1 //===--- Trigram.h - Trigram generation for Fuzzy Matching ------*- C++ -*-===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 ///
9 /// \file
10 /// Trigrams are attributes of the symbol unqualified name used to effectively
11 /// extract symbols which can be fuzzy-matched given user query from the
12 /// inverted index. To match query with the extracted set of trigrams Q, the set
13 /// of generated trigrams T for identifier (unqualified symbol name) should
14 /// contain all items of Q, i.e. Q ⊆ T.
15 ///
16 /// Trigram sets extracted from unqualified name and from query are different:
17 /// the set of query trigrams only contains consecutive sequences of three
18 /// characters (which is only a subset of all trigrams generated for an
19 /// identifier).
20 ///
21 //===----------------------------------------------------------------------===//
22 
23 #ifndef LLVM_CLANG_TOOLS_EXTRA_CLANGD_DEX_TRIGRAM_H
24 #define LLVM_CLANG_TOOLS_EXTRA_CLANGD_DEX_TRIGRAM_H
25 
26 #include "Token.h"
27 #include "llvm/ADT/bit.h"
28 
29 #include <string>
30 
31 namespace clang {
32 namespace clangd {
33 namespace dex {
34 
35 // Compact representation of a trigram (string with up to 3 characters).
36 // Trigram generation is the hot path of indexing, so Token is too wasteful.
37 class Trigram {
38   std::array<char, 4> Data; // Last element is length.
39   // Steal some invalid bit patterns for DenseMap sentinels.
40   enum class Sentinel { Tombstone = 4, Empty = 5 };
Trigram(Sentinel S)41   Trigram(Sentinel S) : Data{0, 0, 0, static_cast<char>(S)} {}
id()42   uint32_t id() const { return llvm::bit_cast<uint32_t>(Data); }
43 
44 public:
Trigram()45   Trigram() : Data{0, 0, 0, 0} {}
Trigram(char A)46   Trigram(char A) : Data{A, 0, 0, 1} {}
Trigram(char A,char B)47   Trigram(char A, char B) : Data{A, B, 0, 2} {}
Trigram(char A,char B,char C)48   Trigram(char A, char B, char C) : Data{A, B, C, 3} {}
str()49   std::string str() const { return std::string(Data.data(), Data[3]); }
50   friend struct ::llvm::DenseMapInfo<Trigram>;
51   friend bool operator==(Trigram L, Trigram R) { return L.id() == R.id(); }
52   friend bool operator<(Trigram L, Trigram R) { return L.id() < R.id(); }
53 };
54 
55 /// Produces list of unique fuzzy-search trigrams from unqualified symbol.
56 /// The trigrams give the 3-character query substrings this symbol can match.
57 ///
58 /// The symbol's name is broken into segments, e.g. "FooBar" has two segments.
59 /// Trigrams can start at any character in the input. Then we can choose to move
60 /// to the next character or to the start of the next segment.
61 ///
62 /// Short trigrams (length 1-2) are used for short queries. These are:
63 ///  - prefixes of the identifier, of length 1 and 2
64 ///  - the first character + next head character
65 ///
66 /// For "FooBar" we get the following trigrams:
67 ///  {f, fo, fb, foo, fob, fba, oob, oba, bar}.
68 ///
69 /// Trigrams are lowercase, as trigram matching is case-insensitive.
70 /// Trigrams in the list are deduplicated.
71 void generateIdentifierTrigrams(llvm::StringRef Identifier,
72                                 std::vector<Trigram> &Out);
73 
74 /// Returns list of unique fuzzy-search trigrams given a query.
75 ///
76 /// Query is segmented using FuzzyMatch API and downcasted to lowercase. Then,
77 /// the simplest trigrams - sequences of three consecutive letters and digits
78 /// are extracted and returned after deduplication.
79 ///
80 /// For short queries (less than 3 characters with Head or Tail roles in Fuzzy
81 /// Matching segmentation) this returns a single trigram with the first
82 /// characters (up to 3) to perform prefix match.
83 std::vector<Token> generateQueryTrigrams(llvm::StringRef Query);
84 
85 } // namespace dex
86 } // namespace clangd
87 } // namespace clang
88 
89 namespace llvm {
90 template <> struct DenseMapInfo<clang::clangd::dex::Trigram> {
91   using Trigram = clang::clangd::dex::Trigram;
92   static inline Trigram getEmptyKey() {
93     return Trigram(Trigram::Sentinel::Empty);
94   }
95   static inline Trigram getTombstoneKey() {
96     return Trigram(Trigram::Sentinel::Tombstone);
97   }
98   static unsigned getHashValue(Trigram V) {
99     // Finalize step from MurmurHash3.
100     uint32_t X = V.id();
101     X ^= X >> 16;
102     X *= uint32_t{0x85ebca6b};
103     X ^= X >> 13;
104     X *= uint32_t{0xc2b2ae35};
105     X ^= X >> 16;
106     return X;
107   }
108   static bool isEqual(const Trigram &LHS, const Trigram &RHS) {
109     return LHS == RHS;
110   }
111 };
112 } // namespace llvm
113 
114 #endif // LLVM_CLANG_TOOLS_EXTRA_CLANGD_DEX_TRIGRAM_H
115