Convert ABNF like grammar to rust-peg build rowan lossless syntax tree

Using rowan library lossless syntax tree

How To Use

Add dependencies:

char-classes = "0.3.4"
peg = "0.8.5"
rowan = "0.15.15"
rowan-peg-utils = "0.2.3"

Edit your grammar declaration:

exports [ pair-list ]

pair-list = whitespace *pair
pair    = key whitespace "=" whitespace val whitespace
key     = text
val     = number / text
number  = int ["." int]
int     = $(+<0-9>) @int ; Add `int` to the expected set when parsing fails
text    = $(+<a-zA-Z_->) @text
whitespace = $(*< \t\r\n>)

Generate grammar from your grammar declaration:

rowan-peg your-lang.abnf > src/grammar.rs

Example

Cargo build

If you want to use cargo to build grammar, you can use rowan-peg-macro

#[rowan_peg_macro::build(path = "grammar.abnf")]
mod grammar {}

A simple json parser example-json

cd examples/example-json
cargo run -- example.json
[examples/example-json/src/main.rs:23:5] &json = JsonText(
    [email protected]
      [email protected] ""
      [email protected]
        [email protected]
          [email protected]
            [email protected] ""
            [email protected] "{"
            [email protected] "\n    "
          [email protected]
            [email protected]
              [email protected] "\""
              [email protected]
                [email protected] "s"
              [email protected]
                [email protected] "t"
              [email protected]
...
...

Generate Grammar Declaration

cargo run -- examples/example-json/json.abnf

Grammar

Exports

Export or rename rules using exports []

Such as exports [ foo ], export foo rule (public it)

Such as exports [ bar = foo ], export bar rule (public and renamed into bar)

Rule

Define a rule using my-rule = "rule"

When a rule belongs to the following situations, it is a token rule:

• rule body only exist a slice and optional expected

  such as: a = $"x", a = $("a" ["b"]), a = $("a" ["b"]) @a

Syntax Elements

String

Hard matched content cannot be escaped, such as "a" "+"

• For identifiers, they will be renamed as {name}_kw, such as "if" -> if_kw
• For some punctuation marks, they will be renamed as names, such as "!" -> bang
• For those that cannot be renamed, an error will be reported. Please refer to the Slice for solution

[!NOTE] Because it cannot be escaped, double quotes are represented by <">

Matches

Match a some of characters or character ranges, can only be used in slice

Escaping equivalent rust string-literal

Syntax equivalent char-classes::any!() macro

Such as <a-z> matches lowercase alpha, <^0-9> matches non number

Slice

Collect any parsing range as a token, must be used in token rule

Such as: a-or-ab = $("a" ["b"]), it can parsing a, ab

Choice

Sequential choice, from PEG

Such as: x = "a" "b" / "c", it can parser ab, c

When a branch fails, it will return to the beginning to try parsing the next branch

If no branch is successful, the this rule fails

Repeat

Repeat parsing specified number of times

• +x: Repeat x 1 to more times
• *x: Repeat x 0 to more times
• 233*x: Repeat x 233 to more times
• *6x: Repeat x 0 to 6 times
• 2*6x: Repeat x 2 to 6 times
• 6x: Repeat x 6 times
• 6(x y): Repeat x y 6 times

Lookahead

Positive lookahead and Negative lookahead, from PEG

• Positive lookahead: Parsing, but success also returns to the beginning, which is a non consumption parsing
• Negative lookahead: Similar to positive lookahead, but fails upon success and succeeds upon failure

Quiet

Map to rust-peg quiet! {}, will not add any expecteds from the target to the error message

Such as: x = "x" ~" " "y", quiet space token expecteds message

Comment

Use semicolons for comment, such as:

x = "a" ; this is a comment