Details | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line | 
|---|---|---|---|
| 14 | pmbaty | 1 | //===-- TrigramIndex.h - a heuristic for SpecialCaseList --------*- 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 | // TrigramIndex implements a heuristic for SpecialCaseList that allows to | ||
| 9 | // filter out ~99% incoming queries when all regular expressions in the | ||
| 10 | // SpecialCaseList are simple wildcards with '*' and '.'. If rules are more | ||
| 11 | // complicated, the check is defeated and it will always pass the queries to a | ||
| 12 | // full regex. | ||
| 13 | // | ||
| 14 | // The basic idea is that in order for a wildcard to match a query, the query | ||
| 15 | // needs to have all trigrams which occur in the wildcard. We create a trigram | ||
| 16 | // index (trigram -> list of rules with it) and then count trigrams in the query | ||
| 17 | // for each rule. If the count for one of the rules reaches the expected value, | ||
| 18 | // the check passes the query to a regex. If none of the rules got enough | ||
| 19 | // trigrams, the check tells that the query is definitely not matched by any | ||
| 20 | // of the rules, and no regex matching is needed. | ||
| 21 | // A similar idea was used in Google Code Search as described in the blog post: | ||
| 22 | // https://swtch.com/~rsc/regexp/regexp4.html | ||
| 23 | // | ||
| 24 | //===----------------------------------------------------------------------===// | ||
| 25 | |||
| 26 | #ifndef LLVM_SUPPORT_TRIGRAMINDEX_H | ||
| 27 | #define LLVM_SUPPORT_TRIGRAMINDEX_H | ||
| 28 | |||
| 29 | #include "llvm/ADT/SmallVector.h" | ||
| 30 | #include <string> | ||
| 31 | #include <unordered_map> | ||
| 32 | #include <vector> | ||
| 33 | |||
| 34 | namespace llvm { | ||
| 35 | class StringRef; | ||
| 36 | |||
| 37 | class TrigramIndex { | ||
| 38 | public: | ||
| 39 |   /// Inserts a new Regex into the index. | ||
| 40 | void insert(const std::string &Regex); | ||
| 41 | |||
| 42 |   /// Returns true, if special case list definitely does not have a line | ||
| 43 |   /// that matches the query. Returns false, if it's not sure. | ||
| 44 | bool isDefinitelyOut(StringRef Query) const; | ||
| 45 | |||
| 46 |   /// Returned true, iff the heuristic is defeated and not useful. | ||
| 47 |   /// In this case isDefinitelyOut always returns false. | ||
| 48 | bool isDefeated() { return Defeated; } | ||
| 49 | private: | ||
| 50 |   // If true, the rules are too complicated for the check to work, and full | ||
| 51 |   // regex matching is needed for every rule. | ||
| 52 | bool Defeated = false; | ||
| 53 |   // The minimum number of trigrams which should match for a rule to have a | ||
| 54 |   // chance to match the query. The number of elements equals the number of | ||
| 55 |   // regex rules in the SpecialCaseList. | ||
| 56 | std::vector<unsigned> Counts; | ||
| 57 |   // Index holds a list of rules indices for each trigram. The same indices | ||
| 58 |   // are used in Counts to store per-rule limits. | ||
| 59 |   // If a trigram is too common (>4 rules with it), we stop tracking it, | ||
| 60 |   // which increases the probability for a need to match using regex, but | ||
| 61 |   // decreases the costs in the regular case. | ||
| 62 | std::unordered_map<unsigned, SmallVector<size_t, 4>> Index{256}; | ||
| 63 | }; | ||
| 64 | |||
| 65 | } // namespace llvm | ||
| 66 | |||
| 67 | #endif  // LLVM_SUPPORT_TRIGRAMINDEX_H |