Geal/nom

Rust parser combinator framework

Jordan Mack

@jordanmack

Any good sources of additional Nom 5 examples?

I'm brand new to it, and I'm struggling a bit with syntax. More examples beyond the color examples would help.

Jordan Mack

@jordanmack

How would I take all a-z until ":", and error if there is something not a-z before the ":"?

Simon Thornington

@sthornington

Jordan Mack

@jordanmack

Thanks! 👍

Wait a sec, it disappeared.

Simon Thornington

@sthornington

https://pastebin.com/zB0iS1NG

because I am not good at gitter

would be curious to hear better ways of doing this

(I've been using nom for like 3 days so this was practice)

working through https://bodil.lol/parser-combinators/ was helpful for me to understand the whole paradigm in Rust

I think you could make a more efficient one_of for ascii A-Z by implementing your own FindToken on 'A' <= c <= 'Z' for instance

or maybe something like that already exists

Jordan Mack

@jordanmack

I actually need a-zA-Z0-9_, which is equiv to \w in regex.

I'm trying to learn Nom as well, so I want to do it the Nom way.

Simon Thornington

@sthornington

maybe implement a FindToken using is_alphanumeric ?

Jordan Mack

@jordanmack

Yes, I think I can tie that with alt or something to get the _ in there.

Simon Thornington

@sthornington

bingo https://docs.rs/nom/5.0.0/nom/character/complete/fn.alphanumeric0.html

that simplifies it to:

fn paz(input: &str) -> IResult<&str, &str>
{
    terminated(alphanumeric0,
               nom::character::complete::char(':'))(input)
}

#[test]
fn test_az() {
    assert_eq!(paz("ABC:"), Ok(("", "ABC")));
}

Alberto

@0X1A

Hi folks. I want to use the error propagation operator in a parser function. The function using the propagator returns a custom error I've defined (we'll call it MyError). The parser function returns the usual IResult. I've tried looking at other examples for nom 5 but I can't quite figure out how this can be done. I can obviously define the From trait for MyError -> nom::Err<_> but then I would have to return an ErrorKind, where the ErrorKind would be totally unrelated since errors returned by MyError have essentially nothing to do with parsing.

peckpeck

@peckpeck

the recommended way is to define a global error type and implement ParseError on it
then all my parsers return an IResult<I,O,MyError>
this is done automatically since I use nom5 functions and use the results directly as a MyError

Alberto

@0X1A

@peckpeck See, that's what I thought as well but then I seem to get :

`?` couldn't convert the error to `nom::Err<MyError>`

When ever I use the prop error with a nom parser function. And I can't exactly impl From for Err<_> for Err<MyError> since I can't impl for a type outside my crate

peckpeck

@peckpeck

did you implement ParseError ?

Alberto

@0X1A

Yep, have impl<'a> ParseError<MyInputType<&'a str>> for MyError in my crate

? does an implicit conversion using the From trait, so I wonder if this simply isn't possible without defining From, which of course I can't do since Err is nom's

peckpeck

@peckpeck

i didn't have to define it in my code

and that would imply defining it for nom::Err not for your Error type

on what did you use the ? ?

Alberto

@0X1A

preceded(_, _)(input)?

peckpeck

@peckpeck

are the _ things nom parsers or ones you wrote ?

Alberto

@0X1A

Oh fml, parser I was using there was still using generic IResult

Alberto

@0X1A

Okay, so that resolved that. I can now use IResult<I, O, MyError>, but now I'm back at not being able to use the ? propagator for my functions that return non-IResults. I'm guessing the only way to fix that would be to convert it manually because of IResult...

peckpeck

@peckpeck

for my part i strictly separated parser code from other code so i have few places where I had to manually translate from my parser errors into my application error

anything i used within parser either returned a MyError or I made it so using an equivalent of cut or i simply matched it for error handling

Evgeny Fomin

@fominok

Hello! I'm new to Nom and have a basic question, I've described it here: https://stackoverflow.com/questions/57608442/rust-nom-many-and-end-of-input. Thank you in advance!

peckpeck

@peckpeck

with nom 5 all macro are incomplete, if you want to use the complete version, your have to switch to functions

Evgeny Fomin

@fominok

@peckpeck is there any alternative to do_parse to deny some of the inputs?

peckpeck

@peckpeck

do_parse is not made to deny some input but to write sequence, the recommended alternative is now to write a parser with a sequence of let (i,xxx) = someParser(i)?;

I created a sequence! alternative in https://github.com/peckpeck/nomplus

Evgeny Fomin

@fominok

Both versions (with take_till and eof) and without macros work, thank you!

matrixbot

@matrixbot

bspeice Is there a good way to handle bit-tuple parsing? As an example, in IPv4, there's a 4-bit version number and 4-bit header size as the first byte. It's not hard to just mask off the upper 4 bits, but I was curious about trying to do it in Nom style.

Denis Lisov

@tanriol

There's the nom::bits module with bit-granularity parsers.

matrixbot

@matrixbot

bspeice Right, but the impression was that I needed to surround things with a nom::bits::bits function in order to actually use bits without resetting the alignment each time.

bspeice Specifically, I want something like tuple((take_bits(4), take_bits(4)))

matrixbot

@matrixbot

bspeice Currently I'm getting type errors with both let (_, (a, b)): (&[u8], (u8, u8)) = tuple((bits(take(4usize)), bits(take(4usize))))(buf).expect("Parse failed"); and let (_, (a, b)): (&[u8], (u8, u8)) = bits(tuple((take(4usize), take(4usize))))(buf).expect("Parse failed");.

bspeice (Noting that the take in question has been imported from the bits::complete)

Denis Lisov

@tanriol

bits(tuple(...)) should work. However, you may need to use some dark magic for type inference to work... for me it required bits::<_, _, (_, _), _, _>(tuple(...))

James Carl

@crazycarl

I like Nom. It's incredibly fast and very maintainable.
I think the documentation for the whitespace module hasn't been updated yet. I just know that in the source code, ws! is marked as deprecated.

Looking at the source... is sep the right function to use here? Is this function finished?

Where communities thrive

People

Repo info

Activity