1 //===--- IdentifierTable.cpp - Hash table for identifier lookup -----------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements the IdentifierInfo, IdentifierVisitor, and
11 // IdentifierTable interfaces.
13 //===----------------------------------------------------------------------===//
15 #include "clang/Basic/CharInfo.h"
16 #include "clang/Basic/IdentifierTable.h"
17 #include "clang/Basic/LangOptions.h"
18 #include "clang/Basic/OperatorKinds.h"
19 #include "clang/Basic/Specifiers.h"
20 #include "llvm/ADT/DenseMap.h"
21 #include "llvm/ADT/FoldingSet.h"
22 #include "llvm/ADT/SmallString.h"
23 #include "llvm/Support/ErrorHandling.h"
24 #include "llvm/Support/raw_ostream.h"
27 using namespace clang;
29 //===----------------------------------------------------------------------===//
30 // IdentifierInfo Implementation
31 //===----------------------------------------------------------------------===//
33 IdentifierInfo::IdentifierInfo() {
34 TokenID = tok::identifier;
39 IsFutureCompatKeyword = false;
41 IsCPPOperatorKeyword = false;
42 NeedsHandleIdentifier = false;
44 ChangedAfterLoad = false;
45 FEChangedAfterLoad = false;
46 RevertedTokenID = false;
48 IsModulesImport = false;
49 FETokenInfo = nullptr;
53 //===----------------------------------------------------------------------===//
54 // IdentifierTable Implementation
55 //===----------------------------------------------------------------------===//
57 IdentifierIterator::~IdentifierIterator() { }
59 IdentifierInfoLookup::~IdentifierInfoLookup() {}
62 /// \brief A simple identifier lookup iterator that represents an
63 /// empty sequence of identifiers.
64 class EmptyLookupIterator : public IdentifierIterator
67 StringRef Next() override { return StringRef(); }
71 IdentifierIterator *IdentifierInfoLookup::getIdentifiers() {
72 return new EmptyLookupIterator();
75 IdentifierTable::IdentifierTable(const LangOptions &LangOpts,
76 IdentifierInfoLookup* externalLookup)
77 : HashTable(8192), // Start with space for 8K identifiers.
78 ExternalLookup(externalLookup) {
80 // Populate the identifier table with info about keywords for the current
82 AddKeywords(LangOpts);
85 // Add the '_experimental_modules_import' contextual keyword.
86 get("import").setModulesImport(true);
89 //===----------------------------------------------------------------------===//
90 // Language Keyword Implementation
91 //===----------------------------------------------------------------------===//
93 // Constants for TokenKinds.def
109 KEYNOOPENCL = 0x02000,
110 WCHARSUPPORT = 0x04000,
111 HALFSUPPORT = 0x08000,
112 KEYCONCEPTS = 0x10000,
114 KEYZVECTOR = 0x40000,
115 KEYCOROUTINES = 0x80000,
116 KEYMODULES = 0x100000,
117 KEYALL = (0x1fffff & ~KEYNOMS18 &
118 ~KEYNOOPENCL) // KEYNOMS18 and KEYNOOPENCL are used to exclude.
121 /// \brief How a keyword is treated in the selected standard.
123 KS_Disabled, // Disabled
124 KS_Extension, // Is an extension
125 KS_Enabled, // Enabled
126 KS_Future // Is a keyword in future standard
130 /// \brief Translates flags as specified in TokenKinds.def into keyword status
131 /// in the given language standard.
132 static KeywordStatus getKeywordStatus(const LangOptions &LangOpts,
134 if (Flags == KEYALL) return KS_Enabled;
135 if (LangOpts.CPlusPlus && (Flags & KEYCXX)) return KS_Enabled;
136 if (LangOpts.CPlusPlus11 && (Flags & KEYCXX11)) return KS_Enabled;
137 if (LangOpts.C99 && (Flags & KEYC99)) return KS_Enabled;
138 if (LangOpts.GNUKeywords && (Flags & KEYGNU)) return KS_Extension;
139 if (LangOpts.MicrosoftExt && (Flags & KEYMS)) return KS_Extension;
140 if (LangOpts.Borland && (Flags & KEYBORLAND)) return KS_Extension;
141 if (LangOpts.Bool && (Flags & BOOLSUPPORT)) return KS_Enabled;
142 if (LangOpts.Half && (Flags & HALFSUPPORT)) return KS_Enabled;
143 if (LangOpts.WChar && (Flags & WCHARSUPPORT)) return KS_Enabled;
144 if (LangOpts.AltiVec && (Flags & KEYALTIVEC)) return KS_Enabled;
145 if (LangOpts.OpenCL && (Flags & KEYOPENCL)) return KS_Enabled;
146 if (!LangOpts.CPlusPlus && (Flags & KEYNOCXX)) return KS_Enabled;
147 if (LangOpts.C11 && (Flags & KEYC11)) return KS_Enabled;
148 // We treat bridge casts as objective-C keywords so we can warn on them
150 if (LangOpts.ObjC2 && (Flags & KEYARC)) return KS_Enabled;
151 if (LangOpts.ObjC2 && (Flags & KEYOBJC2)) return KS_Enabled;
152 if (LangOpts.ConceptsTS && (Flags & KEYCONCEPTS)) return KS_Enabled;
153 if (LangOpts.CoroutinesTS && (Flags & KEYCOROUTINES)) return KS_Enabled;
154 if (LangOpts.ModulesTS && (Flags & KEYMODULES)) return KS_Enabled;
155 if (LangOpts.CPlusPlus && (Flags & KEYCXX11)) return KS_Future;
159 /// AddKeyword - This method is used to associate a token ID with specific
160 /// identifiers because they are language keywords. This causes the lexer to
161 /// automatically map matching identifiers to specialized token codes.
162 static void AddKeyword(StringRef Keyword,
163 tok::TokenKind TokenCode, unsigned Flags,
164 const LangOptions &LangOpts, IdentifierTable &Table) {
165 KeywordStatus AddResult = getKeywordStatus(LangOpts, Flags);
167 // Don't add this keyword under MSVCCompat.
168 if (LangOpts.MSVCCompat && (Flags & KEYNOMS18) &&
169 !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2015))
172 // Don't add this keyword under OpenCL.
173 if (LangOpts.OpenCL && (Flags & KEYNOOPENCL))
176 // Don't add this keyword if disabled in this language.
177 if (AddResult == KS_Disabled) return;
179 IdentifierInfo &Info =
180 Table.get(Keyword, AddResult == KS_Future ? tok::identifier : TokenCode);
181 Info.setIsExtensionToken(AddResult == KS_Extension);
182 Info.setIsFutureCompatKeyword(AddResult == KS_Future);
185 /// AddCXXOperatorKeyword - Register a C++ operator keyword alternative
187 static void AddCXXOperatorKeyword(StringRef Keyword,
188 tok::TokenKind TokenCode,
189 IdentifierTable &Table) {
190 IdentifierInfo &Info = Table.get(Keyword, TokenCode);
191 Info.setIsCPlusPlusOperatorKeyword();
194 /// AddObjCKeyword - Register an Objective-C \@keyword like "class" "selector"
196 static void AddObjCKeyword(StringRef Name,
197 tok::ObjCKeywordKind ObjCID,
198 IdentifierTable &Table) {
199 Table.get(Name).setObjCKeywordID(ObjCID);
202 /// AddKeywords - Add all keywords to the symbol table.
204 void IdentifierTable::AddKeywords(const LangOptions &LangOpts) {
205 // Add keywords and tokens for the current language.
206 #define KEYWORD(NAME, FLAGS) \
207 AddKeyword(StringRef(#NAME), tok::kw_ ## NAME, \
208 FLAGS, LangOpts, *this);
209 #define ALIAS(NAME, TOK, FLAGS) \
210 AddKeyword(StringRef(NAME), tok::kw_ ## TOK, \
211 FLAGS, LangOpts, *this);
212 #define CXX_KEYWORD_OPERATOR(NAME, ALIAS) \
213 if (LangOpts.CXXOperatorNames) \
214 AddCXXOperatorKeyword(StringRef(#NAME), tok::ALIAS, *this);
215 #define OBJC1_AT_KEYWORD(NAME) \
216 if (LangOpts.ObjC1) \
217 AddObjCKeyword(StringRef(#NAME), tok::objc_##NAME, *this);
218 #define OBJC2_AT_KEYWORD(NAME) \
219 if (LangOpts.ObjC2) \
220 AddObjCKeyword(StringRef(#NAME), tok::objc_##NAME, *this);
221 #define TESTING_KEYWORD(NAME, FLAGS)
222 #include "clang/Basic/TokenKinds.def"
224 if (LangOpts.ParseUnknownAnytype)
225 AddKeyword("__unknown_anytype", tok::kw___unknown_anytype, KEYALL,
228 if (LangOpts.DeclSpecKeyword)
229 AddKeyword("__declspec", tok::kw___declspec, KEYALL, LangOpts, *this);
232 /// \brief Checks if the specified token kind represents a keyword in the
233 /// specified language.
234 /// \returns Status of the keyword in the language.
235 static KeywordStatus getTokenKwStatus(const LangOptions &LangOpts,
238 #define KEYWORD(NAME, FLAGS) \
239 case tok::kw_##NAME: return getKeywordStatus(LangOpts, FLAGS);
240 #include "clang/Basic/TokenKinds.def"
241 default: return KS_Disabled;
245 /// \brief Returns true if the identifier represents a keyword in the
246 /// specified language.
247 bool IdentifierInfo::isKeyword(const LangOptions &LangOpts) {
248 switch (getTokenKwStatus(LangOpts, getTokenID())) {
257 tok::PPKeywordKind IdentifierInfo::getPPKeywordID() const {
258 // We use a perfect hash function here involving the length of the keyword,
259 // the first and third character. For preprocessor ID's there are no
260 // collisions (if there were, the switch below would complain about duplicate
261 // case values). Note that this depends on 'if' being null terminated.
263 #define HASH(LEN, FIRST, THIRD) \
264 (LEN << 5) + (((FIRST-'a') + (THIRD-'a')) & 31)
265 #define CASE(LEN, FIRST, THIRD, NAME) \
266 case HASH(LEN, FIRST, THIRD): \
267 return memcmp(Name, #NAME, LEN) ? tok::pp_not_keyword : tok::pp_ ## NAME
269 unsigned Len = getLength();
270 if (Len < 2) return tok::pp_not_keyword;
271 const char *Name = getNameStart();
272 switch (HASH(Len, Name[0], Name[2])) {
273 default: return tok::pp_not_keyword;
274 CASE( 2, 'i', '\0', if);
275 CASE( 4, 'e', 'i', elif);
276 CASE( 4, 'e', 's', else);
277 CASE( 4, 'l', 'n', line);
278 CASE( 4, 's', 'c', sccs);
279 CASE( 5, 'e', 'd', endif);
280 CASE( 5, 'e', 'r', error);
281 CASE( 5, 'i', 'e', ident);
282 CASE( 5, 'i', 'd', ifdef);
283 CASE( 5, 'u', 'd', undef);
285 CASE( 6, 'a', 's', assert);
286 CASE( 6, 'd', 'f', define);
287 CASE( 6, 'i', 'n', ifndef);
288 CASE( 6, 'i', 'p', import);
289 CASE( 6, 'p', 'a', pragma);
291 CASE( 7, 'd', 'f', defined);
292 CASE( 7, 'i', 'c', include);
293 CASE( 7, 'w', 'r', warning);
295 CASE( 8, 'u', 'a', unassert);
296 CASE(12, 'i', 'c', include_next);
298 CASE(14, '_', 'p', __public_macro);
300 CASE(15, '_', 'p', __private_macro);
302 CASE(16, '_', 'i', __include_macros);
308 //===----------------------------------------------------------------------===//
309 // Stats Implementation
310 //===----------------------------------------------------------------------===//
312 /// PrintStats - Print statistics about how well the identifier table is doing
313 /// at hashing identifiers.
314 void IdentifierTable::PrintStats() const {
315 unsigned NumBuckets = HashTable.getNumBuckets();
316 unsigned NumIdentifiers = HashTable.getNumItems();
317 unsigned NumEmptyBuckets = NumBuckets-NumIdentifiers;
318 unsigned AverageIdentifierSize = 0;
319 unsigned MaxIdentifierLength = 0;
321 // TODO: Figure out maximum times an identifier had to probe for -stats.
322 for (llvm::StringMap<IdentifierInfo*, llvm::BumpPtrAllocator>::const_iterator
323 I = HashTable.begin(), E = HashTable.end(); I != E; ++I) {
324 unsigned IdLen = I->getKeyLength();
325 AverageIdentifierSize += IdLen;
326 if (MaxIdentifierLength < IdLen)
327 MaxIdentifierLength = IdLen;
330 fprintf(stderr, "\n*** Identifier Table Stats:\n");
331 fprintf(stderr, "# Identifiers: %d\n", NumIdentifiers);
332 fprintf(stderr, "# Empty Buckets: %d\n", NumEmptyBuckets);
333 fprintf(stderr, "Hash density (#identifiers per bucket): %f\n",
334 NumIdentifiers/(double)NumBuckets);
335 fprintf(stderr, "Ave identifier length: %f\n",
336 (AverageIdentifierSize/(double)NumIdentifiers));
337 fprintf(stderr, "Max identifier length: %d\n", MaxIdentifierLength);
339 // Compute statistics about the memory allocated for identifiers.
340 HashTable.getAllocator().PrintStats();
343 //===----------------------------------------------------------------------===//
344 // SelectorTable Implementation
345 //===----------------------------------------------------------------------===//
347 unsigned llvm::DenseMapInfo<clang::Selector>::getHashValue(clang::Selector S) {
348 return DenseMapInfo<void*>::getHashValue(S.getAsOpaquePtr());
352 /// MultiKeywordSelector - One of these variable length records is kept for each
353 /// selector containing more than one keyword. We use a folding set
354 /// to unique aggregate names (keyword selectors in ObjC parlance). Access to
355 /// this class is provided strictly through Selector.
356 class MultiKeywordSelector
357 : public DeclarationNameExtra, public llvm::FoldingSetNode {
358 MultiKeywordSelector(unsigned nKeys) {
359 ExtraKindOrNumArgs = NUM_EXTRA_KINDS + nKeys;
362 // Constructor for keyword selectors.
363 MultiKeywordSelector(unsigned nKeys, IdentifierInfo **IIV) {
364 assert((nKeys > 1) && "not a multi-keyword selector");
365 ExtraKindOrNumArgs = NUM_EXTRA_KINDS + nKeys;
367 // Fill in the trailing keyword array.
368 IdentifierInfo **KeyInfo = reinterpret_cast<IdentifierInfo **>(this+1);
369 for (unsigned i = 0; i != nKeys; ++i)
373 // getName - Derive the full selector name and return it.
374 std::string getName() const;
376 unsigned getNumArgs() const { return ExtraKindOrNumArgs - NUM_EXTRA_KINDS; }
378 typedef IdentifierInfo *const *keyword_iterator;
379 keyword_iterator keyword_begin() const {
380 return reinterpret_cast<keyword_iterator>(this+1);
382 keyword_iterator keyword_end() const {
383 return keyword_begin()+getNumArgs();
385 IdentifierInfo *getIdentifierInfoForSlot(unsigned i) const {
386 assert(i < getNumArgs() && "getIdentifierInfoForSlot(): illegal index");
387 return keyword_begin()[i];
389 static void Profile(llvm::FoldingSetNodeID &ID,
390 keyword_iterator ArgTys, unsigned NumArgs) {
391 ID.AddInteger(NumArgs);
392 for (unsigned i = 0; i != NumArgs; ++i)
393 ID.AddPointer(ArgTys[i]);
395 void Profile(llvm::FoldingSetNodeID &ID) {
396 Profile(ID, keyword_begin(), getNumArgs());
399 } // end namespace clang.
401 unsigned Selector::getNumArgs() const {
402 unsigned IIF = getIdentifierInfoFlag();
407 // We point to a MultiKeywordSelector.
408 MultiKeywordSelector *SI = getMultiKeywordSelector();
409 return SI->getNumArgs();
412 IdentifierInfo *Selector::getIdentifierInfoForSlot(unsigned argIndex) const {
413 if (getIdentifierInfoFlag() < MultiArg) {
414 assert(argIndex == 0 && "illegal keyword index");
415 return getAsIdentifierInfo();
417 // We point to a MultiKeywordSelector.
418 MultiKeywordSelector *SI = getMultiKeywordSelector();
419 return SI->getIdentifierInfoForSlot(argIndex);
422 StringRef Selector::getNameForSlot(unsigned int argIndex) const {
423 IdentifierInfo *II = getIdentifierInfoForSlot(argIndex);
424 return II? II->getName() : StringRef();
427 std::string MultiKeywordSelector::getName() const {
428 SmallString<256> Str;
429 llvm::raw_svector_ostream OS(Str);
430 for (keyword_iterator I = keyword_begin(), E = keyword_end(); I != E; ++I) {
432 OS << (*I)->getName();
439 std::string Selector::getAsString() const {
441 return "<null selector>";
443 if (getIdentifierInfoFlag() < MultiArg) {
444 IdentifierInfo *II = getAsIdentifierInfo();
446 if (getNumArgs() == 0) {
447 assert(II && "If the number of arguments is 0 then II is guaranteed to "
449 return II->getName();
455 return II->getName().str() + ":";
458 // We have a multiple keyword selector.
459 return getMultiKeywordSelector()->getName();
462 void Selector::print(llvm::raw_ostream &OS) const {
466 /// Interpreting the given string using the normal CamelCase
467 /// conventions, determine whether the given string starts with the
468 /// given "word", which is assumed to end in a lowercase letter.
469 static bool startsWithWord(StringRef name, StringRef word) {
470 if (name.size() < word.size()) return false;
471 return ((name.size() == word.size() || !isLowercase(name[word.size()])) &&
472 name.startswith(word));
475 ObjCMethodFamily Selector::getMethodFamilyImpl(Selector sel) {
476 IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
477 if (!first) return OMF_None;
479 StringRef name = first->getName();
480 if (sel.isUnarySelector()) {
481 if (name == "autorelease") return OMF_autorelease;
482 if (name == "dealloc") return OMF_dealloc;
483 if (name == "finalize") return OMF_finalize;
484 if (name == "release") return OMF_release;
485 if (name == "retain") return OMF_retain;
486 if (name == "retainCount") return OMF_retainCount;
487 if (name == "self") return OMF_self;
488 if (name == "initialize") return OMF_initialize;
491 if (name == "performSelector") return OMF_performSelector;
493 // The other method families may begin with a prefix of underscores.
494 while (!name.empty() && name.front() == '_')
495 name = name.substr(1);
497 if (name.empty()) return OMF_None;
498 switch (name.front()) {
500 if (startsWithWord(name, "alloc")) return OMF_alloc;
503 if (startsWithWord(name, "copy")) return OMF_copy;
506 if (startsWithWord(name, "init")) return OMF_init;
509 if (startsWithWord(name, "mutableCopy")) return OMF_mutableCopy;
512 if (startsWithWord(name, "new")) return OMF_new;
521 ObjCInstanceTypeFamily Selector::getInstTypeMethodFamily(Selector sel) {
522 IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
523 if (!first) return OIT_None;
525 StringRef name = first->getName();
527 if (name.empty()) return OIT_None;
528 switch (name.front()) {
530 if (startsWithWord(name, "array")) return OIT_Array;
533 if (startsWithWord(name, "default")) return OIT_ReturnsSelf;
534 if (startsWithWord(name, "dictionary")) return OIT_Dictionary;
537 if (startsWithWord(name, "shared")) return OIT_ReturnsSelf;
538 if (startsWithWord(name, "standard")) return OIT_Singleton;
540 if (startsWithWord(name, "init")) return OIT_Init;
547 ObjCStringFormatFamily Selector::getStringFormatFamilyImpl(Selector sel) {
548 IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
549 if (!first) return SFF_None;
551 StringRef name = first->getName();
553 switch (name.front()) {
555 if (name == "appendFormat") return SFF_NSString;
559 if (name == "initWithFormat") return SFF_NSString;
563 if (name == "localizedStringWithFormat") return SFF_NSString;
567 if (name == "stringByAppendingFormat" ||
568 name == "stringWithFormat") return SFF_NSString;
575 struct SelectorTableImpl {
576 llvm::FoldingSet<MultiKeywordSelector> Table;
577 llvm::BumpPtrAllocator Allocator;
579 } // end anonymous namespace.
581 static SelectorTableImpl &getSelectorTableImpl(void *P) {
582 return *static_cast<SelectorTableImpl*>(P);
586 SelectorTable::constructSetterName(StringRef Name) {
587 SmallString<64> SetterName("set");
589 SetterName[3] = toUppercase(SetterName[3]);
594 SelectorTable::constructSetterSelector(IdentifierTable &Idents,
595 SelectorTable &SelTable,
596 const IdentifierInfo *Name) {
597 IdentifierInfo *SetterName =
598 &Idents.get(constructSetterName(Name->getName()));
599 return SelTable.getUnarySelector(SetterName);
602 size_t SelectorTable::getTotalMemory() const {
603 SelectorTableImpl &SelTabImpl = getSelectorTableImpl(Impl);
604 return SelTabImpl.Allocator.getTotalMemory();
607 Selector SelectorTable::getSelector(unsigned nKeys, IdentifierInfo **IIV) {
609 return Selector(IIV[0], nKeys);
611 SelectorTableImpl &SelTabImpl = getSelectorTableImpl(Impl);
613 // Unique selector, to guarantee there is one per name.
614 llvm::FoldingSetNodeID ID;
615 MultiKeywordSelector::Profile(ID, IIV, nKeys);
617 void *InsertPos = nullptr;
618 if (MultiKeywordSelector *SI =
619 SelTabImpl.Table.FindNodeOrInsertPos(ID, InsertPos))
622 // MultiKeywordSelector objects are not allocated with new because they have a
623 // variable size array (for parameter types) at the end of them.
624 unsigned Size = sizeof(MultiKeywordSelector) + nKeys*sizeof(IdentifierInfo *);
625 MultiKeywordSelector *SI =
626 (MultiKeywordSelector *)SelTabImpl.Allocator.Allocate(
627 Size, alignof(MultiKeywordSelector));
628 new (SI) MultiKeywordSelector(nKeys, IIV);
629 SelTabImpl.Table.InsertNode(SI, InsertPos);
633 SelectorTable::SelectorTable() {
634 Impl = new SelectorTableImpl();
637 SelectorTable::~SelectorTable() {
638 delete &getSelectorTableImpl(Impl);
641 const char *clang::getOperatorSpelling(OverloadedOperatorKind Operator) {
644 case NUM_OVERLOADED_OPERATORS:
647 #define OVERLOADED_OPERATOR(Name,Spelling,Token,Unary,Binary,MemberOnly) \
648 case OO_##Name: return Spelling;
649 #include "clang/Basic/OperatorKinds.def"
652 llvm_unreachable("Invalid OverloadedOperatorKind!");
655 StringRef clang::getNullabilitySpelling(NullabilityKind kind,
656 bool isContextSensitive) {
658 case NullabilityKind::NonNull:
659 return isContextSensitive ? "nonnull" : "_Nonnull";
661 case NullabilityKind::Nullable:
662 return isContextSensitive ? "nullable" : "_Nullable";
664 case NullabilityKind::Unspecified:
665 return isContextSensitive ? "null_unspecified" : "_Null_unspecified";
667 llvm_unreachable("Unknown nullability kind.");