I'm trying to build a syntax extension that expands an attribute into calls. Before:
#[flame]
fn flamed() {
..
}
After:
fn flamed() {
flame::start_guard("flamed");
..
}
This already works. However, I'd also like it to work if I have the #[flame]
attribute at the crate level (like #![flame]
). Is this possible and if so, how?
@huon's comment
Did you try catching ItemKind::Mod in github.com/llogiq/flamer/blob/… and iterating over its contents (recursively)?
was spot on – I just added a commit that handles mod and trait items by walking them. I'll also probably add code to walk functions to handle inner items and fns.
The code looks like this:
fn flame_item(i: &Item) -> Item {
let base = i.clone();
Item {
node: match i.node {
ItemKind::Mod(ref m) =>
ItemKind::Mod(flame_mod(m)),
ItemKind::Trait(unsafety, ref generic, ref bounds, ref tis) =>
ItemKind::Trait(unsafety,
generic.clone(),
bounds.clone(),
flame_items(tis)),
.. // other item types as usual: items, traitimpls, implitems
_ => return base
},
..base
}
}
fn flame_mod(m: &Mod) -> Mod {
Mod {
inner: m.inner,
items: m.items.iter().map(|i| P(flame_item(i))).collect()
}
}
fn flame_items(items: &[TraitItem]) -> Vec<TraitItem> {
items.iter().map(flame_trait_item).collect()
}