serde-generate

serde-generate

This crate aims to compile the data formats extracted from Rust by serde-reflection into type definitions and (de)serialization methods for other programming languages.

It can be used as a library or as a command-line tool (see serdegen below).

Supported Languages

The following programming languages are fully supported as target languages:

• C++ 17
• Java 8
• Python 3 (requires numpy >= 1.20.1)
• Rust 2018
• Go >= 1.14
• C# (NetCoreApp >= 2.1)
• Swift 5.3
• OCaml

The following languages are partially supported and still considered under development:

• TypeScript 4 (packaged and tested with Deno) (follow-up issue)
• Dart >= 2 (follow-up issue)

Supported Encodings

Type definitions in a target language are meant to be used together with a runtime library that provides (de)serialization in a particular Serde encoding format.

This crate provides easy-to-deploy runtime libraries for the following binary formats, in all supported languages:

• Bincode (default configuration only),
• BCS (short for Binary Canonical Serialization, the main format used in the Diem blockchain).

Quick Start with Python and Bincode

In the following example, we transfer a Test value from Rust to Python using bincode.

use serde::{Deserialize, Serialize};
use serde_reflection::{Registry, Tracer, TracerConfig};
use std::io::Write;

#[derive(Serialize, Deserialize)]
struct Test {
    a: Vec<u64>,
    b: (u32, u32),
}

// Obtain the Serde format of `Test`. (In practice, formats are more likely read from a file.)
let mut tracer = Tracer::new(TracerConfig::default());
tracer.trace_simple_type::<Test>().unwrap();
let registry = tracer.registry().unwrap();

// Create Python class definitions.
let mut source = Vec::new();
let config = serde_generate::CodeGeneratorConfig::new("testing".to_string())
    .with_encodings(vec![serde_generate::Encoding::Bincode]);
let generator = serde_generate::python3::CodeGenerator::new(&config);
generator.output(&mut source, &registry)?;

assert!(
    String::from_utf8_lossy(&source).contains(
    r#"
@dataclass(frozen=True)
class Test:
    a: typing.Sequence[st.uint64]
    b: typing.Tuple[st.uint32, st.uint32]
"#));

// Append some test code to demonstrate Bincode deserialization
// using the runtime in `serde_generate/runtime/python/bincode`.
writeln!(
    source,
    r#"
value = Test.bincode_deserialize(bytes({:?}))
assert value == Test(a=[4, 6], b=(3, 5))
"#,
    bincode::serialize(&Test { a: vec![4, 6], b: (3, 5) }).unwrap(),
)?;

// Execute the Python code.
let mut child = std::process::Command::new("python3")
    .arg("-")
    .env("PYTHONPATH", std::env::var("PYTHONPATH").unwrap_or_default() + ":runtime/python")
    .stdin(std::process::Stdio::piped())
    .spawn()?;
child.stdin.as_mut().unwrap().write_all(&source)?;
let output = child.wait_with_output()?;
assert!(output.status.success());

Binary Tool

In addition to a Rust library, this crate provides a binary tool serdegen to process Serde formats saved on disk.

The tool serdegen assumes that a Rust value of type serde_reflection::Registry has been serialized into a YAML file. The recommended way to generate such a value is to use the library serde-reflection to introspect Rust definitions (see also the example above).

For a quick test, one may create a test file like this:

cat >test.yaml <<EOF
---
Foo:
  ENUM:
    0:
      A:
        NEWTYPE:
          U64
    1:
      B: UNIT
EOF

Then, the following command will generate Python class definitions and write them into test.py:

cargo run -p serde-generate-bin -- --language python3 test.yaml > test.py

To create a python module test and install the bincode runtime in a directory $DEST, you may run:

cargo run -p serde-generate-bin -- --language python3 --with-runtimes serde bincode --module-name test --target-source-dir "$DEST" test.yaml

See the help message of the tool with --help for more options.

Note: Outside of this repository, you may install the tool with cargo install serde-generate-bin then use $HOME/.cargo/bin/serdegen.

Contributing

See the CONTRIBUTING file for how to help out.

License

This project is available under the terms of either the Apache 2.0 license or the MIT license.