10b57cec5SDimitry Andric //===- StringMatcher.cpp - Generate a matcher for input strings -----------===// 20b57cec5SDimitry Andric // 30b57cec5SDimitry Andric // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 40b57cec5SDimitry Andric // See https://llvm.org/LICENSE.txt for license information. 50b57cec5SDimitry Andric // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 60b57cec5SDimitry Andric // 70b57cec5SDimitry Andric //===----------------------------------------------------------------------===// 80b57cec5SDimitry Andric // 90b57cec5SDimitry Andric // This file implements the StringMatcher class. 100b57cec5SDimitry Andric // 110b57cec5SDimitry Andric //===----------------------------------------------------------------------===// 120b57cec5SDimitry Andric 130b57cec5SDimitry Andric #include "llvm/TableGen/StringMatcher.h" 140b57cec5SDimitry Andric #include "llvm/ADT/StringRef.h" 150b57cec5SDimitry Andric #include "llvm/Support/raw_ostream.h" 160b57cec5SDimitry Andric #include <cassert> 170b57cec5SDimitry Andric #include <map> 180b57cec5SDimitry Andric #include <string> 190b57cec5SDimitry Andric #include <utility> 200b57cec5SDimitry Andric #include <vector> 210b57cec5SDimitry Andric 220b57cec5SDimitry Andric using namespace llvm; 230b57cec5SDimitry Andric 240b57cec5SDimitry Andric /// FindFirstNonCommonLetter - Find the first character in the keys of the 250b57cec5SDimitry Andric /// string pairs that is not shared across the whole set of strings. All 260b57cec5SDimitry Andric /// strings are assumed to have the same length. 270b57cec5SDimitry Andric static unsigned 280b57cec5SDimitry Andric FindFirstNonCommonLetter(const std::vector<const 290b57cec5SDimitry Andric StringMatcher::StringPair*> &Matches) { 300b57cec5SDimitry Andric assert(!Matches.empty()); 310b57cec5SDimitry Andric for (unsigned i = 0, e = Matches[0]->first.size(); i != e; ++i) { 320b57cec5SDimitry Andric // Check to see if letter i is the same across the set. 330b57cec5SDimitry Andric char Letter = Matches[0]->first[i]; 340b57cec5SDimitry Andric 350b57cec5SDimitry Andric for (unsigned str = 0, e = Matches.size(); str != e; ++str) 360b57cec5SDimitry Andric if (Matches[str]->first[i] != Letter) 370b57cec5SDimitry Andric return i; 380b57cec5SDimitry Andric } 390b57cec5SDimitry Andric 400b57cec5SDimitry Andric return Matches[0]->first.size(); 410b57cec5SDimitry Andric } 420b57cec5SDimitry Andric 430b57cec5SDimitry Andric /// EmitStringMatcherForChar - Given a set of strings that are known to be the 440b57cec5SDimitry Andric /// same length and whose characters leading up to CharNo are the same, emit 450b57cec5SDimitry Andric /// code to verify that CharNo and later are the same. 460b57cec5SDimitry Andric /// 470b57cec5SDimitry Andric /// \return - True if control can leave the emitted code fragment. 480b57cec5SDimitry Andric bool StringMatcher::EmitStringMatcherForChar( 490b57cec5SDimitry Andric const std::vector<const StringPair *> &Matches, unsigned CharNo, 500b57cec5SDimitry Andric unsigned IndentCount, bool IgnoreDuplicates) const { 510b57cec5SDimitry Andric assert(!Matches.empty() && "Must have at least one string to match!"); 520b57cec5SDimitry Andric std::string Indent(IndentCount * 2 + 4, ' '); 530b57cec5SDimitry Andric 540b57cec5SDimitry Andric // If we have verified that the entire string matches, we're done: output the 550b57cec5SDimitry Andric // matching code. 560b57cec5SDimitry Andric if (CharNo == Matches[0]->first.size()) { 570b57cec5SDimitry Andric if (Matches.size() > 1 && !IgnoreDuplicates) 580b57cec5SDimitry Andric report_fatal_error("Had duplicate keys to match on"); 590b57cec5SDimitry Andric 600b57cec5SDimitry Andric // If the to-execute code has \n's in it, indent each subsequent line. 610b57cec5SDimitry Andric StringRef Code = Matches[0]->second; 620b57cec5SDimitry Andric 630b57cec5SDimitry Andric std::pair<StringRef, StringRef> Split = Code.split('\n'); 640b57cec5SDimitry Andric OS << Indent << Split.first << "\t // \"" << Matches[0]->first << "\"\n"; 650b57cec5SDimitry Andric 660b57cec5SDimitry Andric Code = Split.second; 670b57cec5SDimitry Andric while (!Code.empty()) { 680b57cec5SDimitry Andric Split = Code.split('\n'); 690b57cec5SDimitry Andric OS << Indent << Split.first << "\n"; 700b57cec5SDimitry Andric Code = Split.second; 710b57cec5SDimitry Andric } 720b57cec5SDimitry Andric return false; 730b57cec5SDimitry Andric } 740b57cec5SDimitry Andric 750b57cec5SDimitry Andric // Bucket the matches by the character we are comparing. 760b57cec5SDimitry Andric std::map<char, std::vector<const StringPair*>> MatchesByLetter; 770b57cec5SDimitry Andric 780b57cec5SDimitry Andric for (unsigned i = 0, e = Matches.size(); i != e; ++i) 790b57cec5SDimitry Andric MatchesByLetter[Matches[i]->first[CharNo]].push_back(Matches[i]); 800b57cec5SDimitry Andric 810b57cec5SDimitry Andric 820b57cec5SDimitry Andric // If we have exactly one bucket to match, see how many characters are common 830b57cec5SDimitry Andric // across the whole set and match all of them at once. 840b57cec5SDimitry Andric if (MatchesByLetter.size() == 1) { 850b57cec5SDimitry Andric unsigned FirstNonCommonLetter = FindFirstNonCommonLetter(Matches); 860b57cec5SDimitry Andric unsigned NumChars = FirstNonCommonLetter-CharNo; 870b57cec5SDimitry Andric 880b57cec5SDimitry Andric // Emit code to break out if the prefix doesn't match. 890b57cec5SDimitry Andric if (NumChars == 1) { 900b57cec5SDimitry Andric // Do the comparison with if (Str[1] != 'f') 910b57cec5SDimitry Andric // FIXME: Need to escape general characters. 920b57cec5SDimitry Andric OS << Indent << "if (" << StrVariableName << "[" << CharNo << "] != '" 930b57cec5SDimitry Andric << Matches[0]->first[CharNo] << "')\n"; 940b57cec5SDimitry Andric OS << Indent << " break;\n"; 950b57cec5SDimitry Andric } else { 960b57cec5SDimitry Andric // Do the comparison with if memcmp(Str.data()+1, "foo", 3). 970b57cec5SDimitry Andric // FIXME: Need to escape general strings. 980b57cec5SDimitry Andric OS << Indent << "if (memcmp(" << StrVariableName << ".data()+" << CharNo 990b57cec5SDimitry Andric << ", \"" << Matches[0]->first.substr(CharNo, NumChars) << "\", " 1000b57cec5SDimitry Andric << NumChars << ") != 0)\n"; 1010b57cec5SDimitry Andric OS << Indent << " break;\n"; 1020b57cec5SDimitry Andric } 1030b57cec5SDimitry Andric 1040b57cec5SDimitry Andric return EmitStringMatcherForChar(Matches, FirstNonCommonLetter, IndentCount, 1050b57cec5SDimitry Andric IgnoreDuplicates); 1060b57cec5SDimitry Andric } 1070b57cec5SDimitry Andric 1080b57cec5SDimitry Andric // Otherwise, we have multiple possible things, emit a switch on the 1090b57cec5SDimitry Andric // character. 1100b57cec5SDimitry Andric OS << Indent << "switch (" << StrVariableName << "[" << CharNo << "]) {\n"; 1110b57cec5SDimitry Andric OS << Indent << "default: break;\n"; 1120b57cec5SDimitry Andric 1130b57cec5SDimitry Andric for (std::map<char, std::vector<const StringPair*>>::iterator LI = 1140b57cec5SDimitry Andric MatchesByLetter.begin(), E = MatchesByLetter.end(); LI != E; ++LI) { 1150b57cec5SDimitry Andric // TODO: escape hard stuff (like \n) if we ever care about it. 1160b57cec5SDimitry Andric OS << Indent << "case '" << LI->first << "':\t // " 1170b57cec5SDimitry Andric << LI->second.size() << " string"; 1180b57cec5SDimitry Andric if (LI->second.size() != 1) OS << 's'; 1190b57cec5SDimitry Andric OS << " to match.\n"; 1200b57cec5SDimitry Andric if (EmitStringMatcherForChar(LI->second, CharNo + 1, IndentCount + 1, 1210b57cec5SDimitry Andric IgnoreDuplicates)) 1220b57cec5SDimitry Andric OS << Indent << " break;\n"; 1230b57cec5SDimitry Andric } 1240b57cec5SDimitry Andric 1250b57cec5SDimitry Andric OS << Indent << "}\n"; 1260b57cec5SDimitry Andric return true; 1270b57cec5SDimitry Andric } 1280b57cec5SDimitry Andric 1290b57cec5SDimitry Andric /// Emit - Top level entry point. 1300b57cec5SDimitry Andric /// 1310b57cec5SDimitry Andric void StringMatcher::Emit(unsigned Indent, bool IgnoreDuplicates) const { 1320b57cec5SDimitry Andric // If nothing to match, just fall through. 1330b57cec5SDimitry Andric if (Matches.empty()) return; 1340b57cec5SDimitry Andric 1350b57cec5SDimitry Andric // First level categorization: group strings by length. 1360b57cec5SDimitry Andric std::map<unsigned, std::vector<const StringPair*>> MatchesByLength; 1370b57cec5SDimitry Andric 1380b57cec5SDimitry Andric for (unsigned i = 0, e = Matches.size(); i != e; ++i) 1390b57cec5SDimitry Andric MatchesByLength[Matches[i].first.size()].push_back(&Matches[i]); 1400b57cec5SDimitry Andric 1410b57cec5SDimitry Andric // Output a switch statement on length and categorize the elements within each 1420b57cec5SDimitry Andric // bin. 1430b57cec5SDimitry Andric OS.indent(Indent*2+2) << "switch (" << StrVariableName << ".size()) {\n"; 1440b57cec5SDimitry Andric OS.indent(Indent*2+2) << "default: break;\n"; 1450b57cec5SDimitry Andric 1460b57cec5SDimitry Andric for (std::map<unsigned, std::vector<const StringPair*>>::iterator LI = 1470b57cec5SDimitry Andric MatchesByLength.begin(), E = MatchesByLength.end(); LI != E; ++LI) { 1480b57cec5SDimitry Andric OS.indent(Indent*2+2) << "case " << LI->first << ":\t // " 1490b57cec5SDimitry Andric << LI->second.size() 1500b57cec5SDimitry Andric << " string" << (LI->second.size() == 1 ? "" : "s") << " to match.\n"; 1510b57cec5SDimitry Andric if (EmitStringMatcherForChar(LI->second, 0, Indent, IgnoreDuplicates)) 1520b57cec5SDimitry Andric OS.indent(Indent*2+4) << "break;\n"; 1530b57cec5SDimitry Andric } 1540b57cec5SDimitry Andric 1550b57cec5SDimitry Andric OS.indent(Indent*2+2) << "}\n"; 1560b57cec5SDimitry Andric } 157