aboutsummaryrefslogtreecommitdiff
path: root/crates/ra_ide_api/src/symbol_index.rs
blob: 94fe1d6d7d168712e9bc097a83111f1bd4c32b7d (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
//! This module handles fuzzy-searching of functions, structs and other symbols
//! by name across the whole workspace and dependencies.
//!
//! It works by building an incrementally-updated text-search index of all
//! symbols. The backbone of the index is the **awesome** `fst` crate by
//! @BurntSushi.
//!
//! In a nutshell, you give a set of strings to `fst`, and it builds a
//! finite state machine describing this set of strings. The strings which
//! could fuzzy-match a pattern can also be described by a finite state machine.
//! What is freaking cool is that you can now traverse both state machines in
//! lock-step to enumerate the strings which are both in the input set and
//! fuzz-match the query. Or, more formally, given two languages described by
//! FSTs, one can build a product FST which describes the intersection of the
//! languages.
//!
//! `fst` does not support cheap updating of the index, but it supports unioning
//! of state machines. So, to account for changing source code, we build an FST
//! for each library (which is assumed to never change) and an FST for each Rust
//! file in the current workspace, and run a query against the union of all
//! those FSTs.
use std::{
    cmp::Ordering,
    hash::{Hash, Hasher},
    sync::Arc,
    mem,
    fmt,
};

use fst::{self, Streamer};
use ra_syntax::{
    SyntaxNode, SyntaxNodePtr, SourceFile, SmolStr, TreeArc, AstNode,
    algo::{visit::{visitor, Visitor}, find_covering_node},
    SyntaxKind::{self, *},
    ast::{self, NameOwner},
    WalkEvent,
    TextRange,
};
use ra_db::{
    SourceRootId, SourceDatabase,
    salsa::{self, ParallelDatabase},
};
use rayon::prelude::*;

use crate::{
    FileId, Query,
    db::RootDatabase,
};

#[salsa::query_group(SymbolsDatabaseStorage)]
pub(crate) trait SymbolsDatabase: hir::db::HirDatabase {
    fn file_symbols(&self, file_id: FileId) -> Arc<SymbolIndex>;
    #[salsa::input]
    fn library_symbols(&self, id: SourceRootId) -> Arc<SymbolIndex>;
    /// The set of "local" (that is, from the current workspace) roots.
    /// Files in local roots are assumed to change frequently.
    #[salsa::input]
    fn local_roots(&self) -> Arc<Vec<SourceRootId>>;
    /// The set of roots for crates.io libraries.
    /// Files in libraries are assumed to never change.
    #[salsa::input]
    fn library_roots(&self) -> Arc<Vec<SourceRootId>>;
}

fn file_symbols(db: &impl SymbolsDatabase, file_id: FileId) -> Arc<SymbolIndex> {
    db.check_canceled();
    let source_file = db.parse(file_id);

    let mut symbols = source_file_to_file_symbols(&source_file, file_id);

    for (name, text_range) in hir::source_binder::macro_symbols(db, file_id) {
        let node = find_covering_node(source_file.syntax(), text_range);
        let ptr = SyntaxNodePtr::new(node);
        // TODO: Should we get container name for macro symbols?
        symbols.push(FileSymbol { file_id, name, ptr, name_range: None, container_name: None })
    }

    Arc::new(SymbolIndex::new(symbols))
}

pub(crate) fn world_symbols(db: &RootDatabase, query: Query) -> Vec<FileSymbol> {
    /// Need to wrap Snapshot to provide `Clone` impl for `map_with`
    struct Snap(salsa::Snapshot<RootDatabase>);
    impl Clone for Snap {
        fn clone(&self) -> Snap {
            Snap(self.0.snapshot())
        }
    }

    let buf: Vec<Arc<SymbolIndex>> = if query.libs {
        let snap = Snap(db.snapshot());
        db.library_roots()
            .par_iter()
            .map_with(snap, |db, &lib_id| db.0.library_symbols(lib_id))
            .collect()
    } else {
        let mut files = Vec::new();
        for &root in db.local_roots().iter() {
            let sr = db.source_root(root);
            files.extend(sr.files.values().map(|&it| it))
        }

        let snap = Snap(db.snapshot());
        files.par_iter().map_with(snap, |db, &file_id| db.0.file_symbols(file_id)).collect()
    };
    query.search(&buf)
}

pub(crate) fn index_resolve(db: &RootDatabase, name_ref: &ast::NameRef) -> Vec<FileSymbol> {
    let name = name_ref.text();
    let mut query = Query::new(name.to_string());
    query.exact();
    query.limit(4);
    crate::symbol_index::world_symbols(db, query)
}

#[derive(Default)]
pub(crate) struct SymbolIndex {
    symbols: Vec<FileSymbol>,
    map: fst::Map,
}

impl fmt::Debug for SymbolIndex {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        f.debug_struct("SymbolIndex").field("n_symbols", &self.symbols.len()).finish()
    }
}

impl PartialEq for SymbolIndex {
    fn eq(&self, other: &SymbolIndex) -> bool {
        self.symbols == other.symbols
    }
}

impl Eq for SymbolIndex {}

impl Hash for SymbolIndex {
    fn hash<H: Hasher>(&self, hasher: &mut H) {
        self.symbols.hash(hasher)
    }
}

impl SymbolIndex {
    fn new(mut symbols: Vec<FileSymbol>) -> SymbolIndex {
        fn cmp(s1: &FileSymbol, s2: &FileSymbol) -> Ordering {
            unicase::Ascii::new(s1.name.as_str()).cmp(&unicase::Ascii::new(s2.name.as_str()))
        }
        symbols.par_sort_by(cmp);
        symbols.dedup_by(|s1, s2| cmp(s1, s2) == Ordering::Equal);
        let names = symbols.iter().map(|it| it.name.as_str().to_lowercase());
        let map = fst::Map::from_iter(names.zip(0u64..)).unwrap();
        SymbolIndex { symbols, map }
    }

    pub(crate) fn len(&self) -> usize {
        self.symbols.len()
    }

    pub(crate) fn memory_size(&self) -> usize {
        self.map.as_fst().size() + self.symbols.len() * mem::size_of::<FileSymbol>()
    }

    pub(crate) fn for_files(
        files: impl ParallelIterator<Item = (FileId, TreeArc<SourceFile>)>,
    ) -> SymbolIndex {
        let symbols = files
            .flat_map(|(file_id, file)| source_file_to_file_symbols(&file, file_id))
            .collect::<Vec<_>>();
        SymbolIndex::new(symbols)
    }
}

impl Query {
    pub(crate) fn search(self, indices: &[Arc<SymbolIndex>]) -> Vec<FileSymbol> {
        let mut op = fst::map::OpBuilder::new();
        for file_symbols in indices.iter() {
            let automaton = fst::automaton::Subsequence::new(&self.lowercased);
            op = op.add(file_symbols.map.search(automaton))
        }
        let mut stream = op.union();
        let mut res = Vec::new();
        while let Some((_, indexed_values)) = stream.next() {
            if res.len() >= self.limit {
                break;
            }
            for indexed_value in indexed_values {
                let file_symbols = &indices[indexed_value.index];
                let idx = indexed_value.value as usize;

                let symbol = &file_symbols.symbols[idx];
                if self.only_types && !is_type(symbol.ptr.kind()) {
                    continue;
                }
                if self.exact && symbol.name != self.query {
                    continue;
                }
                res.push(symbol.clone());
            }
        }
        res
    }
}

fn is_type(kind: SyntaxKind) -> bool {
    match kind {
        STRUCT_DEF | ENUM_DEF | TRAIT_DEF | TYPE_ALIAS_DEF => true,
        _ => false,
    }
}

/// The actual data that is stored in the index. It should be as compact as
/// possible.
#[derive(Debug, Clone, PartialEq, Eq, Hash)]
pub(crate) struct FileSymbol {
    pub(crate) file_id: FileId,
    pub(crate) name: SmolStr,
    pub(crate) ptr: SyntaxNodePtr,
    pub(crate) name_range: Option<TextRange>,
    pub(crate) container_name: Option<SmolStr>,
}

fn source_file_to_file_symbols(source_file: &SourceFile, file_id: FileId) -> Vec<FileSymbol> {
    let mut symbols = Vec::new();
    let mut stack = Vec::new();

    for event in source_file.syntax().preorder() {
        match event {
            WalkEvent::Enter(node) => {
                if let Some(mut symbol) = to_file_symbol(node, file_id) {
                    symbol.container_name = stack.last().cloned();

                    stack.push(symbol.name.clone());
                    symbols.push(symbol);
                }
            }

            WalkEvent::Leave(node) => {
                if to_symbol(node).is_some() {
                    stack.pop();
                }
            }
        }
    }

    symbols
}

fn to_symbol(node: &SyntaxNode) -> Option<(SmolStr, SyntaxNodePtr, TextRange)> {
    fn decl<N: NameOwner>(node: &N) -> Option<(SmolStr, SyntaxNodePtr, TextRange)> {
        let name = node.name()?;
        let name_range = name.syntax().range();
        let name = name.text().clone();
        let ptr = SyntaxNodePtr::new(node.syntax());

        Some((name, ptr, name_range))
    }
    visitor()
        .visit(decl::<ast::FnDef>)
        .visit(decl::<ast::StructDef>)
        .visit(decl::<ast::EnumDef>)
        .visit(decl::<ast::TraitDef>)
        .visit(decl::<ast::Module>)
        .visit(decl::<ast::TypeAliasDef>)
        .visit(decl::<ast::ConstDef>)
        .visit(decl::<ast::StaticDef>)
        .accept(node)?
}

fn to_file_symbol(node: &SyntaxNode, file_id: FileId) -> Option<FileSymbol> {
    to_symbol(node).map(move |(name, ptr, name_range)| FileSymbol {
        name,
        ptr,
        file_id,
        name_range: Some(name_range),
        container_name: None,
    })
}