1//===- StringMatcher.cpp - Generate a matcher for input strings -----------===//
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// This file implements the StringMatcher class.
10//
11//===----------------------------------------------------------------------===//
12
13#include "llvm/TableGen/StringMatcher.h"
14#include "llvm/ADT/STLExtras.h"
15#include "llvm/Support/ErrorHandling.h"
16#include "llvm/Support/raw_ostream.h"
17#include <cassert>
18#include <map>
19#include <string>
20#include <utility>
21#include <vector>
22
23using namespace llvm;
24
25/// FindFirstNonCommonLetter - Find the first character in the keys of the
26/// string pairs that is not shared across the whole set of strings. All
27/// strings are assumed to have the same length.
28static unsigned
29FindFirstNonCommonLetter(ArrayRef<const StringMatcher::StringPair *> Matches) {
30 assert(!Matches.empty());
31 for (auto [Idx, Letter] : enumerate(First: Matches[0]->first)) {
32 // Check to see if `Letter` is the same across the set. Since the letter is
33 // from `Matches[0]`, we can skip `Matches[0]` in the loop below.
34 for (const StringMatcher::StringPair *Match : Matches.drop_front())
35 if (Match->first[Idx] != Letter)
36 return Idx;
37 }
38
39 return Matches[0]->first.size();
40}
41
42/// EmitStringMatcherForChar - Given a set of strings that are known to be the
43/// same length and whose characters leading up to CharNo are the same, emit
44/// code to verify that CharNo and later are the same.
45///
46/// \return - True if control can leave the emitted code fragment.
47bool StringMatcher::EmitStringMatcherForChar(
48 ArrayRef<const StringPair *> Matches, unsigned CharNo, unsigned IndentCount,
49 bool IgnoreDuplicates) const {
50 assert(!Matches.empty() && "Must have at least one string to match!");
51 std::string Indent(IndentCount * 2 + 4, ' ');
52
53 // If we have verified that the entire string matches, we're done: output the
54 // matching code.
55 if (CharNo == Matches[0]->first.size()) {
56 if (Matches.size() > 1 && !IgnoreDuplicates)
57 report_fatal_error(reason: "Had duplicate keys to match on");
58
59 // If the to-execute code has \n's in it, indent each subsequent line.
60 StringRef Code = Matches[0]->second;
61
62 std::pair<StringRef, StringRef> Split = Code.split(Separator: '\n');
63 OS << Indent << Split.first << "\t // \"" << Matches[0]->first << "\"\n";
64
65 Code = Split.second;
66 while (!Code.empty()) {
67 Split = Code.split(Separator: '\n');
68 OS << Indent << Split.first << "\n";
69 Code = Split.second;
70 }
71 return false;
72 }
73
74 // Bucket the matches by the character we are comparing.
75 std::map<char, std::vector<const StringPair*>> MatchesByLetter;
76
77 for (const StringPair *Match : Matches)
78 MatchesByLetter[Match->first[CharNo]].push_back(x: Match);
79
80 // If we have exactly one bucket to match, see how many characters are common
81 // across the whole set and match all of them at once.
82 if (MatchesByLetter.size() == 1) {
83 unsigned FirstNonCommonLetter = FindFirstNonCommonLetter(Matches);
84 unsigned NumChars = FirstNonCommonLetter-CharNo;
85
86 // Emit code to break out if the prefix doesn't match.
87 if (NumChars == 1) {
88 // Do the comparison with if (Str[1] != 'f')
89 // FIXME: Need to escape general characters.
90 OS << Indent << "if (" << StrVariableName << "[" << CharNo << "] != '"
91 << Matches[0]->first[CharNo] << "')\n";
92 OS << Indent << " break;\n";
93 } else {
94 // Do the comparison with if memcmp(Str.data()+1, "foo", 3).
95 // FIXME: Need to escape general strings.
96 OS << Indent << "if (memcmp(" << StrVariableName << ".data()+" << CharNo
97 << ", \"" << Matches[0]->first.substr(pos: CharNo, n: NumChars) << "\", "
98 << NumChars << ") != 0)\n";
99 OS << Indent << " break;\n";
100 }
101
102 return EmitStringMatcherForChar(Matches, CharNo: FirstNonCommonLetter, IndentCount,
103 IgnoreDuplicates);
104 }
105
106 // Otherwise, we have multiple possible things, emit a switch on the
107 // character.
108 OS << Indent << "switch (" << StrVariableName << "[" << CharNo << "]) {\n";
109 OS << Indent << "default: break;\n";
110
111 for (const auto &[Letter, Matches] : MatchesByLetter) {
112 // TODO: escape hard stuff (like \n) if we ever care about it.
113 OS << Indent << "case '" << Letter << "':\t // " << Matches.size()
114 << " string";
115 if (Matches.size() != 1)
116 OS << 's';
117 OS << " to match.\n";
118 if (EmitStringMatcherForChar(Matches, CharNo: CharNo + 1, IndentCount: IndentCount + 1,
119 IgnoreDuplicates))
120 OS << Indent << " break;\n";
121 }
122
123 OS << Indent << "}\n";
124 return true;
125}
126
127/// Emit - Top level entry point.
128///
129void StringMatcher::Emit(unsigned Indent, bool IgnoreDuplicates) const {
130 // If nothing to match, just fall through.
131 if (Matches.empty()) return;
132
133 // First level categorization: group strings by length.
134 std::map<unsigned, std::vector<const StringPair*>> MatchesByLength;
135
136 for (const StringPair &Match : Matches)
137 MatchesByLength[Match.first.size()].push_back(x: &Match);
138
139 // Output a switch statement on length and categorize the elements within each
140 // bin.
141 OS.indent(NumSpaces: Indent*2+2) << "switch (" << StrVariableName << ".size()) {\n";
142 OS.indent(NumSpaces: Indent*2+2) << "default: break;\n";
143
144 for (const auto &[Length, Matches] : MatchesByLength) {
145 OS.indent(NumSpaces: Indent * 2 + 2)
146 << "case " << Length << ":\t // " << Matches.size() << " string"
147 << (Matches.size() == 1 ? "" : "s") << " to match.\n";
148 if (EmitStringMatcherForChar(Matches, CharNo: 0, IndentCount: Indent, IgnoreDuplicates))
149 OS.indent(NumSpaces: Indent*2+4) << "break;\n";
150 }
151
152 OS.indent(NumSpaces: Indent*2+2) << "}\n";
153}
154