mirror of
https://github.com/rust-lang/rust-analyzer
synced 2024-12-28 05:53:45 +00:00
Speed up search for short associated functions, especially very common identifiers such as new
The search is used by IDE features such as rename and find all references. The search is slow because we need to verify each candidate, and that requires analyzing it; the key to speeding it up is to avoid the analysis where possible. I did that with a bunch of tricks that exploits knowledge about the language and its possibilities. The first key insight is that associated methods may only be referenced in the form `ContainerName::func_name` (parentheses are not necessary!) (Rust doesn't include a way to `use Container::func_name`, and even if it will in the future most usages are likely to stay in that form. Searching for `::` will help only a bit, but searching for `Container` can help considerably, since it is very rare that there will be two identical instances of both a container and a method of it. However, things are not as simple as they sound. In Rust a container can be aliased in multiple ways, and even aliased from different files/modules. If we will try to resolve the alias, we will lose any gain from the textual search (although very common method names such as `new` will still benefit, most will suffer because there are more instances of a container name than its associated item). This is where the key trick enters the picture. The key insight is that there is still a textual property: a container namer cannot be aliased, unless its name is mentioned in the alias declaration, or a name of alias of it is mentioned in the alias declaration. This becomes a fixpoint algorithm: we expand our list of aliases as we collect more and more (possible) aliases, until we eventually reach a fixpoint. A fixpoint is not guaranteed (and we do have guards for the rare cases where it does not happen), but it is almost so: most types have very few aliases, if at all. We do use some semantic information while analyzing aliases. It's a balance: too much semantic analysis, and the search will become slow. But too few of it, and we will bring many incorrect aliases to our list, and risk it expands and expands and never reach a fixpoint. At the end, based on benchmarks, it seems worth to do a lot to avoid adding an alias (but not too much), while it is worth to do a lot to avoid the need to semantically analyze func_name matches (but again, not too much). After we collected our list of aliases, we filter matches based on this list. Only if a match can be real, we do semantic analysis for it. The results are promising: searching for all references on `new()` in `base-db` in the rust-analyzer repository, which previously took around 60 seconds, now takes as least as two seconds and a half (roughly), while searching for `Vec::new()`, almost an upper bound to how much a symbol can be used, that used to take 7-9 minutes(!) now completes in 100-120 seconds, and with less than half of non-verified results (aka. false positives). This is the less strictly correct (but faster) of this patch; it can miss some (rare) cases (there is a test for that - `goto_ref_on_short_associated_function_complicated_type_magic_can_confuse_our_logic()`). There is another branch that have no false negatives but is slower to search (`Vec::new()` never reaches a fixpoint in aliases collection there). I believe it is possible to create a strategy that will have the best of both worlds, but it will involve significant complexity and I didn't bother, especially considering that in the vast majority of the searches the other branch will be more than enough. But all in all, I decided to bring this branch (of course if the maintainers will agree), since our search is already not 100% accurate (it misses macros), and I believe there is value in the additional perf.
This commit is contained in:
parent
2fa0d4e2a8
commit
d89bdd9b83
2 changed files with 657 additions and 80 deletions
|
@ -4,19 +4,24 @@
|
||||||
//! get a super-set of matches. Then, we confirm each match using precise
|
//! get a super-set of matches. Then, we confirm each match using precise
|
||||||
//! name resolution.
|
//! name resolution.
|
||||||
|
|
||||||
use std::cell::LazyCell;
|
|
||||||
use std::mem;
|
use std::mem;
|
||||||
|
use std::{cell::LazyCell, cmp::Reverse};
|
||||||
|
|
||||||
use base_db::{salsa::Database, SourceDatabase, SourceRootDatabase};
|
use base_db::{salsa::Database, SourceDatabase, SourceRootDatabase};
|
||||||
use hir::{
|
use hir::{
|
||||||
sym, AsAssocItem, DefWithBody, FileRange, HasAttrs, HasSource, HirFileIdExt, InFile,
|
sym, Adt, AsAssocItem, DefWithBody, FileRange, FileRangeWrapper, HasAttrs, HasContainer,
|
||||||
InRealFile, ModuleSource, PathResolution, Semantics, Visibility,
|
HasSource, HirFileIdExt, InFile, InFileWrapper, InRealFile, ItemContainer, ModuleSource,
|
||||||
|
PathResolution, Semantics, Visibility,
|
||||||
};
|
};
|
||||||
use memchr::memmem::Finder;
|
use memchr::memmem::Finder;
|
||||||
use parser::SyntaxKind;
|
use parser::SyntaxKind;
|
||||||
use rustc_hash::FxHashMap;
|
use rustc_hash::{FxHashMap, FxHashSet};
|
||||||
use span::EditionedFileId;
|
use span::EditionedFileId;
|
||||||
use syntax::{ast, match_ast, AstNode, AstToken, SyntaxElement, TextRange, TextSize, ToSmolStr};
|
use syntax::{
|
||||||
|
ast::{self, HasName},
|
||||||
|
match_ast, AstNode, AstToken, SmolStr, SyntaxElement, SyntaxNode, TextRange, TextSize,
|
||||||
|
ToSmolStr,
|
||||||
|
};
|
||||||
use triomphe::Arc;
|
use triomphe::Arc;
|
||||||
|
|
||||||
use crate::{
|
use crate::{
|
||||||
|
@ -442,6 +447,398 @@ impl<'a> FindUsages<'a> {
|
||||||
res
|
res
|
||||||
}
|
}
|
||||||
|
|
||||||
|
fn scope_files<'b>(
|
||||||
|
db: &'b RootDatabase,
|
||||||
|
scope: &'b SearchScope,
|
||||||
|
) -> impl Iterator<Item = (Arc<str>, EditionedFileId, TextRange)> + 'b {
|
||||||
|
scope.entries.iter().map(|(&file_id, &search_range)| {
|
||||||
|
let text = db.file_text(file_id.file_id());
|
||||||
|
let search_range =
|
||||||
|
search_range.unwrap_or_else(|| TextRange::up_to(TextSize::of(&*text)));
|
||||||
|
|
||||||
|
(text, file_id, search_range)
|
||||||
|
})
|
||||||
|
}
|
||||||
|
|
||||||
|
fn match_indices<'b>(
|
||||||
|
text: &'b str,
|
||||||
|
finder: &'b Finder<'b>,
|
||||||
|
search_range: TextRange,
|
||||||
|
) -> impl Iterator<Item = TextSize> + 'b {
|
||||||
|
finder.find_iter(text.as_bytes()).filter_map(move |idx| {
|
||||||
|
let offset: TextSize = idx.try_into().unwrap();
|
||||||
|
if !search_range.contains_inclusive(offset) {
|
||||||
|
return None;
|
||||||
|
}
|
||||||
|
// If this is not a word boundary, that means this is only part of an identifier,
|
||||||
|
// so it can't be what we're looking for.
|
||||||
|
// This speeds up short identifiers significantly.
|
||||||
|
if text[..idx]
|
||||||
|
.chars()
|
||||||
|
.next_back()
|
||||||
|
.is_some_and(|ch| matches!(ch, 'A'..='Z' | 'a'..='z' | '_'))
|
||||||
|
|| text[idx + finder.needle().len()..]
|
||||||
|
.chars()
|
||||||
|
.next()
|
||||||
|
.is_some_and(|ch| matches!(ch, 'A'..='Z' | 'a'..='z' | '_' | '0'..='9'))
|
||||||
|
{
|
||||||
|
return None;
|
||||||
|
}
|
||||||
|
Some(offset)
|
||||||
|
})
|
||||||
|
}
|
||||||
|
|
||||||
|
fn find_nodes<'b>(
|
||||||
|
sema: &'b Semantics<'_, RootDatabase>,
|
||||||
|
name: &str,
|
||||||
|
node: &syntax::SyntaxNode,
|
||||||
|
offset: TextSize,
|
||||||
|
) -> impl Iterator<Item = SyntaxNode> + 'b {
|
||||||
|
node.token_at_offset(offset)
|
||||||
|
.find(|it| {
|
||||||
|
// `name` is stripped of raw ident prefix. See the comment on name retrieval below.
|
||||||
|
it.text().trim_start_matches("r#") == name
|
||||||
|
})
|
||||||
|
.into_iter()
|
||||||
|
.flat_map(move |token| {
|
||||||
|
sema.descend_into_macros_exact_if_in_macro(token)
|
||||||
|
.into_iter()
|
||||||
|
.filter_map(|it| it.parent())
|
||||||
|
})
|
||||||
|
}
|
||||||
|
|
||||||
|
/// Performs a special fast search for associated functions. This is mainly intended
|
||||||
|
/// to speed up `new()` which can take a long time.
|
||||||
|
///
|
||||||
|
/// The trick is instead of searching for `func_name` search for `TypeThatContainsContainerName::func_name`.
|
||||||
|
/// We cannot search exactly that (not even in tokens), because `ContainerName` may be aliased.
|
||||||
|
/// Instead, we perform a textual search for `ContainerName`. Then, we look for all cases where
|
||||||
|
/// `ContainerName` may be aliased (that includes `use ContainerName as Xyz` and
|
||||||
|
/// `type Xyz = ContainerName`). We collect a list of all possible aliases of `ContainerName`.
|
||||||
|
/// The list can have false positives (because there may be multiple types named `ContainerName`),
|
||||||
|
/// but it cannot have false negatives. Then, we look for `TypeThatContainsContainerNameOrAnyAlias::func_name`.
|
||||||
|
/// Those that will be found are of high chance to be actual hits (of course, we will need to verify
|
||||||
|
/// that).
|
||||||
|
///
|
||||||
|
/// Returns true if completed the search.
|
||||||
|
// FIXME: Extend this to other cases, such as associated types/consts/enum variants (note those can be `use`d).
|
||||||
|
fn short_associated_function_fast_search(
|
||||||
|
&self,
|
||||||
|
sink: &mut dyn FnMut(EditionedFileId, FileReference) -> bool,
|
||||||
|
search_scope: &SearchScope,
|
||||||
|
name: &str,
|
||||||
|
) -> bool {
|
||||||
|
let _p = tracing::info_span!("short_associated_function_fast_search").entered();
|
||||||
|
|
||||||
|
let container = (|| {
|
||||||
|
let Definition::Function(function) = self.def else {
|
||||||
|
return None;
|
||||||
|
};
|
||||||
|
if function.has_self_param(self.sema.db) {
|
||||||
|
return None;
|
||||||
|
}
|
||||||
|
match function.container(self.sema.db) {
|
||||||
|
// Only freestanding `impl`s qualify; methods from trait
|
||||||
|
// can be called from within subtraits and bounds.
|
||||||
|
ItemContainer::Impl(impl_) => {
|
||||||
|
let has_trait = impl_.trait_(self.sema.db).is_some();
|
||||||
|
if has_trait {
|
||||||
|
return None;
|
||||||
|
}
|
||||||
|
let adt = impl_.self_ty(self.sema.db).as_adt()?;
|
||||||
|
Some(adt)
|
||||||
|
}
|
||||||
|
_ => None,
|
||||||
|
}
|
||||||
|
})();
|
||||||
|
let Some(container) = container else {
|
||||||
|
return false;
|
||||||
|
};
|
||||||
|
|
||||||
|
fn has_any_name(node: &SyntaxNode, mut predicate: impl FnMut(&str) -> bool) -> bool {
|
||||||
|
node.descendants().any(|node| {
|
||||||
|
match_ast! {
|
||||||
|
match node {
|
||||||
|
ast::Name(it) => predicate(it.text().trim_start_matches("r#")),
|
||||||
|
ast::NameRef(it) => predicate(it.text().trim_start_matches("r#")),
|
||||||
|
_ => false
|
||||||
|
}
|
||||||
|
}
|
||||||
|
})
|
||||||
|
}
|
||||||
|
|
||||||
|
// This is a fixpoint algorithm with O(number of aliases), but most types have no or few aliases,
|
||||||
|
// so this should stay fast.
|
||||||
|
//
|
||||||
|
/// Returns `(aliases, ranges_where_Self_can_refer_to_our_type)`.
|
||||||
|
fn collect_possible_aliases(
|
||||||
|
sema: &Semantics<'_, RootDatabase>,
|
||||||
|
container: Adt,
|
||||||
|
) -> Option<(FxHashSet<SmolStr>, Vec<FileRangeWrapper<EditionedFileId>>)> {
|
||||||
|
fn insert_type_alias(
|
||||||
|
db: &RootDatabase,
|
||||||
|
to_process: &mut Vec<(SmolStr, SearchScope)>,
|
||||||
|
alias_name: &str,
|
||||||
|
def: Definition,
|
||||||
|
) {
|
||||||
|
let alias = alias_name.trim_start_matches("r#").to_smolstr();
|
||||||
|
tracing::debug!("found alias: {alias}");
|
||||||
|
to_process.push((alias, def.search_scope(db)));
|
||||||
|
}
|
||||||
|
|
||||||
|
let _p = tracing::info_span!("collect_possible_aliases").entered();
|
||||||
|
|
||||||
|
let db = sema.db;
|
||||||
|
let container_name = container.name(db).unescaped().display(db).to_smolstr();
|
||||||
|
let search_scope = Definition::from(container).search_scope(db);
|
||||||
|
let mut seen = FxHashSet::default();
|
||||||
|
let mut completed = FxHashSet::default();
|
||||||
|
let mut to_process = vec![(container_name, search_scope)];
|
||||||
|
let mut is_possibly_self = Vec::new();
|
||||||
|
let mut total_files_searched = 0;
|
||||||
|
|
||||||
|
while let Some((current_to_process, current_to_process_search_scope)) = to_process.pop()
|
||||||
|
{
|
||||||
|
let is_alias = |alias: &ast::TypeAlias| {
|
||||||
|
let def = sema.to_def(alias)?;
|
||||||
|
let ty = def.ty(db);
|
||||||
|
let is_alias = ty.as_adt()? == container;
|
||||||
|
is_alias.then_some(def)
|
||||||
|
};
|
||||||
|
|
||||||
|
let finder = Finder::new(current_to_process.as_bytes());
|
||||||
|
for (file_text, file_id, search_range) in
|
||||||
|
FindUsages::scope_files(db, ¤t_to_process_search_scope)
|
||||||
|
{
|
||||||
|
let tree = LazyCell::new(move || sema.parse(file_id).syntax().clone());
|
||||||
|
|
||||||
|
for offset in FindUsages::match_indices(&file_text, &finder, search_range) {
|
||||||
|
let usages =
|
||||||
|
FindUsages::find_nodes(sema, ¤t_to_process, &tree, offset)
|
||||||
|
.filter(|it| {
|
||||||
|
matches!(it.kind(), SyntaxKind::NAME | SyntaxKind::NAME_REF)
|
||||||
|
});
|
||||||
|
for usage in usages {
|
||||||
|
if let Some(alias) = usage.parent().and_then(|it| {
|
||||||
|
let path = ast::PathSegment::cast(it)?.parent_path();
|
||||||
|
let use_tree = ast::UseTree::cast(path.syntax().parent()?)?;
|
||||||
|
use_tree.rename()?.name()
|
||||||
|
}) {
|
||||||
|
if seen.insert(InFileWrapper::new(
|
||||||
|
file_id,
|
||||||
|
alias.syntax().text_range(),
|
||||||
|
)) {
|
||||||
|
tracing::debug!("found alias: {alias}");
|
||||||
|
// FIXME: `use`s have no easy way to determine their search scope, but they are rare.
|
||||||
|
to_process.push((
|
||||||
|
alias.text().to_smolstr(),
|
||||||
|
current_to_process_search_scope.clone(),
|
||||||
|
));
|
||||||
|
}
|
||||||
|
} else if let Some(alias) =
|
||||||
|
usage.ancestors().find_map(ast::TypeAlias::cast)
|
||||||
|
{
|
||||||
|
if let Some(name) = alias.name() {
|
||||||
|
if seen.insert(InFileWrapper::new(
|
||||||
|
file_id,
|
||||||
|
name.syntax().text_range(),
|
||||||
|
)) {
|
||||||
|
if let Some(def) = is_alias(&alias) {
|
||||||
|
insert_type_alias(
|
||||||
|
sema.db,
|
||||||
|
&mut to_process,
|
||||||
|
name.text().as_str(),
|
||||||
|
def.into(),
|
||||||
|
);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
// We need to account for `Self`. It can only refer to our type inside an impl.
|
||||||
|
let impl_ = 'impl_: {
|
||||||
|
for ancestor in usage.ancestors() {
|
||||||
|
if let Some(parent) = ancestor.parent() {
|
||||||
|
if let Some(parent) = ast::Impl::cast(parent) {
|
||||||
|
// Only if the GENERIC_PARAM_LIST is directly under impl, otherwise it may be in the self ty.
|
||||||
|
if matches!(
|
||||||
|
ancestor.kind(),
|
||||||
|
SyntaxKind::ASSOC_ITEM_LIST
|
||||||
|
| SyntaxKind::WHERE_CLAUSE
|
||||||
|
| SyntaxKind::GENERIC_PARAM_LIST
|
||||||
|
) {
|
||||||
|
break;
|
||||||
|
}
|
||||||
|
if parent
|
||||||
|
.trait_()
|
||||||
|
.is_some_and(|trait_| *trait_.syntax() == ancestor)
|
||||||
|
{
|
||||||
|
break;
|
||||||
|
}
|
||||||
|
|
||||||
|
// Otherwise, found an impl where its self ty may be our type.
|
||||||
|
break 'impl_ Some(parent);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
|
None
|
||||||
|
};
|
||||||
|
(|| {
|
||||||
|
let impl_ = impl_?;
|
||||||
|
is_possibly_self.push(sema.original_range(impl_.syntax()));
|
||||||
|
let assoc_items = impl_.assoc_item_list()?;
|
||||||
|
let type_aliases = assoc_items
|
||||||
|
.syntax()
|
||||||
|
.descendants()
|
||||||
|
.filter_map(ast::TypeAlias::cast);
|
||||||
|
for type_alias in type_aliases {
|
||||||
|
let Some(ty) = type_alias.ty() else { continue };
|
||||||
|
let Some(name) = type_alias.name() else { continue };
|
||||||
|
let contains_self = ty
|
||||||
|
.syntax()
|
||||||
|
.descendants_with_tokens()
|
||||||
|
.any(|node| node.kind() == SyntaxKind::SELF_TYPE_KW);
|
||||||
|
if !contains_self {
|
||||||
|
continue;
|
||||||
|
}
|
||||||
|
if seen.insert(InFileWrapper::new(
|
||||||
|
file_id,
|
||||||
|
name.syntax().text_range(),
|
||||||
|
)) {
|
||||||
|
if let Some(def) = is_alias(&type_alias) {
|
||||||
|
insert_type_alias(
|
||||||
|
sema.db,
|
||||||
|
&mut to_process,
|
||||||
|
name.text().as_str(),
|
||||||
|
def.into(),
|
||||||
|
);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
|
Some(())
|
||||||
|
})();
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
completed.insert(current_to_process);
|
||||||
|
|
||||||
|
total_files_searched += current_to_process_search_scope.entries.len();
|
||||||
|
// FIXME: Maybe this needs to be relative to the project size, or at least to the initial search scope?
|
||||||
|
if total_files_searched > 20_000 && completed.len() > 100 {
|
||||||
|
// This case is extremely unlikely (even searching for `Vec::new()` on rust-analyzer does not enter
|
||||||
|
// here - it searches less than 10,000 files, and it does so in five seconds), but if we get here,
|
||||||
|
// we at a risk of entering an almost-infinite loop of growing the aliases list. So just stop and
|
||||||
|
// let normal search handle this case.
|
||||||
|
tracing::info!(aliases_count = %completed.len(), "too much aliases; leaving fast path");
|
||||||
|
return None;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
// Impls can contain each other, so we need to deduplicate their ranges.
|
||||||
|
is_possibly_self.sort_unstable_by_key(|position| {
|
||||||
|
(position.file_id, position.range.start(), Reverse(position.range.end()))
|
||||||
|
});
|
||||||
|
is_possibly_self.dedup_by(|pos2, pos1| {
|
||||||
|
pos1.file_id == pos2.file_id
|
||||||
|
&& pos1.range.start() <= pos2.range.start()
|
||||||
|
&& pos1.range.end() >= pos2.range.end()
|
||||||
|
});
|
||||||
|
|
||||||
|
tracing::info!(aliases_count = %completed.len(), "aliases search completed");
|
||||||
|
|
||||||
|
Some((completed, is_possibly_self))
|
||||||
|
}
|
||||||
|
|
||||||
|
fn search(
|
||||||
|
this: &FindUsages<'_>,
|
||||||
|
finder: &Finder<'_>,
|
||||||
|
name: &str,
|
||||||
|
files: impl Iterator<Item = (Arc<str>, EditionedFileId, TextRange)>,
|
||||||
|
mut container_predicate: impl FnMut(
|
||||||
|
&SyntaxNode,
|
||||||
|
InFileWrapper<EditionedFileId, TextRange>,
|
||||||
|
) -> bool,
|
||||||
|
sink: &mut dyn FnMut(EditionedFileId, FileReference) -> bool,
|
||||||
|
) {
|
||||||
|
for (file_text, file_id, search_range) in files {
|
||||||
|
let tree = LazyCell::new(move || this.sema.parse(file_id).syntax().clone());
|
||||||
|
|
||||||
|
for offset in FindUsages::match_indices(&file_text, finder, search_range) {
|
||||||
|
let usages = FindUsages::find_nodes(this.sema, name, &tree, offset)
|
||||||
|
.filter_map(ast::NameRef::cast);
|
||||||
|
for usage in usages {
|
||||||
|
let found_usage = usage
|
||||||
|
.syntax()
|
||||||
|
.parent()
|
||||||
|
.and_then(ast::PathSegment::cast)
|
||||||
|
.map(|path_segment| {
|
||||||
|
container_predicate(
|
||||||
|
path_segment.parent_path().syntax(),
|
||||||
|
InFileWrapper::new(file_id, usage.syntax().text_range()),
|
||||||
|
)
|
||||||
|
})
|
||||||
|
.unwrap_or(false);
|
||||||
|
if found_usage {
|
||||||
|
this.found_name_ref(&usage, sink);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
let Some((container_possible_aliases, is_possibly_self)) =
|
||||||
|
collect_possible_aliases(self.sema, container)
|
||||||
|
else {
|
||||||
|
return false;
|
||||||
|
};
|
||||||
|
|
||||||
|
// FIXME: If Rust ever gains the ability to `use Struct::method` we'll also need to account for free
|
||||||
|
// functions.
|
||||||
|
let finder = Finder::new(name.as_bytes());
|
||||||
|
// The search for `Self` may return duplicate results with `ContainerName`, so deduplicate them.
|
||||||
|
let mut self_positions = FxHashSet::default();
|
||||||
|
tracing::info_span!("Self_search").in_scope(|| {
|
||||||
|
search(
|
||||||
|
self,
|
||||||
|
&finder,
|
||||||
|
name,
|
||||||
|
is_possibly_self.into_iter().map(|position| {
|
||||||
|
(
|
||||||
|
self.sema.db.file_text(position.file_id.file_id()),
|
||||||
|
position.file_id,
|
||||||
|
position.range,
|
||||||
|
)
|
||||||
|
}),
|
||||||
|
|path, name_position| {
|
||||||
|
let has_self = path
|
||||||
|
.descendants_with_tokens()
|
||||||
|
.any(|node| node.kind() == SyntaxKind::SELF_TYPE_KW);
|
||||||
|
if has_self {
|
||||||
|
self_positions.insert(name_position);
|
||||||
|
}
|
||||||
|
has_self
|
||||||
|
},
|
||||||
|
sink,
|
||||||
|
)
|
||||||
|
});
|
||||||
|
tracing::info_span!("aliases_search").in_scope(|| {
|
||||||
|
search(
|
||||||
|
self,
|
||||||
|
&finder,
|
||||||
|
name,
|
||||||
|
FindUsages::scope_files(self.sema.db, search_scope),
|
||||||
|
|path, name_position| {
|
||||||
|
has_any_name(path, |name| container_possible_aliases.contains(name))
|
||||||
|
&& !self_positions.contains(&name_position)
|
||||||
|
},
|
||||||
|
sink,
|
||||||
|
)
|
||||||
|
});
|
||||||
|
|
||||||
|
true
|
||||||
|
}
|
||||||
|
|
||||||
pub fn search(&self, sink: &mut dyn FnMut(EditionedFileId, FileReference) -> bool) {
|
pub fn search(&self, sink: &mut dyn FnMut(EditionedFileId, FileReference) -> bool) {
|
||||||
let _p = tracing::info_span!("FindUsages:search").entered();
|
let _p = tracing::info_span!("FindUsages:search").entered();
|
||||||
let sema = self.sema;
|
let sema = self.sema;
|
||||||
|
@ -488,77 +885,23 @@ impl<'a> FindUsages<'a> {
|
||||||
Some(s) => s.as_str(),
|
Some(s) => s.as_str(),
|
||||||
None => return,
|
None => return,
|
||||||
};
|
};
|
||||||
|
|
||||||
|
// FIXME: This should probably depend on the number of the results (specifically, the number of false results).
|
||||||
|
if name.len() <= 7 && self.short_associated_function_fast_search(sink, &search_scope, name)
|
||||||
|
{
|
||||||
|
return;
|
||||||
|
}
|
||||||
|
|
||||||
let finder = &Finder::new(name);
|
let finder = &Finder::new(name);
|
||||||
let include_self_kw_refs =
|
let include_self_kw_refs =
|
||||||
self.include_self_kw_refs.as_ref().map(|ty| (ty, Finder::new("Self")));
|
self.include_self_kw_refs.as_ref().map(|ty| (ty, Finder::new("Self")));
|
||||||
|
|
||||||
// for<'a> |text: &'a str, name: &'a str, search_range: TextRange| -> impl Iterator<Item = TextSize> + 'a { ... }
|
for (text, file_id, search_range) in Self::scope_files(sema.db, &search_scope) {
|
||||||
fn match_indices<'a>(
|
|
||||||
text: &'a str,
|
|
||||||
finder: &'a Finder<'a>,
|
|
||||||
search_range: TextRange,
|
|
||||||
) -> impl Iterator<Item = TextSize> + 'a {
|
|
||||||
finder.find_iter(text.as_bytes()).filter_map(move |idx| {
|
|
||||||
let offset: TextSize = idx.try_into().unwrap();
|
|
||||||
if !search_range.contains_inclusive(offset) {
|
|
||||||
return None;
|
|
||||||
}
|
|
||||||
// If this is not a word boundary, that means this is only part of an identifier,
|
|
||||||
// so it can't be what we're looking for.
|
|
||||||
// This speeds up short identifiers significantly.
|
|
||||||
if text[..idx]
|
|
||||||
.chars()
|
|
||||||
.next_back()
|
|
||||||
.is_some_and(|ch| matches!(ch, 'A'..='Z' | 'a'..='z' | '_'))
|
|
||||||
|| text[idx + finder.needle().len()..]
|
|
||||||
.chars()
|
|
||||||
.next()
|
|
||||||
.is_some_and(|ch| matches!(ch, 'A'..='Z' | 'a'..='z' | '_' | '0'..='9'))
|
|
||||||
{
|
|
||||||
return None;
|
|
||||||
}
|
|
||||||
Some(offset)
|
|
||||||
})
|
|
||||||
}
|
|
||||||
|
|
||||||
// for<'a> |scope: &'a SearchScope| -> impl Iterator<Item = (Arc<String>, EditionedFileId, TextRange)> + 'a { ... }
|
|
||||||
fn scope_files<'a>(
|
|
||||||
sema: &'a Semantics<'_, RootDatabase>,
|
|
||||||
scope: &'a SearchScope,
|
|
||||||
) -> impl Iterator<Item = (Arc<str>, EditionedFileId, TextRange)> + 'a {
|
|
||||||
scope.entries.iter().map(|(&file_id, &search_range)| {
|
|
||||||
let text = sema.db.file_text(file_id.file_id());
|
|
||||||
let search_range =
|
|
||||||
search_range.unwrap_or_else(|| TextRange::up_to(TextSize::of(&*text)));
|
|
||||||
|
|
||||||
(text, file_id, search_range)
|
|
||||||
})
|
|
||||||
}
|
|
||||||
|
|
||||||
let find_nodes = move |name: &str, node: &syntax::SyntaxNode, offset: TextSize| {
|
|
||||||
node.token_at_offset(offset)
|
|
||||||
.find(|it| {
|
|
||||||
// `name` is stripped of raw ident prefix. See the comment on name retrieval above.
|
|
||||||
it.text().trim_start_matches("r#") == name
|
|
||||||
})
|
|
||||||
.into_iter()
|
|
||||||
.flat_map(move |token| {
|
|
||||||
// FIXME: There should be optimization potential here
|
|
||||||
// Currently we try to descend everything we find which
|
|
||||||
// means we call `Semantics::descend_into_macros` on
|
|
||||||
// every textual hit. That function is notoriously
|
|
||||||
// expensive even for things that do not get down mapped
|
|
||||||
// into macros.
|
|
||||||
sema.descend_into_macros_exact(token).into_iter().filter_map(|it| it.parent())
|
|
||||||
})
|
|
||||||
};
|
|
||||||
|
|
||||||
for (text, file_id, search_range) in scope_files(sema, &search_scope) {
|
|
||||||
self.sema.db.unwind_if_cancelled();
|
self.sema.db.unwind_if_cancelled();
|
||||||
let tree = LazyCell::new(move || sema.parse(file_id).syntax().clone());
|
let tree = LazyCell::new(move || sema.parse(file_id).syntax().clone());
|
||||||
|
|
||||||
// Search for occurrences of the items name
|
// Search for occurrences of the items name
|
||||||
for offset in match_indices(&text, finder, search_range) {
|
for offset in Self::match_indices(&text, finder, search_range) {
|
||||||
tree.token_at_offset(offset).for_each(|token| {
|
tree.token_at_offset(offset).for_each(|token| {
|
||||||
let Some(str_token) = ast::String::cast(token.clone()) else { return };
|
let Some(str_token) = ast::String::cast(token.clone()) else { return };
|
||||||
if let Some((range, nameres)) =
|
if let Some((range, nameres)) =
|
||||||
|
@ -568,7 +911,9 @@ impl<'a> FindUsages<'a> {
|
||||||
}
|
}
|
||||||
});
|
});
|
||||||
|
|
||||||
for name in find_nodes(name, &tree, offset).filter_map(ast::NameLike::cast) {
|
for name in
|
||||||
|
Self::find_nodes(sema, name, &tree, offset).filter_map(ast::NameLike::cast)
|
||||||
|
{
|
||||||
if match name {
|
if match name {
|
||||||
ast::NameLike::NameRef(name_ref) => self.found_name_ref(&name_ref, sink),
|
ast::NameLike::NameRef(name_ref) => self.found_name_ref(&name_ref, sink),
|
||||||
ast::NameLike::Name(name) => self.found_name(&name, sink),
|
ast::NameLike::Name(name) => self.found_name(&name, sink),
|
||||||
|
@ -580,8 +925,9 @@ impl<'a> FindUsages<'a> {
|
||||||
}
|
}
|
||||||
// Search for occurrences of the `Self` referring to our type
|
// Search for occurrences of the `Self` referring to our type
|
||||||
if let Some((self_ty, finder)) = &include_self_kw_refs {
|
if let Some((self_ty, finder)) = &include_self_kw_refs {
|
||||||
for offset in match_indices(&text, finder, search_range) {
|
for offset in Self::match_indices(&text, finder, search_range) {
|
||||||
for name_ref in find_nodes("Self", &tree, offset).filter_map(ast::NameRef::cast)
|
for name_ref in
|
||||||
|
Self::find_nodes(sema, "Self", &tree, offset).filter_map(ast::NameRef::cast)
|
||||||
{
|
{
|
||||||
if self.found_self_ty_name_ref(self_ty, &name_ref, sink) {
|
if self.found_self_ty_name_ref(self_ty, &name_ref, sink) {
|
||||||
return;
|
return;
|
||||||
|
@ -599,13 +945,13 @@ impl<'a> FindUsages<'a> {
|
||||||
let is_crate_root = module.is_crate_root().then(|| Finder::new("crate"));
|
let is_crate_root = module.is_crate_root().then(|| Finder::new("crate"));
|
||||||
let finder = &Finder::new("super");
|
let finder = &Finder::new("super");
|
||||||
|
|
||||||
for (text, file_id, search_range) in scope_files(sema, &scope) {
|
for (text, file_id, search_range) in Self::scope_files(sema.db, &scope) {
|
||||||
self.sema.db.unwind_if_cancelled();
|
self.sema.db.unwind_if_cancelled();
|
||||||
let tree = LazyCell::new(move || sema.parse(file_id).syntax().clone());
|
let tree = LazyCell::new(move || sema.parse(file_id).syntax().clone());
|
||||||
|
|
||||||
for offset in match_indices(&text, finder, search_range) {
|
for offset in Self::match_indices(&text, finder, search_range) {
|
||||||
for name_ref in
|
for name_ref in Self::find_nodes(sema, "super", &tree, offset)
|
||||||
find_nodes("super", &tree, offset).filter_map(ast::NameRef::cast)
|
.filter_map(ast::NameRef::cast)
|
||||||
{
|
{
|
||||||
if self.found_name_ref(&name_ref, sink) {
|
if self.found_name_ref(&name_ref, sink) {
|
||||||
return;
|
return;
|
||||||
|
@ -613,9 +959,9 @@ impl<'a> FindUsages<'a> {
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
if let Some(finder) = &is_crate_root {
|
if let Some(finder) = &is_crate_root {
|
||||||
for offset in match_indices(&text, finder, search_range) {
|
for offset in Self::match_indices(&text, finder, search_range) {
|
||||||
for name_ref in
|
for name_ref in Self::find_nodes(sema, "crate", &tree, offset)
|
||||||
find_nodes("crate", &tree, offset).filter_map(ast::NameRef::cast)
|
.filter_map(ast::NameRef::cast)
|
||||||
{
|
{
|
||||||
if self.found_name_ref(&name_ref, sink) {
|
if self.found_name_ref(&name_ref, sink) {
|
||||||
return;
|
return;
|
||||||
|
@ -656,8 +1002,9 @@ impl<'a> FindUsages<'a> {
|
||||||
let tree = LazyCell::new(|| sema.parse(file_id).syntax().clone());
|
let tree = LazyCell::new(|| sema.parse(file_id).syntax().clone());
|
||||||
let finder = &Finder::new("self");
|
let finder = &Finder::new("self");
|
||||||
|
|
||||||
for offset in match_indices(&text, finder, search_range) {
|
for offset in Self::match_indices(&text, finder, search_range) {
|
||||||
for name_ref in find_nodes("self", &tree, offset).filter_map(ast::NameRef::cast)
|
for name_ref in
|
||||||
|
Self::find_nodes(sema, "self", &tree, offset).filter_map(ast::NameRef::cast)
|
||||||
{
|
{
|
||||||
if self.found_self_module_name_ref(&name_ref, sink) {
|
if self.found_self_module_name_ref(&name_ref, sink) {
|
||||||
return;
|
return;
|
||||||
|
|
|
@ -2510,4 +2510,234 @@ fn main() {
|
||||||
"#]],
|
"#]],
|
||||||
)
|
)
|
||||||
}
|
}
|
||||||
|
|
||||||
|
// The following are tests for short_associated_function_fast_search() in crates/ide-db/src/search.rs, because find all references
|
||||||
|
// use `FindUsages` and I found it easy to test it here.
|
||||||
|
|
||||||
|
#[test]
|
||||||
|
fn goto_ref_on_short_associated_function() {
|
||||||
|
check(
|
||||||
|
r#"
|
||||||
|
struct Foo;
|
||||||
|
impl Foo {
|
||||||
|
fn new$0() {}
|
||||||
|
}
|
||||||
|
|
||||||
|
fn bar() {
|
||||||
|
Foo::new();
|
||||||
|
}
|
||||||
|
fn baz() {
|
||||||
|
Foo::new;
|
||||||
|
}
|
||||||
|
"#,
|
||||||
|
expect![[r#"
|
||||||
|
new Function FileId(0) 27..38 30..33
|
||||||
|
|
||||||
|
FileId(0) 62..65
|
||||||
|
FileId(0) 91..94
|
||||||
|
"#]],
|
||||||
|
);
|
||||||
|
}
|
||||||
|
|
||||||
|
#[test]
|
||||||
|
fn goto_ref_on_short_associated_function_with_aliases() {
|
||||||
|
check(
|
||||||
|
r#"
|
||||||
|
//- /lib.rs
|
||||||
|
mod a;
|
||||||
|
mod b;
|
||||||
|
|
||||||
|
struct Foo;
|
||||||
|
impl Foo {
|
||||||
|
fn new$0() {}
|
||||||
|
}
|
||||||
|
|
||||||
|
fn bar() {
|
||||||
|
b::c::Baz::new();
|
||||||
|
}
|
||||||
|
|
||||||
|
//- /a.rs
|
||||||
|
use crate::Foo as Bar;
|
||||||
|
|
||||||
|
fn baz() { Bar::new(); }
|
||||||
|
fn quux() { <super::b::Other as super::b::Trait>::Assoc::new(); }
|
||||||
|
|
||||||
|
//- /b.rs
|
||||||
|
pub(crate) mod c;
|
||||||
|
|
||||||
|
pub(crate) struct Other;
|
||||||
|
pub(crate) trait Trait {
|
||||||
|
type Assoc;
|
||||||
|
}
|
||||||
|
impl Trait for Other {
|
||||||
|
type Assoc = super::Foo;
|
||||||
|
}
|
||||||
|
|
||||||
|
//- /b/c.rs
|
||||||
|
type Itself<T> = T;
|
||||||
|
pub(in super::super) type Baz = Itself<crate::Foo>;
|
||||||
|
"#,
|
||||||
|
expect![[r#"
|
||||||
|
new Function FileId(0) 42..53 45..48
|
||||||
|
|
||||||
|
FileId(0) 83..86
|
||||||
|
FileId(1) 40..43
|
||||||
|
FileId(1) 106..109
|
||||||
|
"#]],
|
||||||
|
);
|
||||||
|
}
|
||||||
|
|
||||||
|
#[test]
|
||||||
|
fn goto_ref_on_short_associated_function_self_works() {
|
||||||
|
check(
|
||||||
|
r#"
|
||||||
|
//- /lib.rs
|
||||||
|
mod module;
|
||||||
|
|
||||||
|
struct Foo;
|
||||||
|
impl Foo {
|
||||||
|
fn new$0() {}
|
||||||
|
fn bar() { Self::new(); }
|
||||||
|
}
|
||||||
|
trait Trait {
|
||||||
|
type Assoc;
|
||||||
|
fn baz();
|
||||||
|
}
|
||||||
|
impl Trait for Foo {
|
||||||
|
type Assoc = Self;
|
||||||
|
fn baz() { Self::new(); }
|
||||||
|
}
|
||||||
|
|
||||||
|
//- /module.rs
|
||||||
|
impl super::Foo {
|
||||||
|
fn quux() { Self::new(); }
|
||||||
|
}
|
||||||
|
fn foo() { <super::Foo as super::Trait>::Assoc::new(); }
|
||||||
|
"#,
|
||||||
|
expect![[r#"
|
||||||
|
new Function FileId(0) 40..51 43..46
|
||||||
|
|
||||||
|
FileId(0) 73..76
|
||||||
|
FileId(0) 195..198
|
||||||
|
FileId(1) 40..43
|
||||||
|
FileId(1) 99..102
|
||||||
|
"#]],
|
||||||
|
);
|
||||||
|
}
|
||||||
|
|
||||||
|
#[test]
|
||||||
|
fn goto_ref_on_short_associated_function_overlapping_self_ranges() {
|
||||||
|
check(
|
||||||
|
r#"
|
||||||
|
struct Foo;
|
||||||
|
impl Foo {
|
||||||
|
fn new$0() {}
|
||||||
|
fn bar() {
|
||||||
|
Self::new();
|
||||||
|
impl Foo {
|
||||||
|
fn baz() { Self::new(); }
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
|
"#,
|
||||||
|
expect![[r#"
|
||||||
|
new Function FileId(0) 27..38 30..33
|
||||||
|
|
||||||
|
FileId(0) 68..71
|
||||||
|
FileId(0) 123..126
|
||||||
|
"#]],
|
||||||
|
);
|
||||||
|
}
|
||||||
|
|
||||||
|
#[test]
|
||||||
|
fn goto_ref_on_short_associated_function_no_direct_self_but_path_contains_self() {
|
||||||
|
check(
|
||||||
|
r#"
|
||||||
|
struct Foo;
|
||||||
|
impl Foo {
|
||||||
|
fn new$0() {}
|
||||||
|
}
|
||||||
|
trait Trait {
|
||||||
|
type Assoc;
|
||||||
|
}
|
||||||
|
impl<A, B> Trait for (A, B) {
|
||||||
|
type Assoc = B;
|
||||||
|
}
|
||||||
|
impl Foo {
|
||||||
|
fn bar() {
|
||||||
|
<((), Foo) as Trait>::Assoc::new();
|
||||||
|
<((), Self) as Trait>::Assoc::new();
|
||||||
|
}
|
||||||
|
}
|
||||||
|
"#,
|
||||||
|
expect![[r#"
|
||||||
|
new Function FileId(0) 27..38 30..33
|
||||||
|
|
||||||
|
FileId(0) 188..191
|
||||||
|
FileId(0) 233..236
|
||||||
|
"#]],
|
||||||
|
);
|
||||||
|
}
|
||||||
|
|
||||||
|
// Checks that we can circumvent our fast path logic using complicated type level functions.
|
||||||
|
// This mainly exists as a documentation. I don't believe it is fixable.
|
||||||
|
// Usages search is not 100% accurate anyway; we miss macros.
|
||||||
|
#[test]
|
||||||
|
fn goto_ref_on_short_associated_function_complicated_type_magic_can_confuse_our_logic() {
|
||||||
|
check(
|
||||||
|
r#"
|
||||||
|
struct Foo;
|
||||||
|
impl Foo {
|
||||||
|
fn new$0() {}
|
||||||
|
}
|
||||||
|
|
||||||
|
struct ChoiceA;
|
||||||
|
struct ChoiceB;
|
||||||
|
trait Choice {
|
||||||
|
type Choose<A, B>;
|
||||||
|
}
|
||||||
|
impl Choice for ChoiceA {
|
||||||
|
type Choose<A, B> = A;
|
||||||
|
}
|
||||||
|
impl Choice for ChoiceB {
|
||||||
|
type Choose<A, B> = B;
|
||||||
|
}
|
||||||
|
type Choose<A, C> = <C as Choice>::Choose<A, Foo>;
|
||||||
|
|
||||||
|
fn bar() {
|
||||||
|
Choose::<(), ChoiceB>::new();
|
||||||
|
}
|
||||||
|
"#,
|
||||||
|
expect![[r#"
|
||||||
|
new Function FileId(0) 27..38 30..33
|
||||||
|
|
||||||
|
(no references)
|
||||||
|
"#]],
|
||||||
|
);
|
||||||
|
}
|
||||||
|
|
||||||
|
#[test]
|
||||||
|
fn goto_ref_on_short_associated_function_same_path_mention_alias_and_self() {
|
||||||
|
check(
|
||||||
|
r#"
|
||||||
|
struct Foo;
|
||||||
|
impl Foo {
|
||||||
|
fn new$0() {}
|
||||||
|
}
|
||||||
|
|
||||||
|
type IgnoreFirst<A, B> = B;
|
||||||
|
|
||||||
|
impl Foo {
|
||||||
|
fn bar() {
|
||||||
|
<IgnoreFirst<Foo, Self>>::new();
|
||||||
|
}
|
||||||
|
}
|
||||||
|
"#,
|
||||||
|
expect![[r#"
|
||||||
|
new Function FileId(0) 27..38 30..33
|
||||||
|
|
||||||
|
FileId(0) 131..134
|
||||||
|
"#]],
|
||||||
|
);
|
||||||
|
}
|
||||||
}
|
}
|
||||||
|
|
Loading…
Reference in a new issue