1 //===- IdentifierTable.cpp - Hash table for identifier lookup -------------===//
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 IdentifierInfo, IdentifierVisitor, and
10 // IdentifierTable interfaces.
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "clang/Basic/IdentifierTable.h"
15 #include "clang/Basic/CharInfo.h"
16 #include "clang/Basic/LangOptions.h"
17 #include "clang/Basic/OperatorKinds.h"
18 #include "clang/Basic/Specifiers.h"
19 #include "clang/Basic/TargetBuiltins.h"
20 #include "clang/Basic/TokenKinds.h"
21 #include "llvm/ADT/DenseMapInfo.h"
22 #include "llvm/ADT/FoldingSet.h"
23 #include "llvm/ADT/SmallString.h"
24 #include "llvm/ADT/StringMap.h"
25 #include "llvm/ADT/StringRef.h"
26 #include "llvm/Support/Allocator.h"
27 #include "llvm/Support/ErrorHandling.h"
28 #include "llvm/Support/raw_ostream.h"
29 #include <cassert>
30 #include <cstdio>
31 #include <cstring>
32 #include <string>
33 
34 using namespace clang;
35 
36 // A check to make sure the ObjCOrBuiltinID has sufficient room to store the
37 // largest possible target/aux-target combination. If we exceed this, we likely
38 // need to just change the ObjCOrBuiltinIDBits value in IdentifierTable.h.
39 static_assert(2 * LargestBuiltinID < (2 << (ObjCOrBuiltinIDBits - 1)),
40               "Insufficient ObjCOrBuiltinID Bits");
41 
42 //===----------------------------------------------------------------------===//
43 // IdentifierTable Implementation
44 //===----------------------------------------------------------------------===//
45 
46 IdentifierIterator::~IdentifierIterator() = default;
47 
48 IdentifierInfoLookup::~IdentifierInfoLookup() = default;
49 
50 namespace {
51 
52 /// A simple identifier lookup iterator that represents an
53 /// empty sequence of identifiers.
54 class EmptyLookupIterator : public IdentifierIterator
55 {
56 public:
57   StringRef Next() override { return StringRef(); }
58 };
59 
60 } // namespace
61 
62 IdentifierIterator *IdentifierInfoLookup::getIdentifiers() {
63   return new EmptyLookupIterator();
64 }
65 
66 IdentifierTable::IdentifierTable(IdentifierInfoLookup *ExternalLookup)
67     : HashTable(8192), // Start with space for 8K identifiers.
68       ExternalLookup(ExternalLookup) {}
69 
70 IdentifierTable::IdentifierTable(const LangOptions &LangOpts,
71                                  IdentifierInfoLookup *ExternalLookup)
72     : IdentifierTable(ExternalLookup) {
73   // Populate the identifier table with info about keywords for the current
74   // language.
75   AddKeywords(LangOpts);
76 }
77 
78 //===----------------------------------------------------------------------===//
79 // Language Keyword Implementation
80 //===----------------------------------------------------------------------===//
81 
82 // Constants for TokenKinds.def
83 namespace {
84 
85   enum {
86     KEYC99        = 0x1,
87     KEYCXX        = 0x2,
88     KEYCXX11      = 0x4,
89     KEYGNU        = 0x8,
90     KEYMS         = 0x10,
91     BOOLSUPPORT   = 0x20,
92     KEYALTIVEC    = 0x40,
93     KEYNOCXX      = 0x80,
94     KEYBORLAND    = 0x100,
95     KEYOPENCLC    = 0x200,
96     KEYC11        = 0x400,
97     KEYNOMS18     = 0x800,
98     KEYNOOPENCL   = 0x1000,
99     WCHARSUPPORT  = 0x2000,
100     HALFSUPPORT   = 0x4000,
101     CHAR8SUPPORT  = 0x8000,
102     KEYCONCEPTS   = 0x10000,
103     KEYOBJC       = 0x20000,
104     KEYZVECTOR    = 0x40000,
105     KEYCOROUTINES = 0x80000,
106     KEYMODULES    = 0x100000,
107     KEYCXX20      = 0x200000,
108     KEYOPENCLCXX  = 0x400000,
109     KEYMSCOMPAT   = 0x800000,
110     KEYALLCXX = KEYCXX | KEYCXX11 | KEYCXX20,
111     KEYALL = (0xffffff & ~KEYNOMS18 &
112               ~KEYNOOPENCL) // KEYNOMS18 and KEYNOOPENCL are used to exclude.
113   };
114 
115   /// How a keyword is treated in the selected standard.
116   enum KeywordStatus {
117     KS_Disabled,    // Disabled
118     KS_Extension,   // Is an extension
119     KS_Enabled,     // Enabled
120     KS_Future       // Is a keyword in future standard
121   };
122 
123 } // namespace
124 
125 /// Translates flags as specified in TokenKinds.def into keyword status
126 /// in the given language standard.
127 static KeywordStatus getKeywordStatus(const LangOptions &LangOpts,
128                                       unsigned Flags) {
129   if (Flags == KEYALL) return KS_Enabled;
130   if (LangOpts.CPlusPlus && (Flags & KEYCXX)) return KS_Enabled;
131   if (LangOpts.CPlusPlus11 && (Flags & KEYCXX11)) return KS_Enabled;
132   if (LangOpts.CPlusPlus20 && (Flags & KEYCXX20)) return KS_Enabled;
133   if (LangOpts.C99 && (Flags & KEYC99)) return KS_Enabled;
134   if (LangOpts.GNUKeywords && (Flags & KEYGNU)) return KS_Extension;
135   if (LangOpts.MicrosoftExt && (Flags & KEYMS)) return KS_Extension;
136   if (LangOpts.MSVCCompat && (Flags & KEYMSCOMPAT)) return KS_Enabled;
137   if (LangOpts.Borland && (Flags & KEYBORLAND)) return KS_Extension;
138   if (LangOpts.Bool && (Flags & BOOLSUPPORT)) return KS_Enabled;
139   if (LangOpts.Half && (Flags & HALFSUPPORT)) return KS_Enabled;
140   if (LangOpts.WChar && (Flags & WCHARSUPPORT)) return KS_Enabled;
141   if (LangOpts.Char8 && (Flags & CHAR8SUPPORT)) return KS_Enabled;
142   if (LangOpts.AltiVec && (Flags & KEYALTIVEC)) return KS_Enabled;
143   if (LangOpts.ZVector && (Flags & KEYZVECTOR)) return KS_Enabled;
144   if (LangOpts.OpenCL && !LangOpts.OpenCLCPlusPlus && (Flags & KEYOPENCLC))
145     return KS_Enabled;
146   if (LangOpts.OpenCLCPlusPlus && (Flags & KEYOPENCLCXX)) return KS_Enabled;
147   if (!LangOpts.CPlusPlus && (Flags & KEYNOCXX)) return KS_Enabled;
148   if (LangOpts.C11 && (Flags & KEYC11)) return KS_Enabled;
149   // We treat bridge casts as objective-C keywords so we can warn on them
150   // in non-arc mode.
151   if (LangOpts.ObjC && (Flags & KEYOBJC)) return KS_Enabled;
152   if (LangOpts.CPlusPlus20 && (Flags & KEYCONCEPTS)) return KS_Enabled;
153   if (LangOpts.Coroutines && (Flags & KEYCOROUTINES)) return KS_Enabled;
154   if (LangOpts.ModulesTS && (Flags & KEYMODULES)) return KS_Enabled;
155   if (LangOpts.CPlusPlus && (Flags & KEYALLCXX)) return KS_Future;
156   if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus20 && (Flags & CHAR8SUPPORT))
157     return KS_Future;
158   return KS_Disabled;
159 }
160 
161 /// AddKeyword - This method is used to associate a token ID with specific
162 /// identifiers because they are language keywords.  This causes the lexer to
163 /// automatically map matching identifiers to specialized token codes.
164 static void AddKeyword(StringRef Keyword,
165                        tok::TokenKind TokenCode, unsigned Flags,
166                        const LangOptions &LangOpts, IdentifierTable &Table) {
167   KeywordStatus AddResult = getKeywordStatus(LangOpts, Flags);
168 
169   // Don't add this keyword under MSVCCompat.
170   if (LangOpts.MSVCCompat && (Flags & KEYNOMS18) &&
171       !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2015))
172     return;
173 
174   // Don't add this keyword under OpenCL.
175   if (LangOpts.OpenCL && (Flags & KEYNOOPENCL))
176     return;
177 
178   // Don't add this keyword if disabled in this language.
179   if (AddResult == KS_Disabled) return;
180 
181   IdentifierInfo &Info =
182       Table.get(Keyword, AddResult == KS_Future ? tok::identifier : TokenCode);
183   Info.setIsExtensionToken(AddResult == KS_Extension);
184   Info.setIsFutureCompatKeyword(AddResult == KS_Future);
185 }
186 
187 /// AddCXXOperatorKeyword - Register a C++ operator keyword alternative
188 /// representations.
189 static void AddCXXOperatorKeyword(StringRef Keyword,
190                                   tok::TokenKind TokenCode,
191                                   IdentifierTable &Table) {
192   IdentifierInfo &Info = Table.get(Keyword, TokenCode);
193   Info.setIsCPlusPlusOperatorKeyword();
194 }
195 
196 /// AddObjCKeyword - Register an Objective-C \@keyword like "class" "selector"
197 /// or "property".
198 static void AddObjCKeyword(StringRef Name,
199                            tok::ObjCKeywordKind ObjCID,
200                            IdentifierTable &Table) {
201   Table.get(Name).setObjCKeywordID(ObjCID);
202 }
203 
204 /// AddKeywords - Add all keywords to the symbol table.
205 ///
206 void IdentifierTable::AddKeywords(const LangOptions &LangOpts) {
207   // Add keywords and tokens for the current language.
208 #define KEYWORD(NAME, FLAGS) \
209   AddKeyword(StringRef(#NAME), tok::kw_ ## NAME,  \
210              FLAGS, LangOpts, *this);
211 #define ALIAS(NAME, TOK, FLAGS) \
212   AddKeyword(StringRef(NAME), tok::kw_ ## TOK,  \
213              FLAGS, LangOpts, *this);
214 #define CXX_KEYWORD_OPERATOR(NAME, ALIAS) \
215   if (LangOpts.CXXOperatorNames)          \
216     AddCXXOperatorKeyword(StringRef(#NAME), tok::ALIAS, *this);
217 #define OBJC_AT_KEYWORD(NAME)  \
218   if (LangOpts.ObjC)           \
219     AddObjCKeyword(StringRef(#NAME), tok::objc_##NAME, *this);
220 #define TESTING_KEYWORD(NAME, FLAGS)
221 #include "clang/Basic/TokenKinds.def"
222 
223   if (LangOpts.ParseUnknownAnytype)
224     AddKeyword("__unknown_anytype", tok::kw___unknown_anytype, KEYALL,
225                LangOpts, *this);
226 
227   if (LangOpts.DeclSpecKeyword)
228     AddKeyword("__declspec", tok::kw___declspec, KEYALL, LangOpts, *this);
229 
230   if (LangOpts.IEEE128)
231     AddKeyword("__ieee128", tok::kw___float128, KEYALL, LangOpts, *this);
232 
233   // Add the 'import' contextual keyword.
234   get("import").setModulesImport(true);
235 }
236 
237 /// Checks if the specified token kind represents a keyword in the
238 /// specified language.
239 /// \returns Status of the keyword in the language.
240 static KeywordStatus getTokenKwStatus(const LangOptions &LangOpts,
241                                       tok::TokenKind K) {
242   switch (K) {
243 #define KEYWORD(NAME, FLAGS) \
244   case tok::kw_##NAME: return getKeywordStatus(LangOpts, FLAGS);
245 #include "clang/Basic/TokenKinds.def"
246   default: return KS_Disabled;
247   }
248 }
249 
250 /// Returns true if the identifier represents a keyword in the
251 /// specified language.
252 bool IdentifierInfo::isKeyword(const LangOptions &LangOpts) const {
253   switch (getTokenKwStatus(LangOpts, getTokenID())) {
254   case KS_Enabled:
255   case KS_Extension:
256     return true;
257   default:
258     return false;
259   }
260 }
261 
262 /// Returns true if the identifier represents a C++ keyword in the
263 /// specified language.
264 bool IdentifierInfo::isCPlusPlusKeyword(const LangOptions &LangOpts) const {
265   if (!LangOpts.CPlusPlus || !isKeyword(LangOpts))
266     return false;
267   // This is a C++ keyword if this identifier is not a keyword when checked
268   // using LangOptions without C++ support.
269   LangOptions LangOptsNoCPP = LangOpts;
270   LangOptsNoCPP.CPlusPlus = false;
271   LangOptsNoCPP.CPlusPlus11 = false;
272   LangOptsNoCPP.CPlusPlus20 = false;
273   return !isKeyword(LangOptsNoCPP);
274 }
275 
276 ReservedIdentifierStatus
277 IdentifierInfo::isReserved(const LangOptions &LangOpts) const {
278   StringRef Name = getName();
279 
280   // '_' is a reserved identifier, but its use is so common (e.g. to store
281   // ignored values) that we don't warn on it.
282   if (Name.size() <= 1)
283     return ReservedIdentifierStatus::NotReserved;
284 
285   // [lex.name] p3
286   if (Name[0] == '_') {
287 
288     // Each name that begins with an underscore followed by an uppercase letter
289     // or another underscore is reserved.
290     if (Name[1] == '_')
291       return ReservedIdentifierStatus::StartsWithDoubleUnderscore;
292 
293     if ('A' <= Name[1] && Name[1] <= 'Z')
294       return ReservedIdentifierStatus::
295           StartsWithUnderscoreFollowedByCapitalLetter;
296 
297     // This is a bit misleading: it actually means it's only reserved if we're
298     // at global scope because it starts with an underscore.
299     return ReservedIdentifierStatus::StartsWithUnderscoreAtGlobalScope;
300   }
301 
302   // Each name that contains a double underscore (__) is reserved.
303   if (LangOpts.CPlusPlus && Name.contains("__"))
304     return ReservedIdentifierStatus::ContainsDoubleUnderscore;
305 
306   return ReservedIdentifierStatus::NotReserved;
307 }
308 
309 tok::PPKeywordKind IdentifierInfo::getPPKeywordID() const {
310   // We use a perfect hash function here involving the length of the keyword,
311   // the first and third character.  For preprocessor ID's there are no
312   // collisions (if there were, the switch below would complain about duplicate
313   // case values).  Note that this depends on 'if' being null terminated.
314 
315 #define HASH(LEN, FIRST, THIRD) \
316   (LEN << 5) + (((FIRST-'a') + (THIRD-'a')) & 31)
317 #define CASE(LEN, FIRST, THIRD, NAME) \
318   case HASH(LEN, FIRST, THIRD): \
319     return memcmp(Name, #NAME, LEN) ? tok::pp_not_keyword : tok::pp_ ## NAME
320 
321   unsigned Len = getLength();
322   if (Len < 2) return tok::pp_not_keyword;
323   const char *Name = getNameStart();
324   switch (HASH(Len, Name[0], Name[2])) {
325   default: return tok::pp_not_keyword;
326   CASE( 2, 'i', '\0', if);
327   CASE( 4, 'e', 'i', elif);
328   CASE( 4, 'e', 's', else);
329   CASE( 4, 'l', 'n', line);
330   CASE( 4, 's', 'c', sccs);
331   CASE( 5, 'e', 'd', endif);
332   CASE( 5, 'e', 'r', error);
333   CASE( 5, 'i', 'e', ident);
334   CASE( 5, 'i', 'd', ifdef);
335   CASE( 5, 'u', 'd', undef);
336 
337   CASE( 6, 'a', 's', assert);
338   CASE( 6, 'd', 'f', define);
339   CASE( 6, 'i', 'n', ifndef);
340   CASE( 6, 'i', 'p', import);
341   CASE( 6, 'p', 'a', pragma);
342 
343   CASE( 7, 'd', 'f', defined);
344   CASE( 7, 'e', 'i', elifdef);
345   CASE( 7, 'i', 'c', include);
346   CASE( 7, 'w', 'r', warning);
347 
348   CASE( 8, 'e', 'i', elifndef);
349   CASE( 8, 'u', 'a', unassert);
350   CASE(12, 'i', 'c', include_next);
351 
352   CASE(14, '_', 'p', __public_macro);
353 
354   CASE(15, '_', 'p', __private_macro);
355 
356   CASE(16, '_', 'i', __include_macros);
357 #undef CASE
358 #undef HASH
359   }
360 }
361 
362 //===----------------------------------------------------------------------===//
363 // Stats Implementation
364 //===----------------------------------------------------------------------===//
365 
366 /// PrintStats - Print statistics about how well the identifier table is doing
367 /// at hashing identifiers.
368 void IdentifierTable::PrintStats() const {
369   unsigned NumBuckets = HashTable.getNumBuckets();
370   unsigned NumIdentifiers = HashTable.getNumItems();
371   unsigned NumEmptyBuckets = NumBuckets-NumIdentifiers;
372   unsigned AverageIdentifierSize = 0;
373   unsigned MaxIdentifierLength = 0;
374 
375   // TODO: Figure out maximum times an identifier had to probe for -stats.
376   for (llvm::StringMap<IdentifierInfo*, llvm::BumpPtrAllocator>::const_iterator
377        I = HashTable.begin(), E = HashTable.end(); I != E; ++I) {
378     unsigned IdLen = I->getKeyLength();
379     AverageIdentifierSize += IdLen;
380     if (MaxIdentifierLength < IdLen)
381       MaxIdentifierLength = IdLen;
382   }
383 
384   fprintf(stderr, "\n*** Identifier Table Stats:\n");
385   fprintf(stderr, "# Identifiers:   %d\n", NumIdentifiers);
386   fprintf(stderr, "# Empty Buckets: %d\n", NumEmptyBuckets);
387   fprintf(stderr, "Hash density (#identifiers per bucket): %f\n",
388           NumIdentifiers/(double)NumBuckets);
389   fprintf(stderr, "Ave identifier length: %f\n",
390           (AverageIdentifierSize/(double)NumIdentifiers));
391   fprintf(stderr, "Max identifier length: %d\n", MaxIdentifierLength);
392 
393   // Compute statistics about the memory allocated for identifiers.
394   HashTable.getAllocator().PrintStats();
395 }
396 
397 //===----------------------------------------------------------------------===//
398 // SelectorTable Implementation
399 //===----------------------------------------------------------------------===//
400 
401 unsigned llvm::DenseMapInfo<clang::Selector>::getHashValue(clang::Selector S) {
402   return DenseMapInfo<void*>::getHashValue(S.getAsOpaquePtr());
403 }
404 
405 namespace clang {
406 
407 /// One of these variable length records is kept for each
408 /// selector containing more than one keyword. We use a folding set
409 /// to unique aggregate names (keyword selectors in ObjC parlance). Access to
410 /// this class is provided strictly through Selector.
411 class alignas(IdentifierInfoAlignment) MultiKeywordSelector
412     : public detail::DeclarationNameExtra,
413       public llvm::FoldingSetNode {
414   MultiKeywordSelector(unsigned nKeys) : DeclarationNameExtra(nKeys) {}
415 
416 public:
417   // Constructor for keyword selectors.
418   MultiKeywordSelector(unsigned nKeys, IdentifierInfo **IIV)
419       : DeclarationNameExtra(nKeys) {
420     assert((nKeys > 1) && "not a multi-keyword selector");
421 
422     // Fill in the trailing keyword array.
423     IdentifierInfo **KeyInfo = reinterpret_cast<IdentifierInfo **>(this + 1);
424     for (unsigned i = 0; i != nKeys; ++i)
425       KeyInfo[i] = IIV[i];
426   }
427 
428   // getName - Derive the full selector name and return it.
429   std::string getName() const;
430 
431   using DeclarationNameExtra::getNumArgs;
432 
433   using keyword_iterator = IdentifierInfo *const *;
434 
435   keyword_iterator keyword_begin() const {
436     return reinterpret_cast<keyword_iterator>(this + 1);
437   }
438 
439   keyword_iterator keyword_end() const {
440     return keyword_begin() + getNumArgs();
441   }
442 
443   IdentifierInfo *getIdentifierInfoForSlot(unsigned i) const {
444     assert(i < getNumArgs() && "getIdentifierInfoForSlot(): illegal index");
445     return keyword_begin()[i];
446   }
447 
448   static void Profile(llvm::FoldingSetNodeID &ID, keyword_iterator ArgTys,
449                       unsigned NumArgs) {
450     ID.AddInteger(NumArgs);
451     for (unsigned i = 0; i != NumArgs; ++i)
452       ID.AddPointer(ArgTys[i]);
453   }
454 
455   void Profile(llvm::FoldingSetNodeID &ID) {
456     Profile(ID, keyword_begin(), getNumArgs());
457   }
458 };
459 
460 } // namespace clang.
461 
462 bool Selector::isKeywordSelector(ArrayRef<StringRef> Names) const {
463   assert(!Names.empty() && "must have >= 1 selector slots");
464   if (getNumArgs() != Names.size())
465     return false;
466   for (unsigned I = 0, E = Names.size(); I != E; ++I) {
467     if (getNameForSlot(I) != Names[I])
468       return false;
469   }
470   return true;
471 }
472 
473 bool Selector::isUnarySelector(StringRef Name) const {
474   return isUnarySelector() && getNameForSlot(0) == Name;
475 }
476 
477 unsigned Selector::getNumArgs() const {
478   unsigned IIF = getIdentifierInfoFlag();
479   if (IIF <= ZeroArg)
480     return 0;
481   if (IIF == OneArg)
482     return 1;
483   // We point to a MultiKeywordSelector.
484   MultiKeywordSelector *SI = getMultiKeywordSelector();
485   return SI->getNumArgs();
486 }
487 
488 IdentifierInfo *Selector::getIdentifierInfoForSlot(unsigned argIndex) const {
489   if (getIdentifierInfoFlag() < MultiArg) {
490     assert(argIndex == 0 && "illegal keyword index");
491     return getAsIdentifierInfo();
492   }
493 
494   // We point to a MultiKeywordSelector.
495   MultiKeywordSelector *SI = getMultiKeywordSelector();
496   return SI->getIdentifierInfoForSlot(argIndex);
497 }
498 
499 StringRef Selector::getNameForSlot(unsigned int argIndex) const {
500   IdentifierInfo *II = getIdentifierInfoForSlot(argIndex);
501   return II ? II->getName() : StringRef();
502 }
503 
504 std::string MultiKeywordSelector::getName() const {
505   SmallString<256> Str;
506   llvm::raw_svector_ostream OS(Str);
507   for (keyword_iterator I = keyword_begin(), E = keyword_end(); I != E; ++I) {
508     if (*I)
509       OS << (*I)->getName();
510     OS << ':';
511   }
512 
513   return std::string(OS.str());
514 }
515 
516 std::string Selector::getAsString() const {
517   if (InfoPtr == 0)
518     return "<null selector>";
519 
520   if (getIdentifierInfoFlag() < MultiArg) {
521     IdentifierInfo *II = getAsIdentifierInfo();
522 
523     if (getNumArgs() == 0) {
524       assert(II && "If the number of arguments is 0 then II is guaranteed to "
525                    "not be null.");
526       return std::string(II->getName());
527     }
528 
529     if (!II)
530       return ":";
531 
532     return II->getName().str() + ":";
533   }
534 
535   // We have a multiple keyword selector.
536   return getMultiKeywordSelector()->getName();
537 }
538 
539 void Selector::print(llvm::raw_ostream &OS) const {
540   OS << getAsString();
541 }
542 
543 LLVM_DUMP_METHOD void Selector::dump() const { print(llvm::errs()); }
544 
545 /// Interpreting the given string using the normal CamelCase
546 /// conventions, determine whether the given string starts with the
547 /// given "word", which is assumed to end in a lowercase letter.
548 static bool startsWithWord(StringRef name, StringRef word) {
549   if (name.size() < word.size()) return false;
550   return ((name.size() == word.size() || !isLowercase(name[word.size()])) &&
551           name.startswith(word));
552 }
553 
554 ObjCMethodFamily Selector::getMethodFamilyImpl(Selector sel) {
555   IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
556   if (!first) return OMF_None;
557 
558   StringRef name = first->getName();
559   if (sel.isUnarySelector()) {
560     if (name == "autorelease") return OMF_autorelease;
561     if (name == "dealloc") return OMF_dealloc;
562     if (name == "finalize") return OMF_finalize;
563     if (name == "release") return OMF_release;
564     if (name == "retain") return OMF_retain;
565     if (name == "retainCount") return OMF_retainCount;
566     if (name == "self") return OMF_self;
567     if (name == "initialize") return OMF_initialize;
568   }
569 
570   if (name == "performSelector" || name == "performSelectorInBackground" ||
571       name == "performSelectorOnMainThread")
572     return OMF_performSelector;
573 
574   // The other method families may begin with a prefix of underscores.
575   while (!name.empty() && name.front() == '_')
576     name = name.substr(1);
577 
578   if (name.empty()) return OMF_None;
579   switch (name.front()) {
580   case 'a':
581     if (startsWithWord(name, "alloc")) return OMF_alloc;
582     break;
583   case 'c':
584     if (startsWithWord(name, "copy")) return OMF_copy;
585     break;
586   case 'i':
587     if (startsWithWord(name, "init")) return OMF_init;
588     break;
589   case 'm':
590     if (startsWithWord(name, "mutableCopy")) return OMF_mutableCopy;
591     break;
592   case 'n':
593     if (startsWithWord(name, "new")) return OMF_new;
594     break;
595   default:
596     break;
597   }
598 
599   return OMF_None;
600 }
601 
602 ObjCInstanceTypeFamily Selector::getInstTypeMethodFamily(Selector sel) {
603   IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
604   if (!first) return OIT_None;
605 
606   StringRef name = first->getName();
607 
608   if (name.empty()) return OIT_None;
609   switch (name.front()) {
610     case 'a':
611       if (startsWithWord(name, "array")) return OIT_Array;
612       break;
613     case 'd':
614       if (startsWithWord(name, "default")) return OIT_ReturnsSelf;
615       if (startsWithWord(name, "dictionary")) return OIT_Dictionary;
616       break;
617     case 's':
618       if (startsWithWord(name, "shared")) return OIT_ReturnsSelf;
619       if (startsWithWord(name, "standard")) return OIT_Singleton;
620       break;
621     case 'i':
622       if (startsWithWord(name, "init")) return OIT_Init;
623       break;
624     default:
625       break;
626   }
627   return OIT_None;
628 }
629 
630 ObjCStringFormatFamily Selector::getStringFormatFamilyImpl(Selector sel) {
631   IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
632   if (!first) return SFF_None;
633 
634   StringRef name = first->getName();
635 
636   switch (name.front()) {
637     case 'a':
638       if (name == "appendFormat") return SFF_NSString;
639       break;
640 
641     case 'i':
642       if (name == "initWithFormat") return SFF_NSString;
643       break;
644 
645     case 'l':
646       if (name == "localizedStringWithFormat") return SFF_NSString;
647       break;
648 
649     case 's':
650       if (name == "stringByAppendingFormat" ||
651           name == "stringWithFormat") return SFF_NSString;
652       break;
653   }
654   return SFF_None;
655 }
656 
657 namespace {
658 
659 struct SelectorTableImpl {
660   llvm::FoldingSet<MultiKeywordSelector> Table;
661   llvm::BumpPtrAllocator Allocator;
662 };
663 
664 } // namespace
665 
666 static SelectorTableImpl &getSelectorTableImpl(void *P) {
667   return *static_cast<SelectorTableImpl*>(P);
668 }
669 
670 SmallString<64>
671 SelectorTable::constructSetterName(StringRef Name) {
672   SmallString<64> SetterName("set");
673   SetterName += Name;
674   SetterName[3] = toUppercase(SetterName[3]);
675   return SetterName;
676 }
677 
678 Selector
679 SelectorTable::constructSetterSelector(IdentifierTable &Idents,
680                                        SelectorTable &SelTable,
681                                        const IdentifierInfo *Name) {
682   IdentifierInfo *SetterName =
683     &Idents.get(constructSetterName(Name->getName()));
684   return SelTable.getUnarySelector(SetterName);
685 }
686 
687 std::string SelectorTable::getPropertyNameFromSetterSelector(Selector Sel) {
688   StringRef Name = Sel.getNameForSlot(0);
689   assert(Name.startswith("set") && "invalid setter name");
690   return (Twine(toLowercase(Name[3])) + Name.drop_front(4)).str();
691 }
692 
693 size_t SelectorTable::getTotalMemory() const {
694   SelectorTableImpl &SelTabImpl = getSelectorTableImpl(Impl);
695   return SelTabImpl.Allocator.getTotalMemory();
696 }
697 
698 Selector SelectorTable::getSelector(unsigned nKeys, IdentifierInfo **IIV) {
699   if (nKeys < 2)
700     return Selector(IIV[0], nKeys);
701 
702   SelectorTableImpl &SelTabImpl = getSelectorTableImpl(Impl);
703 
704   // Unique selector, to guarantee there is one per name.
705   llvm::FoldingSetNodeID ID;
706   MultiKeywordSelector::Profile(ID, IIV, nKeys);
707 
708   void *InsertPos = nullptr;
709   if (MultiKeywordSelector *SI =
710         SelTabImpl.Table.FindNodeOrInsertPos(ID, InsertPos))
711     return Selector(SI);
712 
713   // MultiKeywordSelector objects are not allocated with new because they have a
714   // variable size array (for parameter types) at the end of them.
715   unsigned Size = sizeof(MultiKeywordSelector) + nKeys*sizeof(IdentifierInfo *);
716   MultiKeywordSelector *SI =
717       (MultiKeywordSelector *)SelTabImpl.Allocator.Allocate(
718           Size, alignof(MultiKeywordSelector));
719   new (SI) MultiKeywordSelector(nKeys, IIV);
720   SelTabImpl.Table.InsertNode(SI, InsertPos);
721   return Selector(SI);
722 }
723 
724 SelectorTable::SelectorTable() {
725   Impl = new SelectorTableImpl();
726 }
727 
728 SelectorTable::~SelectorTable() {
729   delete &getSelectorTableImpl(Impl);
730 }
731 
732 const char *clang::getOperatorSpelling(OverloadedOperatorKind Operator) {
733   switch (Operator) {
734   case OO_None:
735   case NUM_OVERLOADED_OPERATORS:
736     return nullptr;
737 
738 #define OVERLOADED_OPERATOR(Name,Spelling,Token,Unary,Binary,MemberOnly) \
739   case OO_##Name: return Spelling;
740 #include "clang/Basic/OperatorKinds.def"
741   }
742 
743   llvm_unreachable("Invalid OverloadedOperatorKind!");
744 }
745 
746 StringRef clang::getNullabilitySpelling(NullabilityKind kind,
747                                         bool isContextSensitive) {
748   switch (kind) {
749   case NullabilityKind::NonNull:
750     return isContextSensitive ? "nonnull" : "_Nonnull";
751 
752   case NullabilityKind::Nullable:
753     return isContextSensitive ? "nullable" : "_Nullable";
754 
755   case NullabilityKind::NullableResult:
756     assert(!isContextSensitive &&
757            "_Nullable_result isn't supported as context-sensitive keyword");
758     return "_Nullable_result";
759 
760   case NullabilityKind::Unspecified:
761     return isContextSensitive ? "null_unspecified" : "_Null_unspecified";
762   }
763   llvm_unreachable("Unknown nullability kind.");
764 }
765