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) const {
248 switch (getTokenKwStatus(LangOpts, getTokenID())) {
257 /// \brief Returns true if the identifier represents a C++ keyword in the
258 /// specified language.
259 bool IdentifierInfo::isCPlusPlusKeyword(const LangOptions &LangOpts) const {
260 if (!LangOpts.CPlusPlus || !isKeyword(LangOpts))
262 // This is a C++ keyword if this identifier is not a keyword when checked
263 // using LangOptions without C++ support.
264 LangOptions LangOptsNoCPP = LangOpts;
265 LangOptsNoCPP.CPlusPlus = false;
266 LangOptsNoCPP.CPlusPlus11 = false;
267 return !isKeyword(LangOptsNoCPP);
270 tok::PPKeywordKind IdentifierInfo::getPPKeywordID() const {
271 // We use a perfect hash function here involving the length of the keyword,
272 // the first and third character. For preprocessor ID's there are no
273 // collisions (if there were, the switch below would complain about duplicate
274 // case values). Note that this depends on 'if' being null terminated.
276 #define HASH(LEN, FIRST, THIRD) \
277 (LEN << 5) + (((FIRST-'a') + (THIRD-'a')) & 31)
278 #define CASE(LEN, FIRST, THIRD, NAME) \
279 case HASH(LEN, FIRST, THIRD): \
280 return memcmp(Name, #NAME, LEN) ? tok::pp_not_keyword : tok::pp_ ## NAME
282 unsigned Len = getLength();
283 if (Len < 2) return tok::pp_not_keyword;
284 const char *Name = getNameStart();
285 switch (HASH(Len, Name[0], Name[2])) {
286 default: return tok::pp_not_keyword;
287 CASE( 2, 'i', '\0', if);
288 CASE( 4, 'e', 'i', elif);
289 CASE( 4, 'e', 's', else);
290 CASE( 4, 'l', 'n', line);
291 CASE( 4, 's', 'c', sccs);
292 CASE( 5, 'e', 'd', endif);
293 CASE( 5, 'e', 'r', error);
294 CASE( 5, 'i', 'e', ident);
295 CASE( 5, 'i', 'd', ifdef);
296 CASE( 5, 'u', 'd', undef);
298 CASE( 6, 'a', 's', assert);
299 CASE( 6, 'd', 'f', define);
300 CASE( 6, 'i', 'n', ifndef);
301 CASE( 6, 'i', 'p', import);
302 CASE( 6, 'p', 'a', pragma);
304 CASE( 7, 'd', 'f', defined);
305 CASE( 7, 'i', 'c', include);
306 CASE( 7, 'w', 'r', warning);
308 CASE( 8, 'u', 'a', unassert);
309 CASE(12, 'i', 'c', include_next);
311 CASE(14, '_', 'p', __public_macro);
313 CASE(15, '_', 'p', __private_macro);
315 CASE(16, '_', 'i', __include_macros);
321 //===----------------------------------------------------------------------===//
322 // Stats Implementation
323 //===----------------------------------------------------------------------===//
325 /// PrintStats - Print statistics about how well the identifier table is doing
326 /// at hashing identifiers.
327 void IdentifierTable::PrintStats() const {
328 unsigned NumBuckets = HashTable.getNumBuckets();
329 unsigned NumIdentifiers = HashTable.getNumItems();
330 unsigned NumEmptyBuckets = NumBuckets-NumIdentifiers;
331 unsigned AverageIdentifierSize = 0;
332 unsigned MaxIdentifierLength = 0;
334 // TODO: Figure out maximum times an identifier had to probe for -stats.
335 for (llvm::StringMap<IdentifierInfo*, llvm::BumpPtrAllocator>::const_iterator
336 I = HashTable.begin(), E = HashTable.end(); I != E; ++I) {
337 unsigned IdLen = I->getKeyLength();
338 AverageIdentifierSize += IdLen;
339 if (MaxIdentifierLength < IdLen)
340 MaxIdentifierLength = IdLen;
343 fprintf(stderr, "\n*** Identifier Table Stats:\n");
344 fprintf(stderr, "# Identifiers: %d\n", NumIdentifiers);
345 fprintf(stderr, "# Empty Buckets: %d\n", NumEmptyBuckets);
346 fprintf(stderr, "Hash density (#identifiers per bucket): %f\n",
347 NumIdentifiers/(double)NumBuckets);
348 fprintf(stderr, "Ave identifier length: %f\n",
349 (AverageIdentifierSize/(double)NumIdentifiers));
350 fprintf(stderr, "Max identifier length: %d\n", MaxIdentifierLength);
352 // Compute statistics about the memory allocated for identifiers.
353 HashTable.getAllocator().PrintStats();
356 //===----------------------------------------------------------------------===//
357 // SelectorTable Implementation
358 //===----------------------------------------------------------------------===//
360 unsigned llvm::DenseMapInfo<clang::Selector>::getHashValue(clang::Selector S) {
361 return DenseMapInfo<void*>::getHashValue(S.getAsOpaquePtr());
365 /// MultiKeywordSelector - One of these variable length records is kept for each
366 /// selector containing more than one keyword. We use a folding set
367 /// to unique aggregate names (keyword selectors in ObjC parlance). Access to
368 /// this class is provided strictly through Selector.
369 class MultiKeywordSelector
370 : public DeclarationNameExtra, public llvm::FoldingSetNode {
371 MultiKeywordSelector(unsigned nKeys) {
372 ExtraKindOrNumArgs = NUM_EXTRA_KINDS + nKeys;
375 // Constructor for keyword selectors.
376 MultiKeywordSelector(unsigned nKeys, IdentifierInfo **IIV) {
377 assert((nKeys > 1) && "not a multi-keyword selector");
378 ExtraKindOrNumArgs = NUM_EXTRA_KINDS + nKeys;
380 // Fill in the trailing keyword array.
381 IdentifierInfo **KeyInfo = reinterpret_cast<IdentifierInfo **>(this+1);
382 for (unsigned i = 0; i != nKeys; ++i)
386 // getName - Derive the full selector name and return it.
387 std::string getName() const;
389 unsigned getNumArgs() const { return ExtraKindOrNumArgs - NUM_EXTRA_KINDS; }
391 typedef IdentifierInfo *const *keyword_iterator;
392 keyword_iterator keyword_begin() const {
393 return reinterpret_cast<keyword_iterator>(this+1);
395 keyword_iterator keyword_end() const {
396 return keyword_begin()+getNumArgs();
398 IdentifierInfo *getIdentifierInfoForSlot(unsigned i) const {
399 assert(i < getNumArgs() && "getIdentifierInfoForSlot(): illegal index");
400 return keyword_begin()[i];
402 static void Profile(llvm::FoldingSetNodeID &ID,
403 keyword_iterator ArgTys, unsigned NumArgs) {
404 ID.AddInteger(NumArgs);
405 for (unsigned i = 0; i != NumArgs; ++i)
406 ID.AddPointer(ArgTys[i]);
408 void Profile(llvm::FoldingSetNodeID &ID) {
409 Profile(ID, keyword_begin(), getNumArgs());
412 } // end namespace clang.
414 unsigned Selector::getNumArgs() const {
415 unsigned IIF = getIdentifierInfoFlag();
420 // We point to a MultiKeywordSelector.
421 MultiKeywordSelector *SI = getMultiKeywordSelector();
422 return SI->getNumArgs();
425 IdentifierInfo *Selector::getIdentifierInfoForSlot(unsigned argIndex) const {
426 if (getIdentifierInfoFlag() < MultiArg) {
427 assert(argIndex == 0 && "illegal keyword index");
428 return getAsIdentifierInfo();
430 // We point to a MultiKeywordSelector.
431 MultiKeywordSelector *SI = getMultiKeywordSelector();
432 return SI->getIdentifierInfoForSlot(argIndex);
435 StringRef Selector::getNameForSlot(unsigned int argIndex) const {
436 IdentifierInfo *II = getIdentifierInfoForSlot(argIndex);
437 return II? II->getName() : StringRef();
440 std::string MultiKeywordSelector::getName() const {
441 SmallString<256> Str;
442 llvm::raw_svector_ostream OS(Str);
443 for (keyword_iterator I = keyword_begin(), E = keyword_end(); I != E; ++I) {
445 OS << (*I)->getName();
452 std::string Selector::getAsString() const {
454 return "<null selector>";
456 if (getIdentifierInfoFlag() < MultiArg) {
457 IdentifierInfo *II = getAsIdentifierInfo();
459 if (getNumArgs() == 0) {
460 assert(II && "If the number of arguments is 0 then II is guaranteed to "
462 return II->getName();
468 return II->getName().str() + ":";
471 // We have a multiple keyword selector.
472 return getMultiKeywordSelector()->getName();
475 void Selector::print(llvm::raw_ostream &OS) const {
479 /// Interpreting the given string using the normal CamelCase
480 /// conventions, determine whether the given string starts with the
481 /// given "word", which is assumed to end in a lowercase letter.
482 static bool startsWithWord(StringRef name, StringRef word) {
483 if (name.size() < word.size()) return false;
484 return ((name.size() == word.size() || !isLowercase(name[word.size()])) &&
485 name.startswith(word));
488 ObjCMethodFamily Selector::getMethodFamilyImpl(Selector sel) {
489 IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
490 if (!first) return OMF_None;
492 StringRef name = first->getName();
493 if (sel.isUnarySelector()) {
494 if (name == "autorelease") return OMF_autorelease;
495 if (name == "dealloc") return OMF_dealloc;
496 if (name == "finalize") return OMF_finalize;
497 if (name == "release") return OMF_release;
498 if (name == "retain") return OMF_retain;
499 if (name == "retainCount") return OMF_retainCount;
500 if (name == "self") return OMF_self;
501 if (name == "initialize") return OMF_initialize;
504 if (name == "performSelector" || name == "performSelectorInBackground" ||
505 name == "performSelectorOnMainThread")
506 return OMF_performSelector;
508 // The other method families may begin with a prefix of underscores.
509 while (!name.empty() && name.front() == '_')
510 name = name.substr(1);
512 if (name.empty()) return OMF_None;
513 switch (name.front()) {
515 if (startsWithWord(name, "alloc")) return OMF_alloc;
518 if (startsWithWord(name, "copy")) return OMF_copy;
521 if (startsWithWord(name, "init")) return OMF_init;
524 if (startsWithWord(name, "mutableCopy")) return OMF_mutableCopy;
527 if (startsWithWord(name, "new")) return OMF_new;
536 ObjCInstanceTypeFamily Selector::getInstTypeMethodFamily(Selector sel) {
537 IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
538 if (!first) return OIT_None;
540 StringRef name = first->getName();
542 if (name.empty()) return OIT_None;
543 switch (name.front()) {
545 if (startsWithWord(name, "array")) return OIT_Array;
548 if (startsWithWord(name, "default")) return OIT_ReturnsSelf;
549 if (startsWithWord(name, "dictionary")) return OIT_Dictionary;
552 if (startsWithWord(name, "shared")) return OIT_ReturnsSelf;
553 if (startsWithWord(name, "standard")) return OIT_Singleton;
556 if (startsWithWord(name, "init")) return OIT_Init;
563 ObjCStringFormatFamily Selector::getStringFormatFamilyImpl(Selector sel) {
564 IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
565 if (!first) return SFF_None;
567 StringRef name = first->getName();
569 switch (name.front()) {
571 if (name == "appendFormat") return SFF_NSString;
575 if (name == "initWithFormat") return SFF_NSString;
579 if (name == "localizedStringWithFormat") return SFF_NSString;
583 if (name == "stringByAppendingFormat" ||
584 name == "stringWithFormat") return SFF_NSString;
591 struct SelectorTableImpl {
592 llvm::FoldingSet<MultiKeywordSelector> Table;
593 llvm::BumpPtrAllocator Allocator;
595 } // end anonymous namespace.
597 static SelectorTableImpl &getSelectorTableImpl(void *P) {
598 return *static_cast<SelectorTableImpl*>(P);
602 SelectorTable::constructSetterName(StringRef Name) {
603 SmallString<64> SetterName("set");
605 SetterName[3] = toUppercase(SetterName[3]);
610 SelectorTable::constructSetterSelector(IdentifierTable &Idents,
611 SelectorTable &SelTable,
612 const IdentifierInfo *Name) {
613 IdentifierInfo *SetterName =
614 &Idents.get(constructSetterName(Name->getName()));
615 return SelTable.getUnarySelector(SetterName);
618 size_t SelectorTable::getTotalMemory() const {
619 SelectorTableImpl &SelTabImpl = getSelectorTableImpl(Impl);
620 return SelTabImpl.Allocator.getTotalMemory();
623 Selector SelectorTable::getSelector(unsigned nKeys, IdentifierInfo **IIV) {
625 return Selector(IIV[0], nKeys);
627 SelectorTableImpl &SelTabImpl = getSelectorTableImpl(Impl);
629 // Unique selector, to guarantee there is one per name.
630 llvm::FoldingSetNodeID ID;
631 MultiKeywordSelector::Profile(ID, IIV, nKeys);
633 void *InsertPos = nullptr;
634 if (MultiKeywordSelector *SI =
635 SelTabImpl.Table.FindNodeOrInsertPos(ID, InsertPos))
638 // MultiKeywordSelector objects are not allocated with new because they have a
639 // variable size array (for parameter types) at the end of them.
640 unsigned Size = sizeof(MultiKeywordSelector) + nKeys*sizeof(IdentifierInfo *);
641 MultiKeywordSelector *SI =
642 (MultiKeywordSelector *)SelTabImpl.Allocator.Allocate(
643 Size, alignof(MultiKeywordSelector));
644 new (SI) MultiKeywordSelector(nKeys, IIV);
645 SelTabImpl.Table.InsertNode(SI, InsertPos);
649 SelectorTable::SelectorTable() {
650 Impl = new SelectorTableImpl();
653 SelectorTable::~SelectorTable() {
654 delete &getSelectorTableImpl(Impl);
657 const char *clang::getOperatorSpelling(OverloadedOperatorKind Operator) {
660 case NUM_OVERLOADED_OPERATORS:
663 #define OVERLOADED_OPERATOR(Name,Spelling,Token,Unary,Binary,MemberOnly) \
664 case OO_##Name: return Spelling;
665 #include "clang/Basic/OperatorKinds.def"
668 llvm_unreachable("Invalid OverloadedOperatorKind!");
671 StringRef clang::getNullabilitySpelling(NullabilityKind kind,
672 bool isContextSensitive) {
674 case NullabilityKind::NonNull:
675 return isContextSensitive ? "nonnull" : "_Nonnull";
677 case NullabilityKind::Nullable:
678 return isContextSensitive ? "nullable" : "_Nullable";
680 case NullabilityKind::Unspecified:
681 return isContextSensitive ? "null_unspecified" : "_Null_unspecified";
683 llvm_unreachable("Unknown nullability kind.");