| 1 |
//===--- Iterator.h - Query Symbol Retrieval --------------------*- 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 |
/// Symbol index queries consist of specific requirements for the requested |
| 11 |
/// symbol, such as high fuzzy matching score, scope, type etc. The lists of all |
| 12 |
/// symbols matching some criteria (e.g. belonging to "clang::clangd::" scope) |
| 13 |
/// are expressed in a form of Search Tokens which are stored in the inverted |
| 14 |
/// index. Inverted index maps these tokens to the posting lists - sorted (by |
| 15 |
/// symbol quality) sequences of symbol IDs matching the token, e.g. scope token |
| 16 |
/// "clangd::clangd::" is mapped to the list of IDs of all symbols which are |
| 17 |
/// declared in this namespace. Search queries are build from a set of |
| 18 |
/// requirements which can be combined with each other forming the query trees. |
| 19 |
/// The leafs of such trees are posting lists, and the nodes are operations on |
| 20 |
/// these posting lists, e.g. intersection or union. Efficient processing of |
| 21 |
/// these multi-level queries is handled by Iterators. Iterators advance through |
| 22 |
/// all leaf posting lists producing the result of search query, which preserves |
| 23 |
/// the sorted order of IDs. Having the resulting IDs sorted is important, |
| 24 |
/// because it allows receiving a certain number of the most valuable items |
| 25 |
/// (e.g. symbols with highest quality which was the sorting key in the first |
| 26 |
/// place) without processing all items with requested properties (this might |
| 27 |
/// not be computationally effective if search request is not very restrictive). |
| 28 |
// |
| 29 |
//===----------------------------------------------------------------------===// |
| 30 |
|
| 31 |
#ifndef LLVM_CLANG_TOOLS_EXTRA_CLANGD_INDEX_DEX_ITERATOR_H |
| 32 |
#define LLVM_CLANG_TOOLS_EXTRA_CLANGD_INDEX_DEX_ITERATOR_H |
| 33 |
|
| 34 |
#include "llvm/Support/raw_ostream.h" |
| 35 |
#include <algorithm> |
| 36 |
#include <memory> |
| 37 |
#include <vector> |
| 38 |
|
| 39 |
namespace clang { |
| 40 |
namespace clangd { |
| 41 |
namespace dex { |
| 42 |
|
| 43 |
/// Symbol position in the list of all index symbols sorted by a pre-computed |
| 44 |
/// symbol quality. |
| 45 |
using DocID = uint32_t; |
| 46 |
|
| 47 |
/// Iterator is the interface for Query Tree node. The simplest type of Iterator |
| 48 |
/// is DocumentIterator which is simply a wrapper around PostingList iterator |
| 49 |
/// and serves as the Query Tree leaf. More sophisticated examples of iterators |
| 50 |
/// can manage intersection, union of the elements produced by other iterators |
| 51 |
/// (their children) to form a multi-level Query Tree. The interface is designed |
| 52 |
/// to be extensible in order to support multiple types of iterators. |
| 53 |
class Iterator { |
| 54 |
public: |
| 55 |
/// Returns true if all valid DocIDs were processed and hence the iterator is |
| 56 |
/// exhausted. |
| 57 |
virtual bool reachedEnd() const = 0; |
| 58 |
/// Moves to next valid DocID. If it doesn't exist, the iterator is exhausted |
| 59 |
/// and proceeds to the END. |
| 60 |
/// |
| 61 |
/// Note: reachedEnd() must be false. |
| 62 |
virtual void advance() = 0; |
| 63 |
/// Moves to the first valid DocID which is equal or higher than given ID. If |
| 64 |
/// it doesn't exist, the iterator is exhausted and proceeds to the END. |
| 65 |
/// |
| 66 |
/// Note: reachedEnd() must be false. |
| 67 |
virtual void advanceTo(DocID ID) = 0; |
| 68 |
/// Returns the current element this iterator points to. |
| 69 |
/// |
| 70 |
/// Note: reachedEnd() must be false. |
| 71 |
virtual DocID peek() const = 0; |
| 72 |
/// Informs the iterator that the current document was consumed, and returns |
| 73 |
/// its boost. |
| 74 |
/// |
| 75 |
/// Note: If this iterator has any child iterators that contain the document, |
| 76 |
/// consume() should be called on those and their boosts incorporated. |
| 77 |
/// consume() must *not* be called on children that don't contain the current |
| 78 |
/// doc. |
| 79 |
virtual float consume() = 0; |
| 80 |
/// Returns an estimate of advance() calls before the iterator is exhausted. |
| 81 |
virtual size_t estimateSize() const = 0; |
| 82 |
|
| 83 |
virtual ~Iterator() {} |
| 84 |
|
| 85 |
/// Prints a convenient human-readable iterator representation by recursively |
| 86 |
/// dumping iterators in the following format: |
| 87 |
/// |
| 88 |
/// (Type Child1 Child2 ...) |
| 89 |
/// |
| 90 |
/// Where Type is the iterator type representation: "&" for And, "|" for Or, |
| 91 |
/// ChildN is N-th iterator child. Raw iterators over PostingList are |
| 92 |
/// represented as "[... CurID ...]" where CurID is the current PostingList |
| 93 |
/// entry being inspected. |
| 94 |
friend llvm::raw_ostream &operator<<(llvm::raw_ostream &OS, |
| 95 |
const Iterator &Iterator) { |
| 96 |
return Iterator.dump(OS); |
| 97 |
} |
| 98 |
|
| 99 |
/// Inspect iterator type, used internally for optimizing query trees. |
| 100 |
enum class Kind { And, Or, True, False, Other }; |
| 101 |
Kind kind() const { return MyKind; } |
| 102 |
|
| 103 |
protected: |
| 104 |
Iterator(Kind MyKind = Kind::Other) : MyKind(MyKind) {} |
| 105 |
|
| 106 |
private: |
| 107 |
virtual llvm::raw_ostream &dump(llvm::raw_ostream &OS) const = 0; |
| 108 |
Kind MyKind; |
| 109 |
}; |
| 110 |
|
| 111 |
/// Advances the iterator until it is exhausted. Returns pairs of document IDs |
| 112 |
/// with the corresponding boosting score. |
| 113 |
/// |
| 114 |
/// Boosting can be seen as a compromise between retrieving too many items and |
| 115 |
/// calculating finals score for each of them (which might be very expensive) |
| 116 |
/// and not retrieving enough items so that items with very high final score |
| 117 |
/// would not be processed. Boosting score is a computationally efficient way |
| 118 |
/// to acquire preliminary scores of requested items. |
| 119 |
std::vector<std::pair<DocID, float>> consume(Iterator &It); |
| 120 |
|
| 121 |
namespace detail { |
| 122 |
// Variadic template machinery. |
| 123 |
inline void populateChildren(std::vector<std::unique_ptr<Iterator>> &) {} |
| 124 |
template <typename... TailT> |
| 125 |
void populateChildren(std::vector<std::unique_ptr<Iterator>> &Children, |
| 126 |
std::unique_ptr<Iterator> Head, TailT... Tail) { |
| 127 |
Children.push_back(std::move(Head)); |
| 128 |
populateChildren(Children, std::move(Tail)...); |
| 129 |
} |
| 130 |
} // namespace detail |
| 131 |
|
| 132 |
// A corpus is a set of documents, and a factory for iterators over them. |
| 133 |
class Corpus { |
| 134 |
DocID Size; |
| 135 |
|
| 136 |
public: |
| 137 |
explicit Corpus(DocID Size) : Size(Size) {} |
| 138 |
|
| 139 |
/// Returns AND Iterator which performs the intersection of the PostingLists |
| 140 |
/// of its children. |
| 141 |
/// |
| 142 |
/// consume(): AND Iterator returns the product of Childrens' boosting |
| 143 |
/// scores. |
| 144 |
std::unique_ptr<Iterator> |
| 145 |
intersect(std::vector<std::unique_ptr<Iterator>> Children) const; |
| 146 |
|
| 147 |
/// Returns OR Iterator which performs the union of the PostingLists of its |
| 148 |
/// children. |
| 149 |
/// |
| 150 |
/// consume(): OR Iterator returns the highest boost value among children |
| 151 |
/// containing the requested item. |
| 152 |
std::unique_ptr<Iterator> |
| 153 |
unionOf(std::vector<std::unique_ptr<Iterator>> Children) const; |
| 154 |
|
| 155 |
/// Returns TRUE Iterator which iterates over "virtual" PostingList |
| 156 |
/// containing all items in range [0, Size) in an efficient manner. |
| 157 |
std::unique_ptr<Iterator> all() const; |
| 158 |
|
| 159 |
/// Returns FALSE Iterator which iterates over no documents. |
| 160 |
std::unique_ptr<Iterator> none() const; |
| 161 |
|
| 162 |
/// Returns BOOST iterator which multiplies the score of each item by given |
| 163 |
/// factor. Boosting can be used as a computationally inexpensive filtering. |
| 164 |
/// Users can return significantly more items using consumeAndBoost() and |
| 165 |
/// then trim Top K using retrieval score. |
| 166 |
std::unique_ptr<Iterator> boost(std::unique_ptr<Iterator> Child, |
| 167 |
float Factor) const; |
| 168 |
|
| 169 |
/// Returns LIMIT iterator, which yields up to N elements of its child |
| 170 |
/// iterator. Elements only count towards the limit if they are part of the |
| 171 |
/// final result set. Therefore the following iterator (AND (2) (LIMIT (1 2) |
| 172 |
/// 1)) yields (2), not (). |
| 173 |
std::unique_ptr<Iterator> limit(std::unique_ptr<Iterator> Child, |
| 174 |
size_t Limit) const; |
| 175 |
|
| 176 |
/// This allows intersect(create(...), create(...)) syntax. |
| 177 |
template <typename... Args> |
| 178 |
std::unique_ptr<Iterator> intersect(Args... args) const { |
| 179 |
std::vector<std::unique_ptr<Iterator>> Children; |
| 180 |
detail::populateChildren(Children, std::forward<Args>(args)...); |
| 181 |
return intersect(std::move(Children)); |
| 182 |
} |
| 183 |
|
| 184 |
/// This allows unionOf(create(...), create(...)) syntax. |
| 185 |
template <typename... Args> |
| 186 |
std::unique_ptr<Iterator> unionOf(Args... args) const { |
| 187 |
std::vector<std::unique_ptr<Iterator>> Children; |
| 188 |
detail::populateChildren(Children, std::forward<Args>(args)...); |
| 189 |
return unionOf(std::move(Children)); |
| 190 |
} |
| 191 |
}; |
| 192 |
|
| 193 |
} // namespace dex |
| 194 |
} // namespace clangd |
| 195 |
} // namespace clang |
| 196 |
|
| 197 |
#endif // LLVM_CLANG_TOOLS_EXTRA_CLANGD_INDEX_DEX_ITERATOR_H |