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     KEYSYCL       = 0x1000000,
111     KEYCUDA       = 0x2000000,
112     KEYALLCXX = KEYCXX | KEYCXX11 | KEYCXX20,
113     KEYALL = (0x1ffffff & ~KEYNOMS18 &
114               ~KEYNOOPENCL) // KEYNOMS18 and KEYNOOPENCL are used to exclude.
115   };
116 
117   /// How a keyword is treated in the selected standard.
118   enum KeywordStatus {
119     KS_Disabled,    // Disabled
120     KS_Extension,   // Is an extension
121     KS_Enabled,     // Enabled
122     KS_Future       // Is a keyword in future standard
123   };
124 
125 } // namespace
126 
127 /// Translates flags as specified in TokenKinds.def into keyword status
128 /// in the given language standard.
129 static KeywordStatus getKeywordStatus(const LangOptions &LangOpts,
130                                       unsigned Flags) {
131   if (Flags == KEYALL) return KS_Enabled;
132   if (LangOpts.CPlusPlus && (Flags & KEYCXX)) return KS_Enabled;
133   if (LangOpts.CPlusPlus11 && (Flags & KEYCXX11)) return KS_Enabled;
134   if (LangOpts.CPlusPlus20 && (Flags & KEYCXX20)) return KS_Enabled;
135   if (LangOpts.C99 && (Flags & KEYC99)) return KS_Enabled;
136   if (LangOpts.GNUKeywords && (Flags & KEYGNU)) return KS_Extension;
137   if (LangOpts.MicrosoftExt && (Flags & KEYMS)) return KS_Extension;
138   if (LangOpts.MSVCCompat && (Flags & KEYMSCOMPAT)) return KS_Enabled;
139   if (LangOpts.Borland && (Flags & KEYBORLAND)) return KS_Extension;
140   if (LangOpts.Bool && (Flags & BOOLSUPPORT)) return KS_Enabled;
141   if (LangOpts.Half && (Flags & HALFSUPPORT)) return KS_Enabled;
142   if (LangOpts.WChar && (Flags & WCHARSUPPORT)) return KS_Enabled;
143   if (LangOpts.Char8 && (Flags & CHAR8SUPPORT)) return KS_Enabled;
144   if (LangOpts.AltiVec && (Flags & KEYALTIVEC)) return KS_Enabled;
145   if (LangOpts.ZVector && (Flags & KEYZVECTOR)) return KS_Enabled;
146   if (LangOpts.OpenCL && !LangOpts.OpenCLCPlusPlus && (Flags & KEYOPENCLC))
147     return KS_Enabled;
148   if (LangOpts.OpenCLCPlusPlus && (Flags & KEYOPENCLCXX)) return KS_Enabled;
149   if (!LangOpts.CPlusPlus && (Flags & KEYNOCXX)) return KS_Enabled;
150   if (LangOpts.C11 && (Flags & KEYC11)) return KS_Enabled;
151   // We treat bridge casts as objective-C keywords so we can warn on them
152   // in non-arc mode.
153   if (LangOpts.ObjC && (Flags & KEYOBJC)) return KS_Enabled;
154   if (LangOpts.CPlusPlus20 && (Flags & KEYCONCEPTS)) return KS_Enabled;
155   if (LangOpts.Coroutines && (Flags & KEYCOROUTINES)) return KS_Enabled;
156   if (LangOpts.ModulesTS && (Flags & KEYMODULES)) return KS_Enabled;
157   if (LangOpts.CPlusPlus && (Flags & KEYALLCXX)) return KS_Future;
158   if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus20 && (Flags & CHAR8SUPPORT))
159     return KS_Future;
160   if (LangOpts.isSYCL() && (Flags & KEYSYCL))
161     return KS_Enabled;
162   if (LangOpts.CUDA && (Flags & KEYCUDA))
163     return KS_Enabled;
164   return KS_Disabled;
165 }
166 
167 /// AddKeyword - This method is used to associate a token ID with specific
168 /// identifiers because they are language keywords.  This causes the lexer to
169 /// automatically map matching identifiers to specialized token codes.
170 static void AddKeyword(StringRef Keyword,
171                        tok::TokenKind TokenCode, unsigned Flags,
172                        const LangOptions &LangOpts, IdentifierTable &Table) {
173   KeywordStatus AddResult = getKeywordStatus(LangOpts, Flags);
174 
175   // Don't add this keyword under MSVCCompat.
176   if (LangOpts.MSVCCompat && (Flags & KEYNOMS18) &&
177       !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2015))
178     return;
179 
180   // Don't add this keyword under OpenCL.
181   if (LangOpts.OpenCL && (Flags & KEYNOOPENCL))
182     return;
183 
184   // Don't add this keyword if disabled in this language.
185   if (AddResult == KS_Disabled) return;
186 
187   IdentifierInfo &Info =
188       Table.get(Keyword, AddResult == KS_Future ? tok::identifier : TokenCode);
189   Info.setIsExtensionToken(AddResult == KS_Extension);
190   Info.setIsFutureCompatKeyword(AddResult == KS_Future);
191 }
192 
193 /// AddCXXOperatorKeyword - Register a C++ operator keyword alternative
194 /// representations.
195 static void AddCXXOperatorKeyword(StringRef Keyword,
196                                   tok::TokenKind TokenCode,
197                                   IdentifierTable &Table) {
198   IdentifierInfo &Info = Table.get(Keyword, TokenCode);
199   Info.setIsCPlusPlusOperatorKeyword();
200 }
201 
202 /// AddObjCKeyword - Register an Objective-C \@keyword like "class" "selector"
203 /// or "property".
204 static void AddObjCKeyword(StringRef Name,
205                            tok::ObjCKeywordKind ObjCID,
206                            IdentifierTable &Table) {
207   Table.get(Name).setObjCKeywordID(ObjCID);
208 }
209 
210 /// AddKeywords - Add all keywords to the symbol table.
211 ///
212 void IdentifierTable::AddKeywords(const LangOptions &LangOpts) {
213   // Add keywords and tokens for the current language.
214 #define KEYWORD(NAME, FLAGS) \
215   AddKeyword(StringRef(#NAME), tok::kw_ ## NAME,  \
216              FLAGS, LangOpts, *this);
217 #define ALIAS(NAME, TOK, FLAGS) \
218   AddKeyword(StringRef(NAME), tok::kw_ ## TOK,  \
219              FLAGS, LangOpts, *this);
220 #define CXX_KEYWORD_OPERATOR(NAME, ALIAS) \
221   if (LangOpts.CXXOperatorNames)          \
222     AddCXXOperatorKeyword(StringRef(#NAME), tok::ALIAS, *this);
223 #define OBJC_AT_KEYWORD(NAME)  \
224   if (LangOpts.ObjC)           \
225     AddObjCKeyword(StringRef(#NAME), tok::objc_##NAME, *this);
226 #define TESTING_KEYWORD(NAME, FLAGS)
227 #include "clang/Basic/TokenKinds.def"
228 
229   if (LangOpts.ParseUnknownAnytype)
230     AddKeyword("__unknown_anytype", tok::kw___unknown_anytype, KEYALL,
231                LangOpts, *this);
232 
233   if (LangOpts.DeclSpecKeyword)
234     AddKeyword("__declspec", tok::kw___declspec, KEYALL, LangOpts, *this);
235 
236   if (LangOpts.IEEE128)
237     AddKeyword("__ieee128", tok::kw___float128, KEYALL, LangOpts, *this);
238 
239   // Add the 'import' contextual keyword.
240   get("import").setModulesImport(true);
241 }
242 
243 /// Checks if the specified token kind represents a keyword in the
244 /// specified language.
245 /// \returns Status of the keyword in the language.
246 static KeywordStatus getTokenKwStatus(const LangOptions &LangOpts,
247                                       tok::TokenKind K) {
248   switch (K) {
249 #define KEYWORD(NAME, FLAGS) \
250   case tok::kw_##NAME: return getKeywordStatus(LangOpts, FLAGS);
251 #include "clang/Basic/TokenKinds.def"
252   default: return KS_Disabled;
253   }
254 }
255 
256 /// Returns true if the identifier represents a keyword in the
257 /// specified language.
258 bool IdentifierInfo::isKeyword(const LangOptions &LangOpts) const {
259   switch (getTokenKwStatus(LangOpts, getTokenID())) {
260   case KS_Enabled:
261   case KS_Extension:
262     return true;
263   default:
264     return false;
265   }
266 }
267 
268 /// Returns true if the identifier represents a C++ keyword in the
269 /// specified language.
270 bool IdentifierInfo::isCPlusPlusKeyword(const LangOptions &LangOpts) const {
271   if (!LangOpts.CPlusPlus || !isKeyword(LangOpts))
272     return false;
273   // This is a C++ keyword if this identifier is not a keyword when checked
274   // using LangOptions without C++ support.
275   LangOptions LangOptsNoCPP = LangOpts;
276   LangOptsNoCPP.CPlusPlus = false;
277   LangOptsNoCPP.CPlusPlus11 = false;
278   LangOptsNoCPP.CPlusPlus20 = false;
279   return !isKeyword(LangOptsNoCPP);
280 }
281 
282 ReservedIdentifierStatus
283 IdentifierInfo::isReserved(const LangOptions &LangOpts) const {
284   StringRef Name = getName();
285 
286   // '_' is a reserved identifier, but its use is so common (e.g. to store
287   // ignored values) that we don't warn on it.
288   if (Name.size() <= 1)
289     return ReservedIdentifierStatus::NotReserved;
290 
291   // [lex.name] p3
292   if (Name[0] == '_') {
293 
294     // Each name that begins with an underscore followed by an uppercase letter
295     // or another underscore is reserved.
296     if (Name[1] == '_')
297       return ReservedIdentifierStatus::StartsWithDoubleUnderscore;
298 
299     if ('A' <= Name[1] && Name[1] <= 'Z')
300       return ReservedIdentifierStatus::
301           StartsWithUnderscoreFollowedByCapitalLetter;
302 
303     // This is a bit misleading: it actually means it's only reserved if we're
304     // at global scope because it starts with an underscore.
305     return ReservedIdentifierStatus::StartsWithUnderscoreAtGlobalScope;
306   }
307 
308   // Each name that contains a double underscore (__) is reserved.
309   if (LangOpts.CPlusPlus && Name.contains("__"))
310     return ReservedIdentifierStatus::ContainsDoubleUnderscore;
311 
312   return ReservedIdentifierStatus::NotReserved;
313 }
314 
315 StringRef IdentifierInfo::deuglifiedName() const {
316   StringRef Name = getName();
317   if (Name.size() >= 2 && Name.front() == '_' &&
318       (Name[1] == '_' || (Name[1] >= 'A' && Name[1] <= 'Z')))
319     return Name.ltrim('_');
320   return Name;
321 }
322 
323 tok::PPKeywordKind IdentifierInfo::getPPKeywordID() const {
324   // We use a perfect hash function here involving the length of the keyword,
325   // the first and third character.  For preprocessor ID's there are no
326   // collisions (if there were, the switch below would complain about duplicate
327   // case values).  Note that this depends on 'if' being null terminated.
328 
329 #define HASH(LEN, FIRST, THIRD) \
330   (LEN << 5) + (((FIRST-'a') + (THIRD-'a')) & 31)
331 #define CASE(LEN, FIRST, THIRD, NAME) \
332   case HASH(LEN, FIRST, THIRD): \
333     return memcmp(Name, #NAME, LEN) ? tok::pp_not_keyword : tok::pp_ ## NAME
334 
335   unsigned Len = getLength();
336   if (Len < 2) return tok::pp_not_keyword;
337   const char *Name = getNameStart();
338   switch (HASH(Len, Name[0], Name[2])) {
339   default: return tok::pp_not_keyword;
340   CASE( 2, 'i', '\0', if);
341   CASE( 4, 'e', 'i', elif);
342   CASE( 4, 'e', 's', else);
343   CASE( 4, 'l', 'n', line);
344   CASE( 4, 's', 'c', sccs);
345   CASE( 5, 'e', 'd', endif);
346   CASE( 5, 'e', 'r', error);
347   CASE( 5, 'i', 'e', ident);
348   CASE( 5, 'i', 'd', ifdef);
349   CASE( 5, 'u', 'd', undef);
350 
351   CASE( 6, 'a', 's', assert);
352   CASE( 6, 'd', 'f', define);
353   CASE( 6, 'i', 'n', ifndef);
354   CASE( 6, 'i', 'p', import);
355   CASE( 6, 'p', 'a', pragma);
356 
357   CASE( 7, 'd', 'f', defined);
358   CASE( 7, 'e', 'i', elifdef);
359   CASE( 7, 'i', 'c', include);
360   CASE( 7, 'w', 'r', warning);
361 
362   CASE( 8, 'e', 'i', elifndef);
363   CASE( 8, 'u', 'a', unassert);
364   CASE(12, 'i', 'c', include_next);
365 
366   CASE(14, '_', 'p', __public_macro);
367 
368   CASE(15, '_', 'p', __private_macro);
369 
370   CASE(16, '_', 'i', __include_macros);
371 #undef CASE
372 #undef HASH
373   }
374 }
375 
376 //===----------------------------------------------------------------------===//
377 // Stats Implementation
378 //===----------------------------------------------------------------------===//
379 
380 /// PrintStats - Print statistics about how well the identifier table is doing
381 /// at hashing identifiers.
382 void IdentifierTable::PrintStats() const {
383   unsigned NumBuckets = HashTable.getNumBuckets();
384   unsigned NumIdentifiers = HashTable.getNumItems();
385   unsigned NumEmptyBuckets = NumBuckets-NumIdentifiers;
386   unsigned AverageIdentifierSize = 0;
387   unsigned MaxIdentifierLength = 0;
388 
389   // TODO: Figure out maximum times an identifier had to probe for -stats.
390   for (llvm::StringMap<IdentifierInfo*, llvm::BumpPtrAllocator>::const_iterator
391        I = HashTable.begin(), E = HashTable.end(); I != E; ++I) {
392     unsigned IdLen = I->getKeyLength();
393     AverageIdentifierSize += IdLen;
394     if (MaxIdentifierLength < IdLen)
395       MaxIdentifierLength = IdLen;
396   }
397 
398   fprintf(stderr, "\n*** Identifier Table Stats:\n");
399   fprintf(stderr, "# Identifiers:   %d\n", NumIdentifiers);
400   fprintf(stderr, "# Empty Buckets: %d\n", NumEmptyBuckets);
401   fprintf(stderr, "Hash density (#identifiers per bucket): %f\n",
402           NumIdentifiers/(double)NumBuckets);
403   fprintf(stderr, "Ave identifier length: %f\n",
404           (AverageIdentifierSize/(double)NumIdentifiers));
405   fprintf(stderr, "Max identifier length: %d\n", MaxIdentifierLength);
406 
407   // Compute statistics about the memory allocated for identifiers.
408   HashTable.getAllocator().PrintStats();
409 }
410 
411 //===----------------------------------------------------------------------===//
412 // SelectorTable Implementation
413 //===----------------------------------------------------------------------===//
414 
415 unsigned llvm::DenseMapInfo<clang::Selector>::getHashValue(clang::Selector S) {
416   return DenseMapInfo<void*>::getHashValue(S.getAsOpaquePtr());
417 }
418 
419 namespace clang {
420 
421 /// One of these variable length records is kept for each
422 /// selector containing more than one keyword. We use a folding set
423 /// to unique aggregate names (keyword selectors in ObjC parlance). Access to
424 /// this class is provided strictly through Selector.
425 class alignas(IdentifierInfoAlignment) MultiKeywordSelector
426     : public detail::DeclarationNameExtra,
427       public llvm::FoldingSetNode {
428   MultiKeywordSelector(unsigned nKeys) : DeclarationNameExtra(nKeys) {}
429 
430 public:
431   // Constructor for keyword selectors.
432   MultiKeywordSelector(unsigned nKeys, IdentifierInfo **IIV)
433       : DeclarationNameExtra(nKeys) {
434     assert((nKeys > 1) && "not a multi-keyword selector");
435 
436     // Fill in the trailing keyword array.
437     IdentifierInfo **KeyInfo = reinterpret_cast<IdentifierInfo **>(this + 1);
438     for (unsigned i = 0; i != nKeys; ++i)
439       KeyInfo[i] = IIV[i];
440   }
441 
442   // getName - Derive the full selector name and return it.
443   std::string getName() const;
444 
445   using DeclarationNameExtra::getNumArgs;
446 
447   using keyword_iterator = IdentifierInfo *const *;
448 
449   keyword_iterator keyword_begin() const {
450     return reinterpret_cast<keyword_iterator>(this + 1);
451   }
452 
453   keyword_iterator keyword_end() const {
454     return keyword_begin() + getNumArgs();
455   }
456 
457   IdentifierInfo *getIdentifierInfoForSlot(unsigned i) const {
458     assert(i < getNumArgs() && "getIdentifierInfoForSlot(): illegal index");
459     return keyword_begin()[i];
460   }
461 
462   static void Profile(llvm::FoldingSetNodeID &ID, keyword_iterator ArgTys,
463                       unsigned NumArgs) {
464     ID.AddInteger(NumArgs);
465     for (unsigned i = 0; i != NumArgs; ++i)
466       ID.AddPointer(ArgTys[i]);
467   }
468 
469   void Profile(llvm::FoldingSetNodeID &ID) {
470     Profile(ID, keyword_begin(), getNumArgs());
471   }
472 };
473 
474 } // namespace clang.
475 
476 bool Selector::isKeywordSelector(ArrayRef<StringRef> Names) const {
477   assert(!Names.empty() && "must have >= 1 selector slots");
478   if (getNumArgs() != Names.size())
479     return false;
480   for (unsigned I = 0, E = Names.size(); I != E; ++I) {
481     if (getNameForSlot(I) != Names[I])
482       return false;
483   }
484   return true;
485 }
486 
487 bool Selector::isUnarySelector(StringRef Name) const {
488   return isUnarySelector() && getNameForSlot(0) == Name;
489 }
490 
491 unsigned Selector::getNumArgs() const {
492   unsigned IIF = getIdentifierInfoFlag();
493   if (IIF <= ZeroArg)
494     return 0;
495   if (IIF == OneArg)
496     return 1;
497   // We point to a MultiKeywordSelector.
498   MultiKeywordSelector *SI = getMultiKeywordSelector();
499   return SI->getNumArgs();
500 }
501 
502 IdentifierInfo *Selector::getIdentifierInfoForSlot(unsigned argIndex) const {
503   if (getIdentifierInfoFlag() < MultiArg) {
504     assert(argIndex == 0 && "illegal keyword index");
505     return getAsIdentifierInfo();
506   }
507 
508   // We point to a MultiKeywordSelector.
509   MultiKeywordSelector *SI = getMultiKeywordSelector();
510   return SI->getIdentifierInfoForSlot(argIndex);
511 }
512 
513 StringRef Selector::getNameForSlot(unsigned int argIndex) const {
514   IdentifierInfo *II = getIdentifierInfoForSlot(argIndex);
515   return II ? II->getName() : StringRef();
516 }
517 
518 std::string MultiKeywordSelector::getName() const {
519   SmallString<256> Str;
520   llvm::raw_svector_ostream OS(Str);
521   for (keyword_iterator I = keyword_begin(), E = keyword_end(); I != E; ++I) {
522     if (*I)
523       OS << (*I)->getName();
524     OS << ':';
525   }
526 
527   return std::string(OS.str());
528 }
529 
530 std::string Selector::getAsString() const {
531   if (InfoPtr == 0)
532     return "<null selector>";
533 
534   if (getIdentifierInfoFlag() < MultiArg) {
535     IdentifierInfo *II = getAsIdentifierInfo();
536 
537     if (getNumArgs() == 0) {
538       assert(II && "If the number of arguments is 0 then II is guaranteed to "
539                    "not be null.");
540       return std::string(II->getName());
541     }
542 
543     if (!II)
544       return ":";
545 
546     return II->getName().str() + ":";
547   }
548 
549   // We have a multiple keyword selector.
550   return getMultiKeywordSelector()->getName();
551 }
552 
553 void Selector::print(llvm::raw_ostream &OS) const {
554   OS << getAsString();
555 }
556 
557 LLVM_DUMP_METHOD void Selector::dump() const { print(llvm::errs()); }
558 
559 /// Interpreting the given string using the normal CamelCase
560 /// conventions, determine whether the given string starts with the
561 /// given "word", which is assumed to end in a lowercase letter.
562 static bool startsWithWord(StringRef name, StringRef word) {
563   if (name.size() < word.size()) return false;
564   return ((name.size() == word.size() || !isLowercase(name[word.size()])) &&
565           name.startswith(word));
566 }
567 
568 ObjCMethodFamily Selector::getMethodFamilyImpl(Selector sel) {
569   IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
570   if (!first) return OMF_None;
571 
572   StringRef name = first->getName();
573   if (sel.isUnarySelector()) {
574     if (name == "autorelease") return OMF_autorelease;
575     if (name == "dealloc") return OMF_dealloc;
576     if (name == "finalize") return OMF_finalize;
577     if (name == "release") return OMF_release;
578     if (name == "retain") return OMF_retain;
579     if (name == "retainCount") return OMF_retainCount;
580     if (name == "self") return OMF_self;
581     if (name == "initialize") return OMF_initialize;
582   }
583 
584   if (name == "performSelector" || name == "performSelectorInBackground" ||
585       name == "performSelectorOnMainThread")
586     return OMF_performSelector;
587 
588   // The other method families may begin with a prefix of underscores.
589   while (!name.empty() && name.front() == '_')
590     name = name.substr(1);
591 
592   if (name.empty()) return OMF_None;
593   switch (name.front()) {
594   case 'a':
595     if (startsWithWord(name, "alloc")) return OMF_alloc;
596     break;
597   case 'c':
598     if (startsWithWord(name, "copy")) return OMF_copy;
599     break;
600   case 'i':
601     if (startsWithWord(name, "init")) return OMF_init;
602     break;
603   case 'm':
604     if (startsWithWord(name, "mutableCopy")) return OMF_mutableCopy;
605     break;
606   case 'n':
607     if (startsWithWord(name, "new")) return OMF_new;
608     break;
609   default:
610     break;
611   }
612 
613   return OMF_None;
614 }
615 
616 ObjCInstanceTypeFamily Selector::getInstTypeMethodFamily(Selector sel) {
617   IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
618   if (!first) return OIT_None;
619 
620   StringRef name = first->getName();
621 
622   if (name.empty()) return OIT_None;
623   switch (name.front()) {
624     case 'a':
625       if (startsWithWord(name, "array")) return OIT_Array;
626       break;
627     case 'd':
628       if (startsWithWord(name, "default")) return OIT_ReturnsSelf;
629       if (startsWithWord(name, "dictionary")) return OIT_Dictionary;
630       break;
631     case 's':
632       if (startsWithWord(name, "shared")) return OIT_ReturnsSelf;
633       if (startsWithWord(name, "standard")) return OIT_Singleton;
634       break;
635     case 'i':
636       if (startsWithWord(name, "init")) return OIT_Init;
637       break;
638     default:
639       break;
640   }
641   return OIT_None;
642 }
643 
644 ObjCStringFormatFamily Selector::getStringFormatFamilyImpl(Selector sel) {
645   IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
646   if (!first) return SFF_None;
647 
648   StringRef name = first->getName();
649 
650   switch (name.front()) {
651     case 'a':
652       if (name == "appendFormat") return SFF_NSString;
653       break;
654 
655     case 'i':
656       if (name == "initWithFormat") return SFF_NSString;
657       break;
658 
659     case 'l':
660       if (name == "localizedStringWithFormat") return SFF_NSString;
661       break;
662 
663     case 's':
664       if (name == "stringByAppendingFormat" ||
665           name == "stringWithFormat") return SFF_NSString;
666       break;
667   }
668   return SFF_None;
669 }
670 
671 namespace {
672 
673 struct SelectorTableImpl {
674   llvm::FoldingSet<MultiKeywordSelector> Table;
675   llvm::BumpPtrAllocator Allocator;
676 };
677 
678 } // namespace
679 
680 static SelectorTableImpl &getSelectorTableImpl(void *P) {
681   return *static_cast<SelectorTableImpl*>(P);
682 }
683 
684 SmallString<64>
685 SelectorTable::constructSetterName(StringRef Name) {
686   SmallString<64> SetterName("set");
687   SetterName += Name;
688   SetterName[3] = toUppercase(SetterName[3]);
689   return SetterName;
690 }
691 
692 Selector
693 SelectorTable::constructSetterSelector(IdentifierTable &Idents,
694                                        SelectorTable &SelTable,
695                                        const IdentifierInfo *Name) {
696   IdentifierInfo *SetterName =
697     &Idents.get(constructSetterName(Name->getName()));
698   return SelTable.getUnarySelector(SetterName);
699 }
700 
701 std::string SelectorTable::getPropertyNameFromSetterSelector(Selector Sel) {
702   StringRef Name = Sel.getNameForSlot(0);
703   assert(Name.startswith("set") && "invalid setter name");
704   return (Twine(toLowercase(Name[3])) + Name.drop_front(4)).str();
705 }
706 
707 size_t SelectorTable::getTotalMemory() const {
708   SelectorTableImpl &SelTabImpl = getSelectorTableImpl(Impl);
709   return SelTabImpl.Allocator.getTotalMemory();
710 }
711 
712 Selector SelectorTable::getSelector(unsigned nKeys, IdentifierInfo **IIV) {
713   if (nKeys < 2)
714     return Selector(IIV[0], nKeys);
715 
716   SelectorTableImpl &SelTabImpl = getSelectorTableImpl(Impl);
717 
718   // Unique selector, to guarantee there is one per name.
719   llvm::FoldingSetNodeID ID;
720   MultiKeywordSelector::Profile(ID, IIV, nKeys);
721 
722   void *InsertPos = nullptr;
723   if (MultiKeywordSelector *SI =
724         SelTabImpl.Table.FindNodeOrInsertPos(ID, InsertPos))
725     return Selector(SI);
726 
727   // MultiKeywordSelector objects are not allocated with new because they have a
728   // variable size array (for parameter types) at the end of them.
729   unsigned Size = sizeof(MultiKeywordSelector) + nKeys*sizeof(IdentifierInfo *);
730   MultiKeywordSelector *SI =
731       (MultiKeywordSelector *)SelTabImpl.Allocator.Allocate(
732           Size, alignof(MultiKeywordSelector));
733   new (SI) MultiKeywordSelector(nKeys, IIV);
734   SelTabImpl.Table.InsertNode(SI, InsertPos);
735   return Selector(SI);
736 }
737 
738 SelectorTable::SelectorTable() {
739   Impl = new SelectorTableImpl();
740 }
741 
742 SelectorTable::~SelectorTable() {
743   delete &getSelectorTableImpl(Impl);
744 }
745 
746 const char *clang::getOperatorSpelling(OverloadedOperatorKind Operator) {
747   switch (Operator) {
748   case OO_None:
749   case NUM_OVERLOADED_OPERATORS:
750     return nullptr;
751 
752 #define OVERLOADED_OPERATOR(Name,Spelling,Token,Unary,Binary,MemberOnly) \
753   case OO_##Name: return Spelling;
754 #include "clang/Basic/OperatorKinds.def"
755   }
756 
757   llvm_unreachable("Invalid OverloadedOperatorKind!");
758 }
759 
760 StringRef clang::getNullabilitySpelling(NullabilityKind kind,
761                                         bool isContextSensitive) {
762   switch (kind) {
763   case NullabilityKind::NonNull:
764     return isContextSensitive ? "nonnull" : "_Nonnull";
765 
766   case NullabilityKind::Nullable:
767     return isContextSensitive ? "nullable" : "_Nullable";
768 
769   case NullabilityKind::NullableResult:
770     assert(!isContextSensitive &&
771            "_Nullable_result isn't supported as context-sensitive keyword");
772     return "_Nullable_result";
773 
774   case NullabilityKind::Unspecified:
775     return isContextSensitive ? "null_unspecified" : "_Null_unspecified";
776   }
777   llvm_unreachable("Unknown nullability kind.");
778 }
779