Vyxal

user

00:01

@lyxal I've put globals back into interpreter.py in my PR, as you suggested, and the FizzBuzz test works now

Lemme know if you want me to modify it more or smth

lyxal

00:15

I'll have a look at it once I'm at my computer again

But in regards to github.com/orgs/Vyxal/teams/transpiler-design-team/discussions/…, what do you mean by deep vectorise?

Do you mean [1, 2, 3] + [4, 5, 6] -> [[5, 6, 7], [6, 7, 8], [7, 8, 9]]

Because if so, the existing vectorise function has an expl argument which makes it do the above

@user

user

@lyxal No, just normal addition ([5, 7, 9])

I wasn't thinking about changing the semantics of add, it was to make it easier to define add and shorten the code a bit while keeping it readable and easy to maintain

lyxal

@user it does that normally

user

Honestly, I don't understand vectorization well enough myself, so I didn't open a PR

@lyxal Oh lol

Then why the last four cases for generators and lists?

lyxal

Because they are special

And they need special handling

user

Okay, so vectorise can't handle them normally?

lyxal

00:21

Well they need to be Vectorised element-wise somehow

user

That's why I was proposing a deep_vectorisze function that can handle those special cases

I'll draw up a PR after the code cleanup and restructuring and stuff is done, hopefully I'll grok vectorization myself by then

lyxal

@user I mean I probably don't need those special cases

Because they all just have the same dictionary values

user

Oh

lyxal

Which means a simple .get(types, lambda:...) would fix it

Iirc the special cases are remnants of earlier versions

Earlier and cursed versions

user

Hmm

lyxal

00:27

Because unless there's some sort of vectorisation method I'm missing here, the current function should be relatively clear

lyxal

08:31

Mmm there's a lot of fun for me to have tonight

PR review, issue 147 and molding with generators

lyxal

09:39

Just a friendly reminder that for loops used to make you put the thing you were iterating over inside the for loop

Like if you wanted to do something 8 times, you'd write (8|...)

math

12:41

I have some questions about how Vyxal works. Can I ask it now?

lyxal

sure

I'm only rickrolling myself, so go ahead

math

So, the main file is interpreter, and the first thing it does is pass the code to the parser, right? What will the parser return?

lyxal

@math It returns a list of tokens

math

Can you give an example?

lyxal

in v2.4.x and lower, that was a list of Token objects - a custom class I had

in v2.5.x and higher, it's now a list of (number, dictionary/any)

@math sure

1 2 3W 4 5" v+

turns into

[(7, '1'), (7, '2'), (7, '3'), (0, 'W'), (7, '4'), (7, '5'), (0, '"'), (17, ('v', [(0, '+')]))]

₁ƛ₍₃₅kF½*∑∴

turns into

math

12:46

@lyxal What does the 7 stand for?

lyxal

@math number

[(0, '₁'), (5, {11: '₍₃₅kF½*∑∴'}), (0, 'M')]

@lyxal ^

NONE = 0
IF = 1
FOR = 2
WHILE = 3
FUNCTION = 4
LAMBDA = 5
STRING = 6
NUMBER = 7
MAP = 8
LIST = 9
VAR_GET = 10
VAR_SET = 11
FUNC_REF = 12
COMPRESSED_NUMBER = 13
COMPRESSED_STRING = 14
CHARACTER = 15
DICTIONARY_STRING = 16
MONAD_TRANSFORMER = 17
DYAD_TRANSFORMER = 18
TRIAD_TRANSFORMER = 19
PARA_APPLY = 20
LAMBDA_NEWLINE = 21
FILTER = 23
SORT = 24

those are all the structures

STRING = 1
NUMBER = 2
IF_TRUE = 3
IF_FALSE = 4
FOR_VAR = 5
FOR_BODY = 6
WHILE_COND = 7
WHILE_BODY = 8
FUNC_NAME = 9
FUNC_BODY = 10
LAMBDA_BODY = 11
LIST_ITEM = 12
LIST_ITEMS = 13
VAR_NAME = 14
LAMBDA_ARGS = 15
COM_NUM_VALUE = 16
COM_STR_VALUE = 17
LAMBDA_GROUP = 18

and those are all of the keys

math

oh that's what the variables in the parser file do

lyxal

correct

they are used for classification later on in the compilation process

math

why is the W in the first example a 0?

lyxal

because it is considered NONE

a standard built-in

it isn't a structure

and it isn't a multi-byte value

so it's structure is NONE

math

12:50

ah

so, what does the interpreter do with these tokens?

lyxal

for each token, it matches the first value (the structure type) and processes the second value (the structure data/value) accordingly

an example: to transpile NONE, you simply .get() the structure data from the command dictionary

math

that's clever

lyxal

to transpile a LAMBDA, you first have to include all the juicy lambda boilerplate in the compiled code, then the rest of the structure data

@math I've had almost three years to figure out the most optimal way ;)

math

and the utilities file?

lyxal

those are functions that aren't element helper functions

although to_ten and from_ten should probably be in the main file

utilities is mostly base conversion stuff

I think I decided to keep it separate because I figured that once I got it working, I'd keep it out of sight and out of mind

because base conversion is one of those things that takes me a bit longer to wrap my head around than the average code golfer

math

12:55

really nice, thanks

lyxal

I noticed you asked about splitting the digraphs into separate dictionaries

and the funny thing is that before 2.5, that's how it was

there was the main command dictionary, plus a dictionary for each digraph type

I figured I'd experiment with merging them into one dictionary in 2.5+ and the E branch

math

and last question: Why do you have a dictionary.py and a dictionary.txt?

you could just eval the txt file, right?

lyxal

@math heh, you sure know how to expose my laziness in keeping legacy files ;p (I'm totally fine with the questions btw - this comment is 100% light-hearted)

@math and originally I did

but then I yeeted it all into a python file

because that's what everyone else does

and it's probably better as .py

you could probably just delete the .txt and the dictconv.py files and nothing would change functionality wise

actually I'll do that myself

I'm working on implementing #147, so I'll remove those files in the same commit

math

oh okay, didn't notice that dictionary.py = dictionary.txt+dictconv.py

lyxal

@math in short: legacy files

@user speaking of legacy files: is reference_to_md.py still needed, or does the workflow automatically do it without relying on the file?

math

13:02

@lyxal bye then, I have a Matt Parker video to watch

lyxal

o/

math

14:12

Does Vyxal have commands with the arity 4 or 3?

Aaron Miller

@math There aren’t any arity 4 commands, but these are the arity 3 commands: ¢ V ŀ ẋ Ȧ Ḟ Ŀ Ṁ ∇ øV øṙ ¨M

math

@AaronMiller ok, thanks. also not arity > 4?

Aaron Miller

@math Right, at least none that I can find

math

@AaronMiller thanks

hyper-neutrino

14:34

Announcement: If y'all would like a Vyxal resources site similar to my JHT site, please let me know what sort of stuff you'd like and what categories of builtins there are (elements, modifiers, and syntax?) and I can give the format that data should be in to adapt the site over to Vyxal.

4

math

@hyper-neutrino your site's awesome

user

@lyxal It's still needed. The workflow does do it automatically, but it basically just runs that and then commits. idk bash and stuff well enough to do it inside the workflow itself. In any case, Python's just easier to maintain

math

@hyper-neutrino shall I make a hyper ping?

@Ausername @Underslash @Razetime @UnrelatedString @wasif @pxeger @AviFS ^^^^

user

@lyxal Sorry to ping you again, but could you please merge/reject my PR before more changes are made? It's very hard to resolve conflicts owing to how big interpreter.py is

hyper-neutrino

14:59

is splitting the files into more manageable components part of the cleanup

if not it might not hurt to try to do so

user

Yeah

The problem is that I can't do a math_builtin.py because builtins like ceiling also do stuff for string and aren't really mathy

I'll try an array_builtins.py maybe

obtw cumulative_reduce needs to be redone to be a bit more efficient

Wait, it's not even necessary, I think, there's a scanr_by_axis already

hyper-neutrino

15:15

oh yeah cuz overloading lol

you could make math funcs, string funcs, etc. and then have a unified overloading function that calls each one depending

so for example you could have a ceiling function in math

and a split on space function in str

and then a unified command function that type checks and calls ceiling for numbers and split-on-space for str

that might be more confusing tho so idk if that's actually a good idea

user

15:57

Yeah, that's a good idea

Would also avoid having strings inside commands.py

lyxal

22:44

@user I would have done so last night but it got too late after fixing a few bugs. Dw, I haven't forgotten about doing so

user

Danke

lyxal

23:27

Merged.

Underslash

23:42

Is there (and if there isn't, could it be added) a flag that truncates all array outputs to 100 elements or so?

Transcript for