#parser-generator #context-free-grammar #parser-combinator #syntax-tree #parser #parse-input #ast-parser

rusty_parser

A Generic compile-time Parser generator and pattern matching library written in Rust

14 releases (3 stable)

1.1.0 Jul 6, 2024
0.9.5 Jun 4, 2024

#18 in Parser tooling

Download history 108/week @ 2024-07-30 4/week @ 2024-08-27 9/week @ 2024-09-10 5/week @ 2024-09-17 10/week @ 2024-09-24 9/week @ 2024-10-01

1,281 downloads per month
Used in mini-c-parser

MIT license

240KB
6K SLoC

RustyParser

A Generic compile-time Parser generator and Pattern Matching Library written in Rust

RustyParser provides a set of basic parsers, combinators, and parser-generating functions.

This library is designed to work with general iterators, but some functionalities are limited to specific iterators.

Example

Sample Code

// import rusty_parser
use rusty_parser as rp;

// useful trait member functions
use rp::IntoParser;

#[test]
fn example1() {
    // define pattern
    // digit: [0-9]
    // this will match one digit, and returns (char,), the character it parsed
    let digit_parser = rp::range('0'..='9');

    // define pattern
    // num: digit+
    // this will match one or more digits, and returns (Vec<char>,), the character it parsed
    let num_parser = digit_parser.repeat(1..);

    // map the output
    // Vec<char>  -->  i32
    let num_parser = num_parser.map(|digits:Vec<char>| -> i32 {
        let mut num = 0;
        for ch in digits {
            num = num * 10 + (ch as i32 - '0' as i32);
        }
        num
    });

    // parse input iterator with given pattern, and return the result
    let res = rp::parse(&num_parser, "123456hello_world".chars());

    // res contains the result of parsing
    assert_eq!(res.output.unwrap(), (123456,));

    // res.it: iterator after parsing
    // here, '123456' is parsed, so the rest is "hello_world"
    assert_eq!(res.it.collect::<String>(), "hello_world");
}

Structures

RustyParser provides a set of basic parsers, combinators, and parser-generating functions. Those generated parsers are used to parse the input string, and return the extracted data.

fn parse(pattern:&Pattern, it:It) -> ParseResult<(Parsed Output of Pattern), It>;
fn match_pattern(pattern:&Pattern, it:It) -> ParseResult<(), It>;
  • parse(...) takes a Pattern Object and iterator of input string, then returns ParseResult<Self::Output, It>.

  • match_pattern(...) is used when you only want to check if the pattern is matched or not, without extracting data. For some parsers, like repeat, it is expensive to call parse(...) to get the output since it invokes Vec::push inside.

ParseResult is a struct representing the result of parsing.

pub struct ParseResult<Output, It>
where
    Output: Tuple,
    It: Iterator + Clone,
{
    // the output; extracted data
    // 'None' means parsing failed
    pub output: Option<Output>,

    // iterator after parsing
    // if parsing failed, this will be the same as the input iterator
    pub it: It,
}

Note

  • Since the parse(...) internally clones the iterator, the iterator must be cheaply clonable.
  • Output must be Tuple, including (). If you want to return a single value, use (Value,).

Parsers Overview

Basic(Leaf) Parsers

Parser Description Output
one, one_by Match one charactor (Iterator::Item,)
range Match one charactor in the range (Iterator::Item,)
str, str_by, slice, slice_by Match multiple charactors ()
string, string_by, vec, vec_by Match multiple charactors ()
check Check one charactor with closure (T,)
any Match any charactor (Iterator::Item,)
DictBTree, DictHashMap Trie Dictionary T
DynBoxChars, DynBoxSlice, DynBoxSliceCopied Dynamic Parser that can take any parser with same Output T

Combinators

Combinator Description Output
seq Sequence of parsers ( *<Output of A>, *<Output of B> ... )(Tuple Concatenated )
or Or combinator Output of the all parsers
map Map the output of the parser (T,)
repeat Repeat the parser multiple times (Vec<Output of Self>,)
optional Success whether the pattern is matched or not ( Option<Output of Self>, )
optional_or, or_else Success whether the pattern is matched or not Output of Self
not Match for Pattern1 to success and Pattern2 to fail Output of Self
reduce_left, reduce_right Reduce the output of the parser Output of Self
reduce_with, reduce_right_with Reduce the output of the parser with initial value Init

Others

Parser Description Output
constant Always succeed, and return the constant value ()
end Success if it reached to the end of input ()
fail Always fail ()
void Ignore the output of the parser ()
output Change Parser's Output to (output,) (T,)
string, vec Captures the matched range into String or Vec<T> (String,) or (Vec<Iterator::Item>,)
not_consume Check if the pattern is matched or not, without consuming the input Output of Self

or refer to docs.rs

Detailed explanation and examples of Parsers

Basic Parsers

one, one_by: consumes one character if it is equal to c.

let parser = one( c: CharType )

let a_parser = one('a')
let a_parser = 'a'.into_parser()

let a_parser = one_by('a', |value:char, ch:&char| value.to_ascii_lowercase() == *ch );

Output: (Iterator::Item,)

range: consumes one character if it is in the range r.

let parser = range( r: impl std::ops::RangeBounds )

let digit_parser = range( '0'..='9' )
let digit_parser = ('0'..='9').into_parser()

Output: (Iterator::Item,)

str, str_by, slice, slice_by: consumes multiple characters if it is equal to s.

For borrowing-safety, the lifetime of str or slice must be 'static.

To use with other lifetime, you should use string() or vec() instead. Those functions will clone the items in String, Vec.

// must be 'static
let hello_parser = str("hello");
let hello_parser = "hello".into_parser();
let hello_parser = str_by("hello", |value:char, ch:char| value.to_ascii_lowercase() == ch );

let hello_parser = slice(&[104, 101, 108, 108, 111]);
let hello_parser = (&[104, 101, 108, 108, 111]).into_parser();
let hello_parser = slice_by(&[104, 101, 108, 108, 111], |value:i32, ch:&i32| value == *ch );

Output: ()

string, string_by, vec, vec_by: consumes multiple characters if it is equal to s.

This will copy all the characters into String or Vec, so lifetime belongs to the parser itself.

let hello_parser = string("hello".to_string());
let hello_parser = "hello".to_string().into_parser();
let hello_parser = string_by("hello".to_string(), |value:char, ch:char| value.to_ascii_lowercase() == ch );

let hello_parser = vec(vec![104, 101, 108, 108, 111]);
let hello_parser = (vec![104, 101, 108, 108, 111]).into_parser();
let hello_parser = vec_by(vec![104, 101, 108, 108, 111], |value:i32, ch:&i32| value == *ch );

Output: ()

check: check single character with a closure

The closure must be either of: Fn(Iterator::Item) -> Option<NewOutput> or Fn(Iterator::Item) -> bool.

let parser = check( |ch:char| if ch.is_alphabetic() { Some(ch) }else{ None } ); // returns Option<char> -> `(char,)` as output
let parser = check( |ch:char| ch.is_alphabetic() ); // returns bool -> `()` as output

If the closure returns Option<NewOutput>, the output will be (NewOutput,). If the closure returns bool, the output will be ().

any: Match any character.

let parser = any();

Output: (Iterator::Item,)

Dictionary: build Trie from a list of strings

// let mut parser = rp::DictBTree::new();
let mut parser = rp::DictHashMap::new();

parser.insert("hello".chars(), (1,));
parser.insert("hello_world".chars(), (2,));
parser.insert("world".chars(), (3,));

// this will match as long as possible
let res = rp::parse(&parser, "hello_world_abcdefg".chars());
assert_eq!(res.output.unwrap(), (2,));
// 'hello_world' is parsed, so the rest is "_abcdefg"
assert_eq!(res.it.collect::<String>(), "_abcdefg");

// match 'hello' only
let res = rp::parse(&parser, "hello_wo".chars());
assert_eq!(res.output.unwrap(), (1,));

Output: generic type you inserted

This will match as long as possible, regardless of the order of insertion.

There are two types of Dictionary: DictBTree and DictHashMap for Trie implementation. Both of them have their own Pros and Cons (the memory usage and time complexity of searching), so you can choose one of them.

Combinators

seq: sequence of parsers

// 'a', and then 'b'
let ab_parser = rp::seq!('a', 'b', 'c'); // IntoParser for char

let res = rp::parse(&ab_parser, "abcd".chars());
assert_eq!(res.output.unwrap(), ('a', 'b', 'c')); // Output is concatenated
assert_eq!(res.it.collect::<String>(), "d");

Output: ( A0, A1, ..., B0, B1, ..., C0, C1, ... ) where (A0, A1, ...) are the output of the first parser, and (B0, B1, ...), (C0, C1, ...) are the output of the following parsers.

or: or combinator

// 'a' or 'b'
let ab_parser = rp::or!('a', 'b'); // IntoParser for char

// 'a' is matched
let res = rp::parse(&ab_parser, "abcd".chars());
assert_eq!(res.output.unwrap(), ('a',)); // Output of 'a'
assert_eq!(res.it.clone().collect::<String>(), "bcd");

// continue parsing from the rest
// 'a' is not matched, but 'b' is matched
let res = rp::parse(&ab_parser, res.it);
assert_eq!(res.output.unwrap(), ('b',));
assert_eq!(res.it.clone().collect::<String>(), "cd");

// continue parsing from the rest
// 'a' is not matched, 'b' is not matched; failed
let res = rp::parse(&ab_parser, res.it);
assert_eq!(res.output, None);
assert_eq!(res.it.clone().collect::<String>(), "cd");

Output: Output of the all parsers. Note that the output of all parsers must be the same type.

map: map the output of the parser

Parser's Output(Tuple) will be unpacked and passed to the closure. The value returned from the closure will be new Output.

// map the output
// <Output of 'a'> -> i32
let int_parser = 'a'.map(|ch| -> i32 { ch as i32 - 'a' as i32 }); // IntoParser for char

let res = rp::parse(&int_parser, "abcd".chars());
assert_eq!(res.output.unwrap(), (0,));
assert_eq!(res.it.collect::<String>(), "bcd");

Output: (T,) where T is return type of the closure. The value v returned from the closure will be wrapped into (v,).

repeat: repeat the parser multiple times

// repeat 'a' 3 to 5 times
let multiple_a_parser = 'a'.repeat(3..=5); // IntoParser for char
let res = rp::parse(&multiple_a_parser, "aaaabcd".chars());

// four 'a' is parsed
assert_eq!(res.output.unwrap(), (vec!['a', 'a', 'a', 'a',],));
assert_eq!(res.it.collect::<String>(), "bcd");

Output:

  • if Output of the repeated parser is (), then Output is ()
  • if Output of the repeated parser is (T,), then Output is Vec<T>
  • otherwise, Vec< Output of Self >

optional, optional_or, or_else: success whether the pattern is matched or not

let a_optional_parser = 'a'.optional(); // (Option<char>,)

let res = rp::parse(&a_optional_parser, "abcd".chars()); // success
assert_eq!(res.output.unwrap(), (Some('a'),));

let res = rp::parse(&a_optional_parser, "bcd".chars()); // success, but 'a' is not matched
assert_eq!(res.output.unwrap(), (None,));

// if 'a' failed, return 'x'
let a_optional_or = 'a'.optional_or(('x',)); // (char,)

let res = rp::parse(&a_optional_or, "bcd".chars());
assert_eq!(res.output.unwrap(), ('x',));

// if 'a' failed, evaluate the closure
let a_or_else = 'a'.or_else( || 'x' ); // (char,)
assert_eq!(res.output.unwrap(), ('x',));

Output for optional:

  • if Output of the origin parser is (T0,), (Option<T0>,)
  • otherwise, ( Option<Output of Self>, )

Output for optional_or, or_else:

  • <Output of Self>.

Note

  • The passed value's type to optional_or and or_else must match with the Output of Self
  • For single-value-output ( which's output is (T,) ), passing either T or (T,) is permitted.

not: match for Pattern1 to success and Pattern2 to fail

// all digit but not 4
let digit_parser_except_4 = ('0'..='9').not('4');

let res = rp::parse(&digit_parser_except_4, "3".chars());
assert_eq!(res.output.unwrap(), ('3',));

let res = rp::parse(&digit_parser_except_4, "4".chars());
assert_eq!(res.output, None);

Output: Output of Self

reduce_left: reduce the output of the parser

With given input string self rhs rhs rhs rhs ... and the reducer f, the output will be calculated as f( f( f(self,rhs), rhs ), rhs ), ...

Note

  • The signature of the reducer must be Fn(A0, A1, A2, ..., B0, B1, B2, ...) -> ( A0, A1, A2 ... ). Where (A0, A1, A2, ...) are the output of the first parser, and (B0, B1, B2, ...) are the output of the following parser.

  • For single-value-output ( which's output is (T,) ), returning either T or (T,) is permitted.

let digit_parser = ('0'..='9').into_parser().map(|val: char| -> i32 { val as i32 - '0' as i32 });
let reduced_left = digit_parser.reduce_left(digit_parser, |lhs, rhs| lhs * 10 + rhs);
let res = rp::parse( &reduced_left, "123456abcd".chars() );
assert_eq!(res.output.unwrap(), (123456,));
assert_eq!(res.it.collect::<String>(), "abcd");

Output: Output of Self

reduce_right: reduce the output of the parser

With given input string lhs lhs lhs lhs ... self and the reducer f, the output will be calculated as f(lhs, f(lhs, f(lhs, f( ... f(lhs,self)))

Note

  • The signature of the reducer must be Fn(A0, A1, A2, ..., B0, B1, B2, ...) -> ( B0, B1, B2 ... ). Where (A0, A1, A2, ...) are the output of the first parser, and (B0, B1, B2, ...) are the output of the following parser.

  • For single-value-output ( which's output is (T,) ), returning either T or (T,) is permitted.

let digit_parser =
    ('0'..='9').into_parser().map(|val: char| -> i32 { val as i32 - '0' as i32 });
let alphabet_parser =
    ('a'..='z').into_parser().map(|val: char| -> i32 { val as i32 - 'a' as i32 });
let reduced_right =
    alphabet_parser.reduce_right(digit_parser, |lhs: i32, rhs: i32| -> i32 { rhs * 10 + lhs });

let res = rp::parse(&reduced_right, "123456dcba".chars());
assert_eq!(res.output.unwrap(), (3654321,));
assert_eq!(res.it.collect::<String>(), "cba");

Output: Output of Self

reduce_with: reduce the output of the parser with initial value

With given input string self self self ... and the reducer f, the output will be calculated as f( f( f(init,self), self), self), ...

The signature of the reducer must be Fn(Init, A0, A1, A2, ...) -> Init. Where (A0, A1, A2, ...) are the output of Self.

Output: Init

let digit_parser =
    ('0'..='9').into_parser().map(|val: char| -> i32 { val as i32 - '0' as i32 });
let number_parser =
    digit_parser.reduce_with(0, |acc, rhs| acc * 10 + rhs);

let res = rp::parse(&number_parser, "123456abc".chars());
assert_eq!(res.output.unwrap(), (123456,));
assert_eq!(res.it.collect::<String>(), "abc");

reduce_right_with: reduce the output of the parser with initial value

With given input string self self self ... and the reducer f, the output will be calculated as f(self, f(self, f(self, f( ... f(self,init)))

The signature of the reducer must be Fn(A0, A1, A2, ..., Init) -> Init. Where (A0, A1, A2, ...) are the output of Self.

Output: Init

Example

let digit_parser =
    ('0'..='9').into_parser().map(|val: char| -> i32 { val as i32 - '0' as i32 });
let number_rev_parser =
    digit_parser.reduce_right_with(0, |lhs, acc| acc * 10 + lhs);

let res = rp::parse(&number_rev_parser, "123456abc".chars());
assert_eq!(res.output.unwrap(), (654321,));
assert_eq!(res.it.collect::<String>(), "abc");

Others

Trivial, but useful parsers

constant: This parser will always succeed, and return the constant value

let parser = rp::constant( (1, 2, 3) );

Output: the Tuple value you provided

end: success if it reached to the end of input

let end_parser = rp::end();

Output: ()

fail: This parser will always fail

let parser = rp::fail();

Output: ()

void: ignore the output of the parser

Force the output to be (). It internally calls match_pattern(...) instead of parse(...). This is useful when you only want to check if the pattern is matched or not. For more information, see match_pattern(...) above.

let expensive_parser = 'a'.map(|_| -> i32 {
    // some expensive operations for data extracting...
    panic!("This should not be called");
});
let expensive_parser = expensive_parser.void();

// ignore the output of parser
// this internally calls 'match_pattern(...)' instead of 'parse(...)'
let res = rp::parse(&expensive_parser, "abcd".chars());
assert_eq!(res.output.unwrap(), ());
assert_eq!(res.it.collect::<String>(), "bcd");

Output: ()

output: Change Parser's Output to (output,)

let digit_parser = ('0'..='9').output(2024);

let res = rp::parse(&digit_parser, "123456hello_world".chars());
assert_eq!(res.output.unwrap(), (2024,));
assert_eq!(res.it.collect::<String>(), "23456hello_world");

Output: (T,) where T is the type of the value you provided.

string, vec: captures the matched range into String or Vec<T>

Note

string can be only used for std::str::Chars, and vec can be only used for ExactSizeIterator.

let digits_parser = ('0'..='9').repeat(0..).string();

let res = rp::parse(&digits_parser, "123456hello_world".chars());
assert_eq!(res.output.unwrap(), ("123456".to_string(),));
assert_eq!(res.it.collect::<String>(), "hello_world");

Output: (String,) or (Vec<Iterator::Item>,)

not_consume: check if the pattern is matched or not, without consuming the input

let digit_parser = ('0'..='9').not_consume();

let res = rp::parse(&digit_parser, "12345".chars());
assert_eq!(res.output.unwrap(), ('1',));
assert_eq!(res.it.collect::<String>(), "12345"); // iterator is not consumed

Output: Output of Self

For complex, recursive pattern

By default, all the 'parser-generating' functions consumes input Parser and returns a new instance. These processes create new generic Parser object entirely at compile-time.

However, in some cases, you may want to define a recursive parser. Which involves 'reference-of-parser' or 'virtual-class-like' structure.

Luckily, Rust std provides wrapper for these cases. Rc, RefCell, Box are the most common ones.

For Rc and RefCell, you can wrap any parser with them. They will be treated as a Parser object.

// making shared, interior-mutable parser
let hello_parser = "hello".into_parser();
let hello_parser = std::cell::RefCell::new(hello_parser);
let hello_parser = std::rc::Rc::new(hello_parser);

For Box, you can use DynBox* to wrap any parser. With DynBox*, you can assign any parser with same Output type.

let hello_parser = "hello".into_parser();

let mut dynamic_parser: DynBoxChars<(char,)> = Default::new(); // Default implemented
dynamic_parser.parse( "hello".chars() ); // this will panic, since the parser is not assigned yet

// set dynamic_parser to hello_parser
dynamic_parser.assign( "hello" );
let res = dynamic_parser.parse( "hello".chars() ); // success

// set dynamic_parser to digit_parser
dynamic_parser.assign( '0'..='9' );
let res = dynamic_parser.parse( "01234".chars() ); // success

Default trait is implemented with always-panic-parser. You must assign it later.

For now, there are three types of DynBox*:

  • DynBoxChars<Output>: for std::str::Chars
  • DynBoxSlice<Output,T>: for std::iter::Cloned<std::slice::Iter<T>>
  • DynBoxSliceCopied<Output,T>: for std::iter::Copied<std::slice::Iter<T>> Once you wrap the parser through DynBox*, you can only use corresponding iterator in parse(...).

You can refer HERE to implement for other iterator types.

No runtime deps