simplify and document

This commit is contained in:
Lukas Wirth 2022-02-22 10:45:29 +01:00
parent 94e59c9c56
commit 8db88df758
6 changed files with 53 additions and 48 deletions

View file

@ -675,15 +675,24 @@ impl<'db> SemanticsImpl<'db> {
ast::MacroCall(mcall) => mcall,
// attribute we failed expansion for earlier, this might be a derive invocation
// so try downmapping the token into the pseudo derive expansion
// see [hir_expand::builtin_attr_macro] for how the pseudo derive expansion works
ast::Meta(meta) => {
let attr = meta.parent_attr()?;
let adt = attr.syntax().parent().and_then(ast::Adt::cast)?;
let call_id = self.with_ctx(|ctx| {
let (_, call_id, _) = ctx.attr_to_derive_macro_call(token.with_value(&adt), token.with_value(attr))?;
let (_, call_id, _) = ctx.attr_to_derive_macro_call(
token.with_value(&adt),
token.with_value(attr),
)?;
Some(call_id)
})?;
let file_id = call_id.as_file();
return process_expansion_for_token(&mut stack,file_id,Some(adt.into()),token.as_ref(),);
return process_expansion_for_token(
&mut stack,
file_id,
Some(adt.into()),
token.as_ref(),
);
},
_ => return None,
}

View file

@ -1170,6 +1170,9 @@ impl DefCollector<'_> {
len = idx;
}
// We treat the #[derive] macro as an attribute call, but we do not resolve it for nameres collection.
// This is just a trick to be able to resolve the input to derives as proper paths.
// Check the comment in [`builtin_attr_macro`].
let call_id = attr_macro_as_call_id(
self.db,
file_ast_id,

View file

@ -82,30 +82,47 @@ fn dummy_attr_expand(
ExpandResult::ok(tt.clone())
}
/// We generate a very specific expansion here, as we do not actually expand the `#[derive]` attribute
/// itself in name res, but we do want to expand it to something for the IDE layer, so that the input
/// derive attributes can be downmapped, and resolved as proper paths.
/// This is basically a hack, that simplifies the hacks we need in a lot of ide layer places to
/// somewhat inconsistently resolve derive attributes.
///
/// As such, we expand `#[derive(Foo, bar::Bar)]` into
/// ```
/// #[Foo]
/// #[bar::Bar]
/// ();
/// ```
/// which allows fallback path resolution in hir::Semantics to properly identify our derives.
/// Since we do not expand the attribute in nameres though, we keep the original item.
///
/// The ideal expansion here would be for the `#[derive]` to re-emit the annotated item and somehow
/// use the input paths in its output as well.
/// But that would bring two problems with it, for one every derive would duplicate the item token tree
/// wasting a lot of memory, and it would also require some way to use a path in a way that makes it
/// always resolve as a derive without nameres recollecting them.
/// So this hacky approach is a lot more friendly for us, though it does require a bit of support in
/// [`hir::Semantics`] to make this work.
fn derive_attr_expand(
db: &dyn AstDatabase,
id: MacroCallId,
tt: &tt::Subtree,
) -> ExpandResult<tt::Subtree> {
// we generate a very specific expansion here, as we do not actually expand the `#[derive]` attribute
// itself in name res, but we do want to expand it to something for the IDE layer, so that the input
// derive attributes can be downmapped, and resolved
// This is basically a hack, to get rid of hacks in the IDE layer that slowly accumulate more and more
// in various places.
// we transform the token tree of `#[derive(Foo, bar::Bar)]` into
// ```
// #[Foo]
// #[bar::Bar]
// ();
// ```
// which allows fallback path resolution in hir::Semantics to properly identify our derives
let loc = db.lookup_intern_macro_call(id);
let derives = match &loc.kind {
MacroCallKind::Attr { attr_args, .. } => &attr_args.0,
_ => return ExpandResult::ok(tt.clone()),
};
let mk_leaf = |char| {
tt::TokenTree::Leaf(tt::Leaf::Punct(tt::Punct {
char,
spacing: tt::Spacing::Alone,
id: tt::TokenId::unspecified(),
}))
};
let mut token_trees = Vec::new();
for (comma, group) in &derives
.token_trees
@ -119,38 +136,13 @@ fn derive_attr_expand(
if comma {
continue;
}
let wrap = |leaf| tt::TokenTree::Leaf(tt::Leaf::Punct(leaf));
token_trees.push(wrap(tt::Punct {
char: '#',
spacing: tt::Spacing::Alone,
id: tt::TokenId::unspecified(),
}));
token_trees.push(wrap(tt::Punct {
char: '[',
spacing: tt::Spacing::Alone,
id: tt::TokenId::unspecified(),
}));
token_trees.push(mk_leaf('#'));
token_trees.push(mk_leaf('['));
token_trees.extend(group.cloned().map(tt::TokenTree::Leaf));
token_trees.push(wrap(tt::Punct {
char: ']',
spacing: tt::Spacing::Alone,
id: tt::TokenId::unspecified(),
}));
token_trees.push(wrap(tt::Punct {
char: '(',
spacing: tt::Spacing::Alone,
id: tt::TokenId::unspecified(),
}));
token_trees.push(wrap(tt::Punct {
char: ')',
spacing: tt::Spacing::Alone,
id: tt::TokenId::unspecified(),
}));
token_trees.push(wrap(tt::Punct {
char: ';',
spacing: tt::Spacing::Alone,
id: tt::TokenId::unspecified(),
}));
token_trees.push(mk_leaf(']'));
}
token_trees.push(mk_leaf('('));
token_trees.push(mk_leaf(')'));
token_trees.push(mk_leaf(';'));
ExpandResult::ok(tt::Subtree { delimiter: tt.delimiter, token_trees })
}

View file

@ -166,6 +166,7 @@ pub enum MacroCallKind {
/// Outer attributes are counted first, then inner attributes. This does not support
/// out-of-line modules, which may have attributes spread across 2 files!
invoc_attr_index: u32,
/// Whether this attribute is the `#[derive]` attribute.
is_derive: bool,
},
}

View file

@ -16,7 +16,7 @@ use hir::{ItemInNs, MacroDef, ModuleDef, Name, Semantics};
use itertools::Itertools;
use syntax::{
ast::{self, make, HasLoopBody},
AstNode, SyntaxKind, SyntaxToken, TokenAtOffset, WalkEvent, T,
AstNode, AstToken, SyntaxKind, SyntaxToken, TokenAtOffset, WalkEvent, T,
};
use crate::{defs::Definition, RootDatabase};

View file

@ -111,7 +111,7 @@ env UPDATE_EXPECT=1 cargo qt
After adding a new inline test you need to run `cargo test -p xtask` and also update the test data as described above.
Note [`api_walkthrough`](https://github.com/rust-analyzer/rust-analyzer/blob/2fb6af89eb794f775de60b82afe56b6f986c2a40/crates/ra_syntax/src/lib.rs#L190-L348)
Note [`api_walkthrough`](https://github.com/rust-analyzer/rust-analyzer/blob/2fb6af89eb794f775de60b82afe56b6f986c2a40/crates/ra_syntax/src/lib.rs#L190-L348)
in particular: it shows off various methods of working with syntax tree.
See [#93](https://github.com/rust-analyzer/rust-analyzer/pull/93) for an example PR which fixes a bug in the grammar.