shapely

shapely provides runtime reflection for Rust.

Any type that implements Shapely trait returns a Shape, which describes:

• The memory layout of the type
• Its innards: struct fields, underlying type for newtypes, etc.
• How to drop it in place

The Partial type is able to allocate (or work from a &mut MaybeUninit<T>) any Shapely type, and gradually initialize its fields — until the fully-built value is moved out of the partial.

It comes with a derive macro that uses unsynn for speed of compilation.

Ecosystem

The main shapely crate re-exports symbols from:

• shapely-core, which defines the main Shapely trait and the Shape struct
• shapely-derive, which implements the Shapely derive attribute as a fast/light proc macro powered by unsynn

shapely supports deserialization from multiple data formats through dedicated crates:

• shapely-json: JSON deserialization
• shapely-yaml: YAML deserialization
• shapely-msgpack: MessagePack deserialization
• shapely-urlencoded: URL-encoded form data deserialization

Additionally:

• shapely-pretty is able to pretty-print Shapely types.
• shapely-codegen is internal and generates some of the code of shapely-core

Implementing Your Own Deserializer

To implement a custom deserializer for a new format, you'll need to work with the following key components from shapely:

Key Types

• [Partial]: The central type for building shapely values incrementally
• [Shape]: Describes the memory layout and structure of a type
• [Innards]: Represents the internal structure (Scalar, Struct, etc.)
• [Scalar]: Represents primitive types like String, u64, etc.

Implementation Pattern

1. Create a function that takes a &mut [Partial] and your format's input (string, bytes, etc.)
2. Examine the shape of the partial using [Partial::shape]
3. Handle different shapes based on [Shape::innards]:
   • For [Innards::Scalar], use [Partial::scalar_slot] to get and fill the slot
   • For [Innards::Struct], iterate through fields, using [Partial::slot_by_name] to access each field
   • [Innards::Map] and [Innards::List] come with vtables, they have helper slots as well
   • Create nested [Partial] instances for complex fields and fill them recursively

When in doubt, refer to the shapely-json implementation — it's the most featureful.

Example Implementation Skeleton

use shapely_core::{Partial, Innards, Scalar, FieldError, Shape};
use std::convert::From;

#[derive(Debug)]
pub enum MyFormatError {
    UnsupportedType,
    UnsupportedShape,
    FieldError(FieldError),
}

impl From<FieldError> for MyFormatError {
    fn from(err: FieldError) -> Self {
        MyFormatError::FieldError(err)
    }
}

pub fn from_my_format(partial: &mut Partial, input: &[u8]) -> Result<(), MyFormatError> {
    let shape_desc = partial.shape();
    let shape = shape_desc.get();

    match &shape.innards {
        Innards::Scalar(scalar) => {
            let slot = partial.scalar_slot().expect("Scalar slot");
            // Parse scalar value from input and fill slot
            match scalar {
                Scalar::String => slot.fill("/* parsed string */".to_string()),
                Scalar::U64 => slot.fill(0u64),
                // Handle other scalar types
                _ => return Err(MyFormatError::UnsupportedType),
            }
        },
        Innards::Struct { .. } => {
            // Parse struct fields from input
            for (field_name, field_value) in [("field1", "value1"), ("field2", "value2")].iter() {
                let slot = partial.slot_by_name(field_name)?;

                // Create a partial for the field and fill it recursively
                let mut partial_field = Partial::alloc(slot.shape());
                // Recursively deserialize field_value into partial_field
                // ...
                slot.fill_from_partial(partial_field);
            }
        },
        // Handle other shapes as needed
        _ => return Err(MyFormatError::UnsupportedShape),
    }

    Ok(())
}

For more detailed examples, examine the implementation of existing deserializers like shapely-json or shapely-msgpack.

Funding

Thanks to Namespace for providing fast GitHub Actions workers:

License

Licensed under either of:

• Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.