feat(convert): add derive macros for To and FromBytes

Cargo.toml

@@ -1,6 +1,6 @@
 [workspace]
 resolver = "2"
-members = ["extism-maturin", "manifest", "runtime", "libextism", "convert"]
+members = ["extism-maturin", "manifest", "runtime", "libextism", "convert", "convert-macros"]
 exclude = ["kernel"]
 
 [workspace.package]
@@ -14,4 +14,5 @@ version = "0.0.0+replaced-by-ci"
 [workspace.dependencies]
 extism = { path = "./runtime", version = "0.0.0+replaced-by-ci" }
 extism-convert = { path = "./convert", version = "0.0.0+replaced-by-ci" }
+extism-convert-macros = { path = "./convert-macros", version = "0.0.0+replaced-by-ci" }
 extism-manifest = { path = "./manifest", version = "0.0.0+replaced-by-ci" }

convert-macros/Cargo.toml

@@ -0,0 +1,28 @@
+[package]
+name = "extism-convert-macros"
+readme = "./README.md"
+edition.workspace = true
+authors.workspace = true
+license.workspace = true
+homepage.workspace = true
+repository.workspace = true
+version.workspace = true
+description = "Macros to remove boilerplate with Extism"
+
+[lib]
+proc-macro = true
+
+[features]
+extism-path = []
+extism-pdk-path = []
+
+[dependencies]
+# manyhow = "0.10.4"
+manyhow.version = "0.11.0"
+proc-macro-crate = "3.1.0"
+proc-macro2 = "1.0.78"
+quote = "1.0.35"
+syn = { version = "2.0.48", features = ["derive"] }
+
+[dev-dependencies]
+trybuild = "1.0.89"

convert-macros/src/lib.rs

@@ -0,0 +1,108 @@
+use std::iter;
+
+use manyhow::{ensure, error_message, manyhow, Result};
+use proc_macro_crate::{crate_name, FoundCrate};
+use quote::{format_ident, quote, ToTokens};
+use syn::{parse_quote, Attribute, DeriveInput, Path};
+
+/// Tries to resolve the path to `extism_convert` dynamically, falling back to feature flags when unsuccessful.
+fn convert_path() -> Path {
+    match (
+        crate_name("extism"),
+        crate_name("extism-convert"),
+        crate_name("extism-pdk"),
+    ) {
+        (Ok(FoundCrate::Name(name)), ..) => {
+            let ident = format_ident!("{name}");
+            parse_quote!(::#ident::convert)
+        }
+        (_, Ok(FoundCrate::Name(name)), ..) | (.., Ok(FoundCrate::Name(name))) => {
+            let ident = format_ident!("{name}");
+            parse_quote!(::#ident)
+        }
+        (Ok(FoundCrate::Itself), ..) => parse_quote!(::extism::convert),
+        (_, Ok(FoundCrate::Itself), ..) => parse_quote!(::extism_convert),
+        (.., Ok(FoundCrate::Itself)) => parse_quote!(::extism_pdk),
+        _ if cfg!(feature = "extism-path") => parse_quote!(::extism::convert),
+        _ if cfg!(feature = "extism-pdk-path") => parse_quote!(::extism_pdk),
+        _ => parse_quote!(::extism_convert),
+    }
+}
+
+fn extract_encoding(attrs: &[Attribute]) -> Result<Path> {
+    let encodings: Vec<_> = attrs
+        .iter()
+        .filter(|attr| attr.path().is_ident("encoding"))
+        .collect();
+    ensure!(!encodings.is_empty(), "encoding needs to be specified"; try = "`#[encoding(ToJson)]`");
+    ensure!(encodings.len() < 2, encodings[1], "only one encoding can be specified"; try = "remove `{}`", encodings[1].to_token_stream());
+
+    Ok(encodings[0].parse_args().map_err(
+        |e| error_message!(e.span(), "{e}"; note= "expects a path"; try = "`#[encoding(ToJson)]`"),
+    )?)
+}
+
+#[manyhow]
+#[proc_macro_derive(ToBytes, attributes(encoding))]
+pub fn to_bytes(
+    DeriveInput {
+        attrs,
+        ident,
+        generics,
+        ..
+    }: DeriveInput,
+) -> Result {
+    let encoding = extract_encoding(&attrs)?;
+    let convert = convert_path();
+
+    let (_, type_generics, _) = generics.split_for_impl();
+
+    let mut generics = generics.clone();
+    generics.make_where_clause().predicates.push(
+        parse_quote!(for<'__to_bytes_b> #encoding<&'__to_bytes_b Self>: #convert::ToBytes<'__to_bytes_b>)
+    );
+    generics.params = iter::once(parse_quote!('__to_bytes_a))
+        .chain(generics.params)
+        .collect();
+    let (impl_generics, _, where_clause) = generics.split_for_impl();
+
+    Ok(quote! {
+        impl #impl_generics #convert::ToBytes<'__to_bytes_a> for #ident #type_generics #where_clause
+        {
+            type Bytes = ::std::vec::Vec<u8>;
+
+            fn to_bytes(&self) -> Result<Self::Bytes, #convert::Error> {
+                #convert::ToBytes::to_bytes(&#encoding(self)).map(|__bytes| __bytes.as_ref().to_vec())
+            }
+        }
+
+    })
+}
+
+#[manyhow]
+#[proc_macro_derive(FromBytes, attributes(encoding))]
+pub fn from_bytes(
+    DeriveInput {
+        attrs,
+        ident,
+        mut generics,
+        ..
+    }: DeriveInput,
+) -> Result {
+    let encoding = extract_encoding(&attrs)?;
+    let convert = convert_path();
+    generics
+        .make_where_clause()
+        .predicates
+        .push(parse_quote!(#encoding<Self>: #convert::FromBytesOwned));
+    let (impl_generics, type_generics, where_clause) = generics.split_for_impl();
+    Ok(quote! {
+        impl #impl_generics #convert::FromBytesOwned for #ident #type_generics #where_clause
+        {
+            fn from_bytes_owned(__data: &[u8]) -> Result<Self, #convert::Error> {
+                <#encoding<Self> as #convert::FromBytesOwned>::from_bytes_owned(__data).map(|__encoding| __encoding.0)
+            }
+        }
+
+    })
+}

convert-macros/tests/ui.rs

@@ -0,0 +1,5 @@
+#[test]
+fn ui() {
+    let t = trybuild::TestCases::new();
+    t.compile_fail("tests/ui/*.rs");
+}

convert-macros/tests/ui/invalid-encoding.rs

@@ -0,0 +1,23 @@
+use extism_convert_macros::ToBytes;
+
+#[derive(ToBytes)]
+struct MissingEncoding;
+
+#[derive(ToBytes)]
+#[encoding]
+struct EmptyAttr;
+
+#[derive(ToBytes)]
+#[encoding = "string"]
+struct EqNoParen;
+
+#[derive(ToBytes)]
+#[encoding(something, else)]
+struct NotAPath;
+
+#[derive(ToBytes)]
+#[encoding(Multiple)]
+#[encoding(Encodings)]
+struct MultipleEncodings;
+
+fn main() {}

convert-macros/tests/ui/invalid-encoding.stderr

@@ -0,0 +1,44 @@
+error: encoding needs to be specified
+
+         = try: `#[encoding(ToJson)]`
+ --> tests/ui/invalid-encoding.rs:3:10
+  |
+3 | #[derive(ToBytes)]
+  |          ^^^^^^^
+  |
+  = note: this error originates in the derive macro `ToBytes` (in Nightly builds, run with -Z macro-backtrace for more info)
+
+error: expected attribute arguments in parentheses: #[encoding(...)]
+
+         = note: expects a path
+         = try: `#[encoding(ToJson)]`
+ --> tests/ui/invalid-encoding.rs:7:3
+  |
+7 | #[encoding]
+  |   ^^^^^^^^
+
+error: expected parentheses: #[encoding(...)]
+
+         = note: expects a path
+         = try: `#[encoding(ToJson)]`
+  --> tests/ui/invalid-encoding.rs:11:12
+   |
+11 | #[encoding = "string"]
+   |            ^
+
+error: unexpected token
+
+         = note: expects a path
+         = try: `#[encoding(ToJson)]`
+  --> tests/ui/invalid-encoding.rs:15:21
+   |
+15 | #[encoding(something, else)]
+   |                     ^
+
+error: only one encoding can be specified
+
+         = try: remove `#[encoding(Encodings)]`
+  --> tests/ui/invalid-encoding.rs:20:1
+   |
+20 | #[encoding(Encodings)]
+   | ^^^^^^^^^^^^^^^^^^^^^^

convert/Cargo.toml

@@ -18,11 +18,15 @@ protobuf = { version = "3.2.0", optional = true }
 rmp-serde = { version = "1.1.2", optional = true }
 serde = "1.0.186"
 serde_json = "1.0.105"
+extism-convert-macros.workspace = true
 
 [dev-dependencies]
+# extism = { path = "../runtime" }
 serde = { version = "1.0.186", features = ["derive"] }
 
 [features]
 default = ["msgpack", "prost", "raw"]
 msgpack = ["rmp-serde"]
 raw = ["bytemuck"]
+extism-path = ["extism-convert-macros/extism-path"]
+extism-pdk-path = ["extism-convert-macros/extism-pdk-path"]

convert/src/encoding.rs

@@ -8,11 +8,22 @@ use base64::Engine;
 /// For example, the following line creates a new JSON encoding using serde_json:
 ///
 /// ```
-/// extism_convert::encoding!(MyJson, serde_json::to_vec, serde_json::from_slice);
+#[cfg_attr(
+    feature = "extism-path",
+    doc = "extism::convert::encoding!(MyJson, serde_json::to_vec, serde_json::from_slice);"
+)]
+#[cfg_attr(
+    feature = "extism-pdk-path",
+    doc = "extism_pdk::encoding!(MyJson, serde_json::to_vec, serde_json::from_slice);"
+)]
+#[cfg_attr(
+    not(any(feature = "extism-path", feature = "extism-pdk-path")),
+    doc = "extism_convert::encoding!(MyJson, serde_json::to_vec, serde_json::from_slice);"
+)]
 /// ```
 ///
-/// This will create a struct `struct MyJson<T>(pub T)` and implement `ToBytes` using `serde_json::to_vec`
-/// and `FromBytesOwned` using `serde_json::from_vec`
+/// This will create a struct `struct MyJson<T>(pub T)` and implement [`ToBytes`] using [`serde_json::to_vec`]
+/// and [`FromBytesOwned`] using [`serde_json::from_slice`]
 #[macro_export]
 macro_rules! encoding {
     ($pub:vis $name:ident, $to_vec:expr, $from_slice:expr) => {

convert/src/from_bytes.rs

@@ -1,14 +1,90 @@
 use crate::*;
 
+pub use extism_convert_macros::FromBytes;
+
 /// `FromBytes` is used to define how a type should be decoded when working with
 /// Extism memory. It is used for plugin output and host function input.
+///
+/// `FromBytes` can be derived by delegating encoding to generic type implementing
+/// `FromBytes`, e.g., [`Json`], [`Msgpack`].
+///
+/// ```
+#[cfg_attr(
+    feature = "extism-path",
+    doc = "use extism::convert::{Json, FromBytes};"
+)]
+#[cfg_attr(
+    feature = "extism-pdk-path",
+    doc = "use extism_pdk::{Json, FromBytes};"
+)]
+#[cfg_attr(
+    not(any(feature = "extism-path", feature = "extism-pdk-path")),
+    doc = "use extism_convert::{Json, FromBytes};"
+)]
+/// use serde::Deserialize;
+///
+/// #[derive(FromBytes, Deserialize, PartialEq, Debug)]
+/// #[encoding(Json)]
+/// struct Struct {
+///     hello: String,
+/// }
+///
+/// assert_eq!(Struct::from_bytes(br#"{"hello":"hi"}"#)?, Struct { hello: "hi".into() });
+/// # Ok::<(), extism_convert::Error>(())
+/// ```
+///
+/// Custom encodings can also be used, through new-types with a single generic
+/// argument, i.e., `Type<T>(T)`, that implement `FromBytesOwned` for the struct.
+///
+/// ```
+/// use std::str::{self, FromStr};
+/// use std::convert::Infallible;
+#[cfg_attr(
+    feature = "extism-path",
+    doc = "use extism::convert::{Error, FromBytes, FromBytesOwned};"
+)]
+#[cfg_attr(
+    feature = "extism-pdk-path",
+    doc = "use extism_pdk::{Error, FromBytes, FromBytesOwned};"
+)]
+#[cfg_attr(
+    not(any(feature = "extism-path", feature = "extism-pdk-path")),
+    doc = "use extism_convert::{Error, FromBytes, FromBytesOwned};"
+)]
+///
+/// // Custom serialization using `FromStr`
+/// struct StringEnc<T>(T);
+/// impl<T: FromStr> FromBytesOwned for StringEnc<T> where Error: From<<T as FromStr>::Err> {
+///     fn from_bytes_owned(data: &[u8]) -> Result<Self, Error> {
+///         Ok(Self(str::from_utf8(data)?.parse()?))
+///     }
+/// }
+///
+/// #[derive(FromBytes, PartialEq, Debug)]
+/// #[encoding(StringEnc)]
+/// struct Struct {
+///     hello: String,
+/// }
+///
+/// impl FromStr for Struct {
+///     type Err = Infallible;
+///     fn from_str(s: &str) -> Result<Self, Infallible> {
+///         Ok(Self { hello: s.to_owned() })
+///     }
+/// }
+///
+/// assert_eq!(Struct::from_bytes(b"hi")?, Struct { hello: "hi".into() });
+/// # Ok::<(), extism_convert::Error>(())
+/// ```
 pub trait FromBytes<'a>: Sized {
     /// Decode a value from a slice of bytes
     fn from_bytes(data: &'a [u8]) -> Result<Self, Error>;
 }
 
-/// `FromBytesOwned` is similar to `FromBytes` but it doesn't borrow from the input slice.
-/// `FromBytes` is automatically implemented for all types that implement `FromBytesOwned`
+/// `FromBytesOwned` is similar to [`FromBytes`] but it doesn't borrow from the input slice.
+/// [`FromBytes`] is automatically implemented for all types that implement `FromBytesOwned`.
+///
+/// `FromBytesOwned` can be derived through [`#[derive(FromBytes)]`](FromBytes).
 pub trait FromBytesOwned: Sized {
     /// Decode a value from a slice of bytes, the resulting value should not borrow the input
     /// data.

convert/src/lib.rs

@@ -5,6 +5,9 @@
 //! similar to [axum extractors](https://docs.rs/axum/latest/axum/extract/index.html#intro) - they are
 //! implemented as a tuple struct with a single field that is meant to be extracted using pattern matching.
 
+// Makes proc-macros able to resolve `::extism_convert` correctly
+extern crate self as extism_convert;
+
 pub use anyhow::Error;
 
 mod encoding;