode plans

@user oh yeah literate mode

That needs a bit of infrastructure planning

Probably should get that working first

hyper-neutrino

oh okay

cool

actually this doesn't need a spec. the ID is not needed cuz that was used for routing in the path but rn the routes are just the element symbols themselves

I kind of want to request separation between elements and structures because I still think they're different enough to warrant being termed differently

otherwise name, arity, keywords for searching, tags for filtering and info, a tagline for the table/list, and a full description would be nice

user

@lyxal New lexer for that, right? :|

lyxal

@user no

Same lexer just extended to recognise words

Transliterate the words to symbols and lex

user

@hyper-neutrino Perhaps a boolean field structure that's false by default?

lyxal

@hyper-neutrino what tags should there be?

Because vectorisation is already an attribute rather than a tag

@hyper-neutrino the only things not already there are the tags and keywords

@hyper-neutrino no spaces in keywords?

user

02:45

By the way, nice library for searching: Fuse.js

@lyxal Maybe it's so someone can select "logical" and get all the logical operators?

Actually, no, that'd work even if you had a "logical negate" keyword with spaces

lyxal

@user I mean so I know how to write the keywords

It's more of a yes or no question rather than a why question :p

user

Well I wanna make it a why question :P

lyxal

03:11

boolean
number
string
literal
constant
syntax
list
stack
split
remove

some of the tags I have thusfar

@hyper-neutrino what do you think about the tags?

Are they what you're thinking of tag wise?

lyxal

03:57

just realised the number/string/list tags are useless

because almost everything works on numbers/strings/lists

because type overloads

lyxal

04:20

Tags:

- base
- bitwise
- boolean
- constant
- digraph
- function
- io
- lambda
- list
- literal
- number
- prime
- register
- stack
- string
- vectorise

Vyxal Bot

Lyxal deleted branch Vyxal/add-keywords-to-v2

Lyxal created branch Vyxal/prep-v2-for-vyxapedia

lyxal

I've decided to start again doing things a little differently

mostly because I want to automate it all

"but lyxal, how are you going to automate it all? copilot doesn't have an API!"

raw codex does

+ it's free

hyper-neutrino

@user I would be okay with that

lyxal

better idea

structure instead of element

hyper-neutrino

@lyxal not sure, but the jelly site has arguments, arithmetic, array, base, bitwise, boolean, combinator, combinatorics, complex, constant, hyper, literal, math, matrix, misc, nth-link, number, optimization, prime, random, register, separator, stdio, string, and vectorize

@lyxal that is what I was originally thinking

although it would be more convenient to have symbol be the field and type: "structure" | "element" | "modifier"

emanresu A

04:27

Isn't almost everything string, number and list?

hyper-neutrino

the tags refer to what type of stuff they do

@lyxal this looks like a pretty good set of tags

some could be added to it from the jelly list but having too many isn't good either

lyxal

time to figure out a) how to handle all this yaml with python and b) what prompt to use for keyword and tag gen

hyper-neutrino

@lyxal oh do we have a short and long description separately?

lyxal

ye

description for short, overloads for long

hyper-neutrino

oh

description IMO is still too long for the table and too short for a full expanded description though lol

also for keywords, it's okay to have spaces but then if someone puts two spaces in their search it won't match the keyword anymore

lyxal

04:32

@hyper-neutrino guess I'll add that to the list of things to automate with codex :p

lyxal

04:53

@hyper-neutrino on second thought, I have added a few more

purely because there wasn't that many codex could decide

tags like number, string, list can be added based on overloads

hyper-neutrino

alright

lyxal

wow it actually works

I'm actually good at prompt engineering

# From the following YAML, create a list of tags that might apply to the function from its name, description and overload tags.
# The possible tags to choose from are:

# arithmetic - for functions that look like they perform mathematical functions
# base-conversion - for functions that look like they perform base conversion
# bitwise - for functions that look like they perform bitwise operations
# boolean - for functions that look like they return boolean results
# combinatorics - for functions that look like they involve combinatorics

and you give it the name field, description field and overloads

on experimentation, it's a little jank with some things

lyxal

05:22

well

let's just say the ai isn't as good at this as I thought

that, or the categories aren't general enough

@lyxal gonna stick with this list

lyxal

06:06

Okay so here's the story

The whole auto generation thing didn't work out so keywording and tagging is being done manually

It's on the github.com/Vyxal/Vyxal/tree/prep-v2-for-vyxapedia branch

There's a text file with all the tags that can be used

Number, string, list and function are related to type overloads

The rest are hopefully self explanatory

The keywords field goes under the description field

And the tags field under that

Keywords is a list of strings. One word per string unless unavoidable

Keywords are what people might use in a sentence to find a particular overload

Protip use copilot for keywords and edit if need be

Tags is also a list of strings

Each string is a tag from the tags file

Copilot can do the tags but you need to manually check the list it generates

Secondly, there's some renaming happening too

If something starts a structure, it's structure, not element

If something is syntax, it's syntax not element

Keywords of structures is an empty list

As are the tags

You can have keywords for syntax though

But the tags are empty list for syntax

I've done up to C

So if you need inspiration on how to format things, elements lambda through C serve as examples

If you've got any questions don't hesitate to ask

And no, this doesn't break the txt file, test and md file generators

The keyboard description generator will have to be changed a little, but not too much

Also, the vectorisation field of elements serves as the vectorise tag, so no need to include it in the tags list

hyper-neutrino

06:59

oh nice

looks good

syntax refers to stuff like arrays and strings I presume?

lyxal

Yes

hyper-neutrino

alright

lyxal

But also decimal points, branches and that sort of thing

hyper-neutrino

ah, okay

also, the tests being in the elements.yml is very convenient cuz it gives me free examples :P

lyxal

Lol

Finally a big brain reason for having them there :p

Alas that won't be possible in v3

Because there's different test types lol

user

07:22

:(

Actually, we could have a tests.yaml and a test suite that reads tests from that yaml file at runtime

And then the more complex tests would continue being defined the same way as they are now

I'll make a branch for that tomorrow

lyxal

@user yeah you don't need a new lexer at all. Just split on words, things wrapped in curly brackets turned into lambdas, things in square brackets turned into a list and then look up each word in a word to symbol hash map (auto gen'd from the implementations object before hand) and sbcs lex

Vyxal Bot

12:16

Version 2.19.1

5

lyxal

12:38

All the monographs have been keyworded and tagged

now for the 300 digraphs

k digraphs done

now for the math, string, list and misc digraphs

digraph-math
digraph-string
digraph-list
digraph-misc

those are the 4 tag types for digraphs

why

why the hell

are there 2

views from core.tcl-lang.org

when there is a 99% chance

that there is no link to vyxal there at all

~~which one of you idiots visited the tcl lang site with cookies on and then directly visited vyxal?~~

AndrovT

13:48

CMC: Sort a list of complex numbers by the real part

in particular it has to work for ⟨ 1.5 | 1°1 ⟩

I feel like working with complex numbers is unnecessarily hard because of the overloads of complex number built-ins

Vyxal Bot

16:19

hyper-neutrino forked Vyxal/vyxapedia into hyper-neutrino/jht-site

user

21:18

@AndrovT There's a sort by element/structure, right?

user

21:31

@lyxal I'm making a tests.yaml so that hyper can get examples, but should it be a tests.json instead?

Or some other format?

AndrovT

@user The problem is that ⌊ is real part for complex numbers but floor for reals so µ⌊; doesn't work

user

Oh, that sucks

AndrovT

!!/run ⟨ 1.5 | 1°1 ⟩ µ⌊;

Vyxal Bot

@AndrovT ⟨ 3/2 | 1 + I ⟩

AndrovT

!!/run ⟨ 1.5 | 1°1 ⟩ µ⌊;

Vyxal Bot

21:39

@AndrovT ⟨ 3/2 | 1 + I ⟩

user

@lyxal Looks like there needs to be a different way to get the real/imaginary components of complex numbers

lyxal

21:50

@user yaml should be fine I think

It can always be converted to json easily with code

user

22:15

👍

lyxal

Y'all the Programming Language Design proposal is 3 upvotes away from leaving the first phase. Come distribute your votes to the last 3 9-scores

!!/hyperping

Vyxal Bot

@AaroneousMiller @Allxy @Ausername @BgilMidol @DialFrost @Ginger @Milk @Niko @NumberBasher @PyGamer0 @Seggan @Steffan @Wansen @astonearachnid @emanresuA @hyper-neutrino @mathcat @pxeger @rues @tybocopperkettle @user ^

lyxal

(cc @lesobrod @97.100.97.109 @AndrovT)

user

22:58

Now it's gotten to the commit phase, so come commit

If you want, you can use this link that includes me as the referrer so I can get...fake internet points?

emanresu A

Imagine being able to create an account

So apparently I do have an account, but it's unlinked

Transcript for