@blueberry wait…people put all identifiers under one token?

you can…do that?

yeah??

were you planning to make a token for every possible identifier

…yeah, kinda

that would be unpleasant, seeing as there's an infinite number of them

oh

like var identifiers

that…changes things a bit

oh, i'm using identifier to mean "variable name"

I thought you were talking about keywords like for, while, etc.

that's generally what they're called

OH NO LYXAL'S HERE

I'M ABOUT TO GET ROASTED TO OBLIVION

yeah you make a token for each keyword

makes sense

@Someone not for the reason you'd think :p

I mean, everyone compares Java vs C#, but what about Rust vs Swift?

@lyxal uh oh

make a token for each keyword, make one token that contains a string for identifier

that's terrifying

I need y'all to keep a keen eye on the users in room list

@DannyuNDos why compare rust and swift

Star spammer is back

do they even have similar goals

uh oh

star spammer

If anyone suspicious is in the user list, and stars start flooding in, ping me and tell me who it is

@blueberry now that I know the vernacular, everything makes sense

@blueberry Because they both are better C++.

The Nineteenth Byte got hit as did a few other rooms

uh oh

@DannyuNDos doesn't mean you should compare them though

do we have an automod system for preventing the flood from actually happening?

prolog and ocaml are both better then c++ in their domains, comparing them is nonsense

@Someone no

that's unfortunate

womp womp

why do people do that btw?

like, what do they stand to gain from it?

shrug

they enjoy it

No clue

I don't get it

why do some people do stuff that harms everyone, including themselves?

doesn't really harm them

well, it kinda does

@blueberry Not just better than C++, but better C++. Unlike C++, they successfully incorporated OOP into C.

I use C++ for its fancy containers and stream syntax :shrug:

@DannyuNDos saying rust is OOP is interesting

Aw.

rust is wack imo

very broad definition of OOP

@Someone why so

idk

it's very strange in terms of syntax

not really

at least compared to what I know/similar langs

and cargo/crates also feel different from other module/package solutions

there we go

Few months ago, I learnt that Unix and Linux do I/O by exposing devices in /dev, but what the heck about Windows?

@Someone they're not that different from, say, python packages, they're just significantly better organized

better tooling too

I/O is one major reason that prevents me from designing my own lang, so

not to mention the fact that it was a lang specifically to suit one person's tastes, and then a whole cult formed around it for some reason, obsessed with the idea that it could replace all other code forever (or maybe just c(++))

> a lang specifically to suit one person's tastes

... like C++?

idk about cpp

it's not as cultish as crablang is

in any case, just like every other time this kicks off for some reason, there's probably not going to be anything constructive come of it

As a wise man once said, C++ is bloated.

again, I just use it because stream syntax and std::string go brrr

and also because I don't like Rust as much

That said, in Haskell, String is defined as a typedef for [Char]. It was a mistake, IMO.

Should've been newtype String = String [Char], or something like that.

wait

what

how does that work???

I mean, just because of that mistake, the Show typeclass has showList :: [a] -> ShowS. That's a bad interface design.

what is ShowS?

ShowS is a typedef for String -> String. It was introduced so the showing functions would compose conveniently.

huh

interesting

what's the best way to actually process the AST once finished? a member function named something like execute?

That is a reasonable way to build an interpreter, yes

You could also use a visitor or build some sort of tree-walker

visitor?

Often there will be multiple layers of transformation from the initial parse tree to the final result, so it might not be "execute" straight away, for example if you are going to build a compiler and you need to lower the semantics of the input language to generate machine code. For an initial interpreter, whatever is the easiest way for you to get something to happen within the structure will be good

@Someone Visitor pattern

@DannyuNDos massive mistake

imo it should have just been an abstract builtin

not even a newtype

linked lists of chars are hilariously inefficient

so if I had an AST node type for floats, and one for strings, and I wanted to have a type/union/class/whatever that encompasses all AST node types that return a value via, say, a member function, how would I say "idk the return type of the member function, but it does return a thing"?

and would I use a template? a union? some other random thing?

You need to decide on what the run-time semantics of the language are to establish that

I'm specifically using this for an output statement, among other things

Tagged unions are one way of implementing dynamically-typed languages that you could use

But if you have objects, static types, etc, you might go a different way

If you only have the two types of value, a tagged union seems easiest to go forward with

well

so far I have two

idk if that'll change, since I have no standards for the thing yet

I just noticed that there is no reason to have the I monad and the O monad. They must be imbedded into the IO monad anyway. Dang it.

yep

This is why both std::istream and std::ostream are std::ios, I guess?

oh in c++ i have no clue

in a lazy language though, they both have to be IO anyway

> getUnsafePerformIO :: IO (IO a -> a)

This is why Haskell doesn't expose main like Python does.

if I'm separating chunks of code for my program, but the things I'm importing to main aren't intended to be used anywhere else, can I just import it directly without a header file?

@DannyuNDos define "exposing main"

what, the runtime setup before your main is run? yeah of course

*Cough*

@blueberry so if multiple modular chunks need each others' content, how would I handle that? just import everything everywhere and use header guards to clean up my mess?

However it is that you usually structure your programs will also work fine here

@blueberry The main function's crimes have been exposed. To jail it goes! 10 years.

Is that the real Peter Norvig who just joined the site? He doesn't have any other SE accounts.

I don't know how we could tell at this stage

yeah, there's no indicators to verify/disprove the authenticity of the account

I believe that it is not an impersonator

I was more asking out of surprise than disbelief

Yeah, hopefully he sticks around

I feel ashamed of myself for not knowing who Peter Norvig is

I don't really see why you should

fair

Quite a number of stack-based golflangs have word order of OSV, but there are so few natlangs with that word order.

I wonder speakers of such natlangs have stack-based thoughts?

Sapir-Whorf intensifies.

Another load of stars

good luck finding a OSV native speaker

oh ok warao might work

there's 33k native speakers

who is starring everything?

Dunno

this person? not sure, just in room list

@Redz pretty sure you need 20 rep to star

Maybe

I don't actually know

@DannyuNDos ASL is OSV

Slavic languages have a fluid sentence structure, and OSV is often used (myself included) if you want to emphasize the object (although IME OVS is more often used for that)

@DannyuNDos I'm inclined to say stack-based syntax isn't OSV, it's O1 O2 V and the verbs are imperative.

I'm under situation where tail call optimization is a savior.

or some sort of search

linked list

let me guess

Huh?

that's the most common situations where I would expect TCO to be a lifesaver