1 unstable release
0.1.0 | Aug 25, 2019 |
---|
#1993 in Procedural macros
4,428 downloads per month
Used in 3 crates
(2 directly)
47KB
868 lines
serde_syn
serde_syn is a serde backend for parsing Rust syntax inside procedural macros. For example, you can deserialize parameters for a custom derive out of attributes and directly into structs. The goal is to eliminate the parsing boilerplate that goes into writing procedural macros.
The interface to serde_syn is fairly minimal. Here are a few ways to use it:
- syn requires a
Parser
implementation to process syntax inside already parsed/visited attributes. Theparser
function creates exactly that! - If you are working directly with
proc_macro
/proc_macro2
token streams or strings, you should also useparser
. - If you are implementing syn's
Parse
trait yourself, you should use thefrom_stream
function which takes in aParseStream
.
Lots of pre-made configurations exist inside the config
module for
common syntaxes (JSON-like, attribute-like, expression-like, etc) or you can
combine flags to build your own.
Example derive implementation
Here you can see a simple derive macro implementation. For more examples, see the examples directory.
#
#
/// The format of `named` attributes.
#[derive(Deserialize)]
struct Props {
rename: Option<String>, // #[named(rename="hello")]
lowercase: Option<()>, // #[named(lowercase)]
}
#[proc_macro_derive(NamedType, attributes(named))]
pub fn my_macro(input: TokenStream) -> TokenStream {
let input = parse_macro_input!(input as DeriveInput);
let mut name = input.ident.to_string();
for attr in input.attrs.iter().filter(|a| a.path.is_ident("named")) {
let parser = parser::<Props>(config::RUSTY_META);
let props = match attr.parse_args_with(parser) {
Ok(props) => props,
Err(err) => return err.to_compile_error().into(),
};
if let Some(rename) = props.rename { name = rename; }
if props.lowercase.is_some() { name = name.to_lowercase(); }
}
let ident = &input.ident;
(quote! {
impl NamedType for #ident {
fn name() -> &'static str { #name }
}
}).into()
}
Error handling
Deserialization errors are automatically assigned a "span" (the area of the
source code that could not be parsed) before being returned from parser
and from_stream
as ordinary
syn::Error
s. When that
error is reported to the Rust compiler, the correct regions of code will be
highlighted:
error: unknown field `lowrcase`, expected `rename` or `lowercase`
--> named_type.rs:4:13
|
4 | #[named(lowrcase)]
| ^^^^^^^^
If you use Deserializer
directly, serde_syn will do its best to assign a
span but it is always possible to create an error with no span using serde's
required custom
function.
Limitations
serde_syn is early in development and so still has some gotchyas. For
example, serde_syn will throw an error if you try to deserialize into a
serde_json Value
since it doesn't yet support self-description.
If you find any bugs, have any ideas, or wind up with free time to help random open source projects, please drop by the repository.
Dependencies
~2MB
~44K SLoC