feat: Add MoonBit nom-like parser combinator framework (cogna-dev/parkit/nom) with JSON parser, Rust criterion benchmarks, and cross-language comparison

[Copilot]

Implements a Rust nom-inspired parser combinator library in MoonBit as a monorepo package under cogna-dev/parkit, including an ECMA-404 compliant JSON parser sample, Rust nom reference cross-validation tests, MoonBit benchmarks (native target), and Rust criterion.rs benchmarks with a cross-language performance comparison.

src/nom/ — parkit core (cogna-dev/parkit/nom)

• Parser[O] wraps (String) -> IResult[O]; IResult[O] is Result[(String, O), ParseError]
• Character matchers: tag, satisfy, char_, take_while, take_while1, take, take_while_m_n
• Combinators: map, map_res, alt, pair, tuple3, preceded, terminated, delimited, many0, many1, separated_list0/1, opt, verify, success, fail, eof, context
• Numeric parsers: uint, int, double (manual computation, no external deps)

let p = @nom.delimited(
  @nom.char_('"'),
  @nom.take_while(fn(c) { c != '"' }),
  @nom.char_('"'),
)
assert_eq!(p.parse("\"hello\" rest"), Ok((" rest", "hello")))

examples/json/ — JSON parser sample (not published to mooncakes)

• JsonValue enum (Null | Bool | Number | Str | Array | Object) with Show + Eq
• Handles full string escape sequences including \uXXXX
• Top-level parse(String) -> Result[JsonValue, ParseError]
• Lives in a separate cogna-dev/parkit-examples workspace module so it is not published to mooncakes

examples/benchmark/ — MoonBit benchmarks (not published to mooncakes)

• Uses MoonBit @bench.T API (moon bench --target native)
• Covers JSON parse, tag, take_while1, separated_list0

Benchmark results

Benchmarks run on ubuntu-latest. MoonBit targets the native backend; Rust uses criterion.rs with a native release build. Both use the same 3-object JSON array input.

Benchmark	MoonBit (native)	Rust nom (native)
json_parse (3-object array)	1.17 ms	5.53 µs

MoonBit native benchmark details:

Benchmark	Mean	σ	Range
json_parse (3-object array)	1.17 ms	±6.29 µs	1.16 ms … 1.19 ms
tag	0.21 µs	±0.00 µs	0.21 µs … 0.21 µs
take_while1_digits	0.43 µs	±0.00 µs	0.43 µs … 0.43 µs
separated_list0	2.49 µs	±0.02 µs	2.47 µs … 2.54 µs

The performance difference reflects MoonBit's runtime and memory model overhead compared to Rust's zero-cost abstractions — both parsers implement the same combinator logic and produce identical results.

reference/nom-json/ — Rust nom reference tests and criterion benchmarks

• Full JSON parser built with Rust nom 8, mirroring the MoonBit implementation
• 18 test cases matching every case in examples/json/json_test.mbt, confirming identical results across both implementations
• criterion.rs benchmark in benches/json_bench.rs using the same input as the MoonBit benchmark suite
• Run tests: cargo test --manifest-path reference/nom-json/Cargo.toml
• Run benchmarks: cargo bench --manifest-path reference/nom-json/Cargo.toml

CI & infrastructure

• .github/workflows/ci.yml — moonbit job via cogna-dev/moonbit-actions/setup@v0 with moon check, moon build, moon test all using --manifest-path moon.work.json; rust-reference job runs cargo test on reference/nom-json
• moon.work.json — MoonBit workspace linking the publishable library (.) and the examples module (examples/)
• examples/moon.mod.json — separate cogna-dev/parkit-examples module for JSON sample and benchmarks
• .github/workflows/copilot-setup-steps.yml — agent env via cogna-dev/moonbit-actions/setup@v0
• .agents/skills/moonbit — cogna-dev/moonbit-skill registered and initialized as a git submodule
• Cargo.toml — Cargo workspace root including reference/nom-json
• .gitignore — excludes _build/, target/, .mooncakes/, Cargo.lock

Original prompt

Create a Rust Nom-like parser combinator framework:

• Same API with Rust nom
• A JSON parser as sample
• MoonBit benchmark

Use:

• https://github.com/cogna-dev/moonbit-skill as git submodule at .agents/skills/moonbit
• https://github.com/cogna-dev/moonbit-actions use this action for moonbit ci testing

References:

1. https://ecma-international.org/publications-and-standards/standards/ecma-404/
2. https://github.com/rust-bakery/nom
3. https://docs.moonbitlang.com/en/latest/language/benchmarks.html

[yufeiminds]

@copilot I have already set cogna-dev/moonbit-actions and https://github.com/cogna-dev/moonbit-skill to public. You can try re-importing them; just configure MoonBit using cogna-dev/moonbit-actions/setup@v0. You can also try running the agent again, as I have added moonbitlang.com to the allowed list.

[Copilot]

@copilot I have already set cogna-dev/moonbit-actions and https://github.com/cogna-dev/moonbit-skill to public. You can try re-importing them; just conf...

Done in commit a320bac. Updated both workflows to use cogna-dev/moonbit-actions/setup@v0 (replacing the old hustcer/setup-moonbit@v1 in copilot-setup-steps and cogna-dev/moonbit-actions@main in CI), and re-initialized the cogna-dev/moonbit-skill submodule at .agents/skills/moonbit now that the repo is public.

[yufeiminds]

@copilot See follows:

1. This repo should as MoonBit monorepo, and current parsec as a package. parsec might be confused with Haskell parsec. Please adjust the path so that the final referenced package name is cogna-dev/parskit/nom. I will rename this repo as cogna-dev/parskit.
2. Please add JSON parser reference testing with Rust nom, check it got same result
3. Run benchmark and put the result at README

[Copilot]

@copilot See follows:

1. This repo should as MoonBit monorepo, and current parsec as a package. parsec might be confused with Haskell parsec. Please adjust the path so that the final referenced pac...

Done in commit b986e43. Here's a summary of the three changes:

1. Monorepo restructure — module renamed cogna-dev/parskit, src/lib/ → src/nom/. The parser library is now referenced as cogna-dev/parskit/nom (alias @nom). Also fixed several deprecated MoonBit APIs (unsafe_char_at → get_char, substring → slice syntax, typealias → type =, pub enum → pub(all) enum).

2. Rust nom reference tests — added reference/nom-json/ (Rust nom 8), mirroring every test case in src/json/json_test.mbt. Both suites now pass: 58 MoonBit tests and 18 Rust reference tests, confirming identical results. Run with cargo test --manifest-path reference/nom-json/Cargo.toml. CI has a new rust-reference job.

3. Benchmark results in README — ran moon bench (MoonBit wasm-gc backend) and added the results table:

   Benchmark	Mean
   json_parse	431.20 µs
   tag	0.05 µs
   take_while1_digits	0.11 µs
   separated_list0	0.65 µs

[yufeiminds]

@copilot

1. Use https://github.com/bheisler/criterion.rs for Rust nom JSON parser benchmark, and compare with MoonBit implementation, the table write into README
2. I had changed this repo's name as cogna-dev/parkit, please change it on README and other files. the parser combinator library will import as cogna-dev/parkit/nom

[Copilot]

@copilot

1. Use https://github.com/bheisler/criterion.rs for Rust nom JSON parser benchmark, and compare with MoonBit implementation, the table write into README
   ...

Done in commit 254b260.

1. Criterion.rs benchmarks — added reference/nom-json/benches/json_bench.rs using criterion.rs with the same 3-object JSON input as the MoonBit benchmark. Run with cargo bench --manifest-path reference/nom-json/Cargo.toml. The comparison table in README:

   Benchmark	MoonBit (wasm-gc)	Rust nom (native)
   json_parse (3-object array)	431.20 µs	5.53 µs

   The difference reflects wasm-gc runtime overhead — both parsers implement identical combinator logic and produce identical results.

2. Repo rename — updated all cogna-dev/parskit → cogna-dev/parkit in moon.mod.json, moon.pkg.json files (both parkit/nom and parkit/json), and README.

[yufeiminds]

@copilot

1. Please use moonbit native target to benchmark
2. Please seperate moonbit json parser sample with src，don't pushlish it to mooncakes

[Copilot]

@copilot

1. Please use moonbit native target to benchmark
   ...

Done in commit d1c05da.

1. Native target benchmarks — switched moon bench to --target native. Updated CI and README with native results:

   Benchmark	MoonBit (native)	Rust nom (native)
   json_parse (3-object array)	1.17 ms	5.53 µs
   tag	0.21 µs	—
   take_while1_digits	0.43 µs	—
   separated_list0	2.49 µs	—

2. JSON parser separated from src/ — moved src/json/ and src/benchmark/ to examples/ as a separate MoonBit workspace module (cogna-dev/parkit-examples). Only src/nom/ remains under src/ and will be published to mooncakes as cogna-dev/parkit/nom. The workspace is defined in moon.work.json; all commands now use --manifest-path moon.work.json.

Warning

Firewall rules blocked me from connecting to one or more addresses (expand for details)

I tried to connect to the following addresses, but was blocked by firewall rules:

• mooncakes.io
  • Triggering command: /usr/lib/git-core/git-remote-https /usr/lib/git-core/git-remote-https origin REDACTED moonbitlang/core/error -i /home/REDACTED/.moon/lib/core/_build/native/release/bundle/builtin/builtin.mi:builtin -i /home/REDACTED/.moon/lib/core/_build/native/release/bundle/debug/debug.mi:debug -pkg-sources moonbitlang/core/error:/home/REDACTED/.moon/lib/core/error -target native (dns block)

If you need me to access, download, or install something from one of these locations, you can either:

• Configure Actions setup steps to set up my environment, which run before the firewall is enabled
• Add the appropriate URLs or hosts to the custom allowlist in this repository's Copilot coding agent settings (admins only)