internal: Rewrite ImportMap::search_dependencies

This commit is contained in:
Lukas Wirth 2024-01-04 17:19:02 +01:00
parent 9279c6566b
commit 9b3052104c
2 changed files with 254 additions and 275 deletions

View file

@ -3,13 +3,13 @@
use std::{fmt, hash::BuildHasherDefault};
use base_db::CrateId;
use fst::{self, raw::IndexedValue, Streamer};
use fst::{self, raw::IndexedValue, Automaton, Streamer};
use hir_expand::name::Name;
use indexmap::IndexMap;
use itertools::Itertools;
use rustc_hash::{FxHashSet, FxHasher};
use smallvec::SmallVec;
use stdx::format_to;
use stdx::{format_to, TupleExt};
use triomphe::Arc;
use crate::{
@ -20,12 +20,10 @@ use crate::{
AssocItemId, ModuleDefId, ModuleId, TraitId,
};
type FxIndexMap<K, V> = IndexMap<K, V, BuildHasherDefault<FxHasher>>;
/// Item import details stored in the `ImportMap`.
#[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord)]
pub struct ImportInfo {
/// A name that can be used to import the item, relative to the crate's root.
/// A name that can be used to import the item, relative to the container.
pub name: Name,
/// The module containing this item.
pub container: ModuleId,
@ -35,22 +33,22 @@ pub struct ImportInfo {
pub is_unstable: bool,
}
type ImportMapIndex = FxIndexMap<ItemInNs, (SmallVec<[ImportInfo; 1]>, IsTraitAssocItem)>;
/// A map from publicly exported items to its name.
///
/// Reexports of items are taken into account, ie. if something is exported under multiple
/// names, the one with the shortest import path will be used.
/// Reexports of items are taken into account.
#[derive(Default)]
pub struct ImportMap {
map: ImportMapIndex,
/// List of keys stored in `map`, sorted lexicographically by their `ModPath`. Indexed by the
/// values returned by running `fst`.
/// Maps from `ItemInNs` to information of imports that bring the item into scope.
item_to_info_map: ImportMapIndex,
/// List of keys stored in [`Self::item_to_info_map`], sorted lexicographically by their
/// [`Name`]. Indexed by the values returned by running `fst`.
///
/// Since a name can refer to multiple items due to namespacing, we store all items with the
/// same name right after each other. This allows us to find all items after the FST gives us
/// the index of the first one.
importables: Vec<ItemInNs>,
/// Since a name can refer to multiple items due to namespacing and import aliases, we store all
/// items with the same name right after each other. This allows us to find all items after the
/// fst gives us the index of the first one.
///
/// The [`u32`] is the index into the smallvec in the value of [`Self::item_to_info_map`].
importables: Vec<(ItemInNs, u32)>,
fst: fst::Map<Vec<u8>>,
}
@ -60,10 +58,13 @@ enum IsTraitAssocItem {
No,
}
type FxIndexMap<K, V> = IndexMap<K, V, BuildHasherDefault<FxHasher>>;
type ImportMapIndex = FxIndexMap<ItemInNs, (SmallVec<[ImportInfo; 1]>, IsTraitAssocItem)>;
impl ImportMap {
pub fn dump(&self, db: &dyn DefDatabase) -> String {
let mut out = String::new();
for (k, v) in self.map.iter() {
for (k, v) in self.item_to_info_map.iter() {
format_to!(out, "{:?} ({:?}) -> ", k, v.1);
for v in &v.0 {
format_to!(out, "{}:{:?}, ", v.name.display(db.upcast()), v.container);
@ -76,20 +77,18 @@ impl ImportMap {
pub(crate) fn import_map_query(db: &dyn DefDatabase, krate: CrateId) -> Arc<Self> {
let _p = profile::span("import_map_query");
let map = collect_import_map(db, krate);
let map = Self::collect_import_map(db, krate);
let mut importables: Vec<_> = map
.iter()
// We've only collected items, whose name cannot be tuple field.
.flat_map(|(&item, (info, is_assoc))| {
info.iter().map(move |info| {
(item, *is_assoc, info.name.as_str().unwrap().to_ascii_lowercase())
// We've only collected items, whose name cannot be tuple field so unwrapping is fine.
.flat_map(|(&item, (info, _))| {
info.iter().enumerate().map(move |(idx, info)| {
(item, info.name.as_str().unwrap().to_ascii_lowercase(), idx as u32)
})
})
.collect();
importables.sort_by(|(_, l_is_assoc, lhs_name), (_, r_is_assoc, rhs_name)| {
lhs_name.cmp(rhs_name).then_with(|| l_is_assoc.cmp(r_is_assoc))
});
importables.sort_by(|(_, lhs_name, _), (_, rhs_name, _)| lhs_name.cmp(rhs_name));
importables.dedup();
// Build the FST, taking care not to insert duplicate values.
@ -97,160 +96,159 @@ impl ImportMap {
let iter = importables
.iter()
.enumerate()
.dedup_by(|(_, (_, _, lhs)), (_, (_, _, rhs))| lhs == rhs);
for (start_idx, (_, _, name)) in iter {
.dedup_by(|(_, (_, lhs, _)), (_, (_, rhs, _))| lhs == rhs);
for (start_idx, (_, name, _)) in iter {
let _ = builder.insert(name, start_idx as u64);
}
Arc::new(ImportMap {
map,
item_to_info_map: map,
fst: builder.into_map(),
importables: importables.into_iter().map(|(item, _, _)| item).collect(),
importables: importables.into_iter().map(|(item, _, idx)| (item, idx)).collect(),
})
}
pub fn import_info_for(&self, item: ItemInNs) -> Option<&[ImportInfo]> {
self.map.get(&item).map(|(info, _)| &**info)
self.item_to_info_map.get(&item).map(|(info, _)| &**info)
}
}
fn collect_import_map(db: &dyn DefDatabase, krate: CrateId) -> ImportMapIndex {
let _p = profile::span("collect_import_map");
fn collect_import_map(db: &dyn DefDatabase, krate: CrateId) -> ImportMapIndex {
let _p = profile::span("collect_import_map");
let def_map = db.crate_def_map(krate);
let mut map = FxIndexMap::default();
let def_map = db.crate_def_map(krate);
let mut map = FxIndexMap::default();
// We look only into modules that are public(ly reexported), starting with the crate root.
let root = def_map.module_id(DefMap::ROOT);
let mut worklist = vec![root];
let mut visited = FxHashSet::default();
// We look only into modules that are public(ly reexported), starting with the crate root.
let root = def_map.module_id(DefMap::ROOT);
let mut worklist = vec![root];
let mut visited = FxHashSet::default();
while let Some(module) = worklist.pop() {
if !visited.insert(module) {
continue;
}
let ext_def_map;
let mod_data = if module.krate == krate {
&def_map[module.local_id]
} else {
// The crate might reexport a module defined in another crate.
ext_def_map = module.def_map(db);
&ext_def_map[module.local_id]
};
let visible_items = mod_data.scope.entries().filter_map(|(name, per_ns)| {
let per_ns = per_ns.filter_visibility(|vis| vis == Visibility::Public);
if per_ns.is_none() {
None
} else {
Some((name, per_ns))
}
});
for (name, per_ns) in visible_items {
for (item, import) in per_ns.iter_items() {
let attr_id = if let Some(import) = import {
match import {
ImportOrExternCrate::ExternCrate(id) => Some(id.into()),
ImportOrExternCrate::Import(id) => Some(id.import.into()),
}
} else {
match item {
ItemInNs::Types(id) | ItemInNs::Values(id) => id.try_into().ok(),
ItemInNs::Macros(id) => Some(id.into()),
}
};
let (is_doc_hidden, is_unstable) = attr_id.map_or((false, false), |attr_id| {
let attrs = db.attrs(attr_id);
(attrs.has_doc_hidden(), attrs.is_unstable())
});
let import_info = ImportInfo {
name: name.clone(),
container: module,
is_doc_hidden,
is_unstable,
};
if let Some(ModuleDefId::TraitId(tr)) = item.as_module_def_id() {
collect_trait_assoc_items(
db,
&mut map,
tr,
matches!(item, ItemInNs::Types(_)),
&import_info,
);
}
let (infos, _) =
map.entry(item).or_insert_with(|| (SmallVec::new(), IsTraitAssocItem::No));
infos.reserve_exact(1);
infos.push(import_info);
// If we've just added a module, descend into it.
if let Some(ModuleDefId::ModuleId(mod_id)) = item.as_module_def_id() {
worklist.push(mod_id);
}
}
}
}
map.shrink_to_fit();
map
}
fn collect_trait_assoc_items(
db: &dyn DefDatabase,
map: &mut ImportMapIndex,
tr: TraitId,
is_type_in_ns: bool,
trait_import_info: &ImportInfo,
) {
let _p = profile::span("collect_trait_assoc_items");
for &(ref assoc_item_name, item) in &db.trait_data(tr).items {
let module_def_id = match item {
AssocItemId::FunctionId(f) => ModuleDefId::from(f),
AssocItemId::ConstId(c) => ModuleDefId::from(c),
// cannot use associated type aliases directly: need a `<Struct as Trait>::TypeAlias`
// qualifier, ergo no need to store it for imports in import_map
AssocItemId::TypeAliasId(_) => {
cov_mark::hit!(type_aliases_ignored);
while let Some(module) = worklist.pop() {
if !visited.insert(module) {
continue;
}
};
let assoc_item = if is_type_in_ns {
ItemInNs::Types(module_def_id)
} else {
ItemInNs::Values(module_def_id)
};
let ext_def_map;
let mod_data = if module.krate == krate {
&def_map[module.local_id]
} else {
// The crate might reexport a module defined in another crate.
ext_def_map = module.def_map(db);
&ext_def_map[module.local_id]
};
let attrs = &db.attrs(item.into());
let assoc_item_info = ImportInfo {
container: trait_import_info.container,
name: assoc_item_name.clone(),
is_doc_hidden: attrs.has_doc_hidden(),
is_unstable: attrs.is_unstable(),
};
let visible_items = mod_data.scope.entries().filter_map(|(name, per_ns)| {
let per_ns = per_ns.filter_visibility(|vis| vis == Visibility::Public);
if per_ns.is_none() {
None
} else {
Some((name, per_ns))
}
});
let (infos, _) =
map.entry(assoc_item).or_insert_with(|| (SmallVec::new(), IsTraitAssocItem::Yes));
infos.reserve_exact(1);
infos.push(assoc_item_info);
for (name, per_ns) in visible_items {
for (item, import) in per_ns.iter_items() {
let attr_id = if let Some(import) = import {
match import {
ImportOrExternCrate::ExternCrate(id) => Some(id.into()),
ImportOrExternCrate::Import(id) => Some(id.import.into()),
}
} else {
match item {
ItemInNs::Types(id) | ItemInNs::Values(id) => id.try_into().ok(),
ItemInNs::Macros(id) => Some(id.into()),
}
};
let (is_doc_hidden, is_unstable) = attr_id.map_or((false, false), |attr_id| {
let attrs = db.attrs(attr_id);
(attrs.has_doc_hidden(), attrs.is_unstable())
});
let import_info = ImportInfo {
name: name.clone(),
container: module,
is_doc_hidden,
is_unstable,
};
if let Some(ModuleDefId::TraitId(tr)) = item.as_module_def_id() {
Self::collect_trait_assoc_items(
db,
&mut map,
tr,
matches!(item, ItemInNs::Types(_)),
&import_info,
);
}
let (infos, _) =
map.entry(item).or_insert_with(|| (SmallVec::new(), IsTraitAssocItem::No));
infos.reserve_exact(1);
infos.push(import_info);
// If we've just added a module, descend into it.
if let Some(ModuleDefId::ModuleId(mod_id)) = item.as_module_def_id() {
worklist.push(mod_id);
}
}
}
}
map.shrink_to_fit();
map
}
}
impl PartialEq for ImportMap {
fn eq(&self, other: &Self) -> bool {
// `fst` and `importables` are built from `map`, so we don't need to compare them.
self.map == other.map
fn collect_trait_assoc_items(
db: &dyn DefDatabase,
map: &mut ImportMapIndex,
tr: TraitId,
is_type_in_ns: bool,
trait_import_info: &ImportInfo,
) {
let _p = profile::span("collect_trait_assoc_items");
for &(ref assoc_item_name, item) in &db.trait_data(tr).items {
let module_def_id = match item {
AssocItemId::FunctionId(f) => ModuleDefId::from(f),
AssocItemId::ConstId(c) => ModuleDefId::from(c),
// cannot use associated type aliases directly: need a `<Struct as Trait>::TypeAlias`
// qualifier, ergo no need to store it for imports in import_map
AssocItemId::TypeAliasId(_) => {
cov_mark::hit!(type_aliases_ignored);
continue;
}
};
let assoc_item = if is_type_in_ns {
ItemInNs::Types(module_def_id)
} else {
ItemInNs::Values(module_def_id)
};
let attrs = &db.attrs(item.into());
let assoc_item_info = ImportInfo {
container: trait_import_info.container,
name: assoc_item_name.clone(),
is_doc_hidden: attrs.has_doc_hidden(),
is_unstable: attrs.is_unstable(),
};
let (infos, _) =
map.entry(assoc_item).or_insert_with(|| (SmallVec::new(), IsTraitAssocItem::Yes));
infos.reserve_exact(1);
infos.push(assoc_item_info);
}
}
}
impl Eq for ImportMap {}
impl PartialEq for ImportMap {
fn eq(&self, other: &Self) -> bool {
// `fst` and `importables` are built from `map`, so we don't need to compare them.
self.item_to_info_map == other.item_to_info_map
}
}
impl fmt::Debug for ImportMap {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
let mut importable_names: Vec<_> = self
.map
.item_to_info_map
.iter()
.map(|(item, (infos, _))| {
let l = infos.len();
@ -348,39 +346,6 @@ impl Query {
_ => true,
}
}
/// Checks whether the import map entry matches the query.
fn import_matches(&self, import: &ImportInfo, enforce_lowercase: bool) -> bool {
let _p = profile::span("import_map::Query::import_matches");
// FIXME: Can we get rid of the alloc here?
let input = import.name.to_smol_str();
let mut _s_slot;
let case_insensitive = enforce_lowercase || !self.case_sensitive;
let input = if case_insensitive {
_s_slot = String::from(input);
_s_slot.make_ascii_lowercase();
&*_s_slot
} else {
&*input
};
let query_string = if case_insensitive { &self.lowercased } else { &self.query };
match self.search_mode {
SearchMode::Exact => input == *query_string,
SearchMode::Prefix => input.starts_with(query_string),
SearchMode::Fuzzy => {
let mut input_chars = input.chars();
for query_char in query_string.chars() {
if !input_chars.any(|it| it == query_char) {
return false;
}
}
true
}
}
}
}
/// Searches dependencies of `krate` for an importable name matching `query`.
@ -398,69 +363,85 @@ pub fn search_dependencies(
let import_maps: Vec<_> =
graph[krate].dependencies.iter().map(|dep| db.import_map(dep.crate_id)).collect();
let automaton = fst::automaton::Subsequence::new(&query.lowercased);
let mut op = fst::map::OpBuilder::new();
for map in &import_maps {
op = op.add(map.fst.search(&automaton));
match query.search_mode {
SearchMode::Exact => {
let automaton = fst::automaton::Str::new(&query.lowercased);
for map in &import_maps {
op = op.add(map.fst.search(&automaton));
}
search_maps(&import_maps, op.union(), query)
}
SearchMode::Fuzzy => {
let automaton = fst::automaton::Subsequence::new(&query.lowercased);
for map in &import_maps {
op = op.add(map.fst.search(&automaton));
}
search_maps(&import_maps, op.union(), query)
}
SearchMode::Prefix => {
let automaton = fst::automaton::Str::new(&query.lowercased).starts_with();
for map in &import_maps {
op = op.add(map.fst.search(&automaton));
}
search_maps(&import_maps, op.union(), query)
}
}
}
let mut stream = op.union();
fn search_maps(
import_maps: &[Arc<ImportMap>],
mut stream: fst::map::Union<'_>,
query: &Query,
) -> FxHashSet<ItemInNs> {
let mut res = FxHashSet::default();
let mut common_importable_data_scratch = vec![];
// FIXME: Improve this, its rather unreadable and does duplicate amount of work
while let Some((_, indexed_values)) = stream.next() {
for &IndexedValue { index, value } in indexed_values {
let import_map = &import_maps[index];
let importables @ [importable, ..] = &import_map.importables[value as usize..] else {
continue;
};
let &(ref importable_data, is_trait_assoc_item) = &import_map.map[importable];
if !query.matches_assoc_mode(is_trait_assoc_item) {
continue;
}
while let Some((key, indexed_values)) = stream.next() {
for &IndexedValue { index: import_map_idx, value } in indexed_values {
let import_map = &import_maps[import_map_idx];
let importables = &import_map.importables[value as usize..];
// Fetch all the known names of this importable item (to handle import aliases/renames)
common_importable_data_scratch.extend(
importable_data
.iter()
.filter(|&info| query.import_matches(info, true))
// Name shared by the importable items in this group.
.map(|info| info.name.to_smol_str()),
);
if common_importable_data_scratch.is_empty() {
continue;
}
common_importable_data_scratch.sort();
common_importable_data_scratch.dedup();
let iter =
common_importable_data_scratch.drain(..).flat_map(|common_importable_name| {
// Add the items from this name group. Those are all subsequent items in
// `importables` whose name match `common_importable_name`.
importables
.iter()
.copied()
.take_while(move |item| {
let &(ref import_infos, assoc_mode) = &import_map.map[item];
query.matches_assoc_mode(assoc_mode)
&& import_infos.iter().any(|info| {
info.name
.to_smol_str()
.eq_ignore_ascii_case(&common_importable_name)
})
})
.filter(move |item| {
!query.case_sensitive || {
// we've already checked the common importables name case-insensitively
let &(ref import_infos, _) = &import_map.map[item];
import_infos.iter().any(|info| query.import_matches(info, false))
}
})
let iter = importables
.iter()
.copied()
.map(|(item, info_idx)| {
let (import_infos, assoc_mode) = &import_map.item_to_info_map[&item];
(item, &import_infos[info_idx as usize], *assoc_mode)
})
// we put all entries with the same lowercased name in a row, so stop once we find a
// different name in the importables
.take_while(|&(_, info, _)| {
info.name.to_smol_str().as_bytes().eq_ignore_ascii_case(&key)
})
.filter(|&(_, info, assoc_mode)| {
if !query.matches_assoc_mode(assoc_mode) {
return false;
}
if !query.case_sensitive {
return true;
}
let name = info.name.to_smol_str();
match query.search_mode {
SearchMode::Exact => name == query.query,
SearchMode::Prefix => name.starts_with(&query.query),
SearchMode::Fuzzy => {
let mut name = &*name;
query.query.chars().all(|query_char| {
match name.match_indices(query_char).next() {
Some((index, _)) => {
name = &name[index + 1..];
true
}
None => false,
}
})
}
}
});
res.extend(iter);
res.extend(iter.map(TupleExt::head));
if res.len() >= query.limit {
return res;
@ -484,7 +465,7 @@ mod tests {
impl ImportMap {
fn fmt_for_test(&self, db: &dyn DefDatabase) -> String {
let mut importable_paths: Vec<_> = self
.map
.item_to_info_map
.iter()
.flat_map(|(item, (info, _))| info.iter().map(move |info| (item, info)))
.map(|(item, info)| {
@ -911,28 +892,28 @@ mod tests {
#[test]
fn search_mode() {
let ra_fixture = r#"
//- /main.rs crate:main deps:dep
//- /dep.rs crate:dep deps:tdep
use tdep::fmt as fmt_dep;
pub mod fmt {
pub trait Display {
fn fmt();
}
}
#[macro_export]
macro_rules! Fmt {
() => {};
}
pub struct Fmt;
//- /main.rs crate:main deps:dep
//- /dep.rs crate:dep deps:tdep
use tdep::fmt as fmt_dep;
pub mod fmt {
pub trait Display {
fn fmt();
}
}
#[macro_export]
macro_rules! Fmt {
() => {};
}
pub struct Fmt;
pub fn format() {}
pub fn no() {}
pub fn format() {}
pub fn no() {}
//- /tdep.rs crate:tdep
pub mod fmt {
pub struct NotImportableFromMain;
}
"#;
//- /tdep.rs crate:tdep
pub mod fmt {
pub struct NotImportableFromMain;
}
"#;
check_search(
ra_fixture,
@ -1000,19 +981,6 @@ mod tests {
dep::fmt::Display::fmt (a)
"#]],
);
check_search(
ra_fixture,
"main",
Query::new("fmt".to_string()),
expect![[r#"
dep::Fmt (m)
dep::Fmt (t)
dep::Fmt (v)
dep::fmt (t)
dep::fmt::Display::fmt (a)
"#]],
);
}
#[test]

View file

@ -59,6 +59,17 @@ impl<T, U> TupleExt for (T, U) {
}
}
impl<T, U, V> TupleExt for (T, U, V) {
type Head = T;
type Tail = V;
fn head(self) -> Self::Head {
self.0
}
fn tail(self) -> Self::Tail {
self.2
}
}
pub fn to_lower_snake_case(s: &str) -> String {
to_snake_case(s, char::to_lowercase)
}