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     KEYMAX        = KEYCUDA, // The maximum key
113     KEYALLCXX = KEYCXX | KEYCXX11 | KEYCXX20,
114     KEYALL = (KEYMAX | (KEYMAX-1)) & ~KEYNOMS18 &
115              ~KEYNOOPENCL // KEYNOMS18 and KEYNOOPENCL are used to exclude.
116   };
117 
118   /// How a keyword is treated in the selected standard.
119   enum KeywordStatus {
120     KS_Disabled,    // Disabled
121     KS_Extension,   // Is an extension
122     KS_Enabled,     // Enabled
123     KS_Future       // Is a keyword in future standard
124   };
125 
126 } // namespace
127 
128 /// Translates flags as specified in TokenKinds.def into keyword status
129 /// in the given language standard.
130 static KeywordStatus getKeywordStatus(const LangOptions &LangOpts,
131                                       unsigned Flags) {
132   if (Flags == KEYALL) return KS_Enabled;
133   if (LangOpts.CPlusPlus && (Flags & KEYCXX)) return KS_Enabled;
134   if (LangOpts.CPlusPlus11 && (Flags & KEYCXX11)) return KS_Enabled;
135   if (LangOpts.CPlusPlus20 && (Flags & KEYCXX20)) return KS_Enabled;
136   if (LangOpts.C99 && (Flags & KEYC99)) return KS_Enabled;
137   if (LangOpts.GNUKeywords && (Flags & KEYGNU)) return KS_Extension;
138   if (LangOpts.MicrosoftExt && (Flags & KEYMS)) return KS_Extension;
139   if (LangOpts.MSVCCompat && (Flags & KEYMSCOMPAT)) return KS_Enabled;
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.Char8 && (Flags & CHAR8SUPPORT)) return KS_Enabled;
145   if (LangOpts.AltiVec && (Flags & KEYALTIVEC)) return KS_Enabled;
146   if (LangOpts.ZVector && (Flags & KEYZVECTOR)) return KS_Enabled;
147   if (LangOpts.OpenCL && !LangOpts.OpenCLCPlusPlus && (Flags & KEYOPENCLC))
148     return KS_Enabled;
149   if (LangOpts.OpenCLCPlusPlus && (Flags & KEYOPENCLCXX)) return KS_Enabled;
150   if (!LangOpts.CPlusPlus && (Flags & KEYNOCXX)) return KS_Enabled;
151   if (LangOpts.C11 && (Flags & KEYC11)) return KS_Enabled;
152   // We treat bridge casts as objective-C keywords so we can warn on them
153   // in non-arc mode.
154   if (LangOpts.ObjC && (Flags & KEYOBJC)) return KS_Enabled;
155   if (LangOpts.CPlusPlus20 && (Flags & KEYCONCEPTS)) return KS_Enabled;
156   if (LangOpts.Coroutines && (Flags & KEYCOROUTINES)) return KS_Enabled;
157   if (LangOpts.ModulesTS && (Flags & KEYMODULES)) return KS_Enabled;
158   if (LangOpts.CPlusPlus && (Flags & KEYALLCXX)) return KS_Future;
159   if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus20 && (Flags & CHAR8SUPPORT))
160     return KS_Future;
161   if (LangOpts.isSYCL() && (Flags & KEYSYCL))
162     return KS_Enabled;
163   if (LangOpts.CUDA && (Flags & KEYCUDA))
164     return KS_Enabled;
165   return KS_Disabled;
166 }
167 
168 /// AddKeyword - This method is used to associate a token ID with specific
169 /// identifiers because they are language keywords.  This causes the lexer to
170 /// automatically map matching identifiers to specialized token codes.
171 static void AddKeyword(StringRef Keyword,
172                        tok::TokenKind TokenCode, unsigned Flags,
173                        const LangOptions &LangOpts, IdentifierTable &Table) {
174   KeywordStatus AddResult = getKeywordStatus(LangOpts, Flags);
175 
176   // Don't add this keyword under MSVCCompat.
177   if (LangOpts.MSVCCompat && (Flags & KEYNOMS18) &&
178       !LangOpts.isCompatibleWithMSVC(LangOptions::MSVC2015))
179     return;
180 
181   // Don't add this keyword under OpenCL.
182   if (LangOpts.OpenCL && (Flags & KEYNOOPENCL))
183     return;
184 
185   // Don't add this keyword if disabled in this language.
186   if (AddResult == KS_Disabled) return;
187 
188   IdentifierInfo &Info =
189       Table.get(Keyword, AddResult == KS_Future ? tok::identifier : TokenCode);
190   Info.setIsExtensionToken(AddResult == KS_Extension);
191   Info.setIsFutureCompatKeyword(AddResult == KS_Future);
192 }
193 
194 /// AddCXXOperatorKeyword - Register a C++ operator keyword alternative
195 /// representations.
196 static void AddCXXOperatorKeyword(StringRef Keyword,
197                                   tok::TokenKind TokenCode,
198                                   IdentifierTable &Table) {
199   IdentifierInfo &Info = Table.get(Keyword, TokenCode);
200   Info.setIsCPlusPlusOperatorKeyword();
201 }
202 
203 /// AddObjCKeyword - Register an Objective-C \@keyword like "class" "selector"
204 /// or "property".
205 static void AddObjCKeyword(StringRef Name,
206                            tok::ObjCKeywordKind ObjCID,
207                            IdentifierTable &Table) {
208   Table.get(Name).setObjCKeywordID(ObjCID);
209 }
210 
211 /// AddKeywords - Add all keywords to the symbol table.
212 ///
213 void IdentifierTable::AddKeywords(const LangOptions &LangOpts) {
214   // Add keywords and tokens for the current language.
215 #define KEYWORD(NAME, FLAGS) \
216   AddKeyword(StringRef(#NAME), tok::kw_ ## NAME,  \
217              FLAGS, LangOpts, *this);
218 #define ALIAS(NAME, TOK, FLAGS) \
219   AddKeyword(StringRef(NAME), tok::kw_ ## TOK,  \
220              FLAGS, LangOpts, *this);
221 #define CXX_KEYWORD_OPERATOR(NAME, ALIAS) \
222   if (LangOpts.CXXOperatorNames)          \
223     AddCXXOperatorKeyword(StringRef(#NAME), tok::ALIAS, *this);
224 #define OBJC_AT_KEYWORD(NAME)  \
225   if (LangOpts.ObjC)           \
226     AddObjCKeyword(StringRef(#NAME), tok::objc_##NAME, *this);
227 #define TESTING_KEYWORD(NAME, FLAGS)
228 #include "clang/Basic/TokenKinds.def"
229 
230   if (LangOpts.ParseUnknownAnytype)
231     AddKeyword("__unknown_anytype", tok::kw___unknown_anytype, KEYALL,
232                LangOpts, *this);
233 
234   if (LangOpts.DeclSpecKeyword)
235     AddKeyword("__declspec", tok::kw___declspec, KEYALL, LangOpts, *this);
236 
237   if (LangOpts.IEEE128)
238     AddKeyword("__ieee128", tok::kw___float128, KEYALL, LangOpts, *this);
239 
240   // Add the 'import' contextual keyword.
241   get("import").setModulesImport(true);
242 }
243 
244 /// Checks if the specified token kind represents a keyword in the
245 /// specified language.
246 /// \returns Status of the keyword in the language.
247 static KeywordStatus getTokenKwStatus(const LangOptions &LangOpts,
248                                       tok::TokenKind K) {
249   switch (K) {
250 #define KEYWORD(NAME, FLAGS) \
251   case tok::kw_##NAME: return getKeywordStatus(LangOpts, FLAGS);
252 #include "clang/Basic/TokenKinds.def"
253   default: return KS_Disabled;
254   }
255 }
256 
257 /// Returns true if the identifier represents a keyword in the
258 /// specified language.
259 bool IdentifierInfo::isKeyword(const LangOptions &LangOpts) const {
260   switch (getTokenKwStatus(LangOpts, getTokenID())) {
261   case KS_Enabled:
262   case KS_Extension:
263     return true;
264   default:
265     return false;
266   }
267 }
268 
269 /// Returns true if the identifier represents a C++ keyword in the
270 /// specified language.
271 bool IdentifierInfo::isCPlusPlusKeyword(const LangOptions &LangOpts) const {
272   if (!LangOpts.CPlusPlus || !isKeyword(LangOpts))
273     return false;
274   // This is a C++ keyword if this identifier is not a keyword when checked
275   // using LangOptions without C++ support.
276   LangOptions LangOptsNoCPP = LangOpts;
277   LangOptsNoCPP.CPlusPlus = false;
278   LangOptsNoCPP.CPlusPlus11 = false;
279   LangOptsNoCPP.CPlusPlus20 = false;
280   return !isKeyword(LangOptsNoCPP);
281 }
282 
283 ReservedIdentifierStatus
284 IdentifierInfo::isReserved(const LangOptions &LangOpts) const {
285   StringRef Name = getName();
286 
287   // '_' is a reserved identifier, but its use is so common (e.g. to store
288   // ignored values) that we don't warn on it.
289   if (Name.size() <= 1)
290     return ReservedIdentifierStatus::NotReserved;
291 
292   // [lex.name] p3
293   if (Name[0] == '_') {
294 
295     // Each name that begins with an underscore followed by an uppercase letter
296     // or another underscore is reserved.
297     if (Name[1] == '_')
298       return ReservedIdentifierStatus::StartsWithDoubleUnderscore;
299 
300     if ('A' <= Name[1] && Name[1] <= 'Z')
301       return ReservedIdentifierStatus::
302           StartsWithUnderscoreFollowedByCapitalLetter;
303 
304     // This is a bit misleading: it actually means it's only reserved if we're
305     // at global scope because it starts with an underscore.
306     return ReservedIdentifierStatus::StartsWithUnderscoreAtGlobalScope;
307   }
308 
309   // Each name that contains a double underscore (__) is reserved.
310   if (LangOpts.CPlusPlus && Name.contains("__"))
311     return ReservedIdentifierStatus::ContainsDoubleUnderscore;
312 
313   return ReservedIdentifierStatus::NotReserved;
314 }
315 
316 StringRef IdentifierInfo::deuglifiedName() const {
317   StringRef Name = getName();
318   if (Name.size() >= 2 && Name.front() == '_' &&
319       (Name[1] == '_' || (Name[1] >= 'A' && Name[1] <= 'Z')))
320     return Name.ltrim('_');
321   return Name;
322 }
323 
324 tok::PPKeywordKind IdentifierInfo::getPPKeywordID() const {
325   // We use a perfect hash function here involving the length of the keyword,
326   // the first and third character.  For preprocessor ID's there are no
327   // collisions (if there were, the switch below would complain about duplicate
328   // case values).  Note that this depends on 'if' being null terminated.
329 
330 #define HASH(LEN, FIRST, THIRD) \
331   (LEN << 5) + (((FIRST-'a') + (THIRD-'a')) & 31)
332 #define CASE(LEN, FIRST, THIRD, NAME) \
333   case HASH(LEN, FIRST, THIRD): \
334     return memcmp(Name, #NAME, LEN) ? tok::pp_not_keyword : tok::pp_ ## NAME
335 
336   unsigned Len = getLength();
337   if (Len < 2) return tok::pp_not_keyword;
338   const char *Name = getNameStart();
339   switch (HASH(Len, Name[0], Name[2])) {
340   default: return tok::pp_not_keyword;
341   CASE( 2, 'i', '\0', if);
342   CASE( 4, 'e', 'i', elif);
343   CASE( 4, 'e', 's', else);
344   CASE( 4, 'l', 'n', line);
345   CASE( 4, 's', 'c', sccs);
346   CASE( 5, 'e', 'd', endif);
347   CASE( 5, 'e', 'r', error);
348   CASE( 5, 'i', 'e', ident);
349   CASE( 5, 'i', 'd', ifdef);
350   CASE( 5, 'u', 'd', undef);
351 
352   CASE( 6, 'a', 's', assert);
353   CASE( 6, 'd', 'f', define);
354   CASE( 6, 'i', 'n', ifndef);
355   CASE( 6, 'i', 'p', import);
356   CASE( 6, 'p', 'a', pragma);
357 
358   CASE( 7, 'd', 'f', defined);
359   CASE( 7, 'e', 'i', elifdef);
360   CASE( 7, 'i', 'c', include);
361   CASE( 7, 'w', 'r', warning);
362 
363   CASE( 8, 'e', 'i', elifndef);
364   CASE( 8, 'u', 'a', unassert);
365   CASE(12, 'i', 'c', include_next);
366 
367   CASE(14, '_', 'p', __public_macro);
368 
369   CASE(15, '_', 'p', __private_macro);
370 
371   CASE(16, '_', 'i', __include_macros);
372 #undef CASE
373 #undef HASH
374   }
375 }
376 
377 //===----------------------------------------------------------------------===//
378 // Stats Implementation
379 //===----------------------------------------------------------------------===//
380 
381 /// PrintStats - Print statistics about how well the identifier table is doing
382 /// at hashing identifiers.
383 void IdentifierTable::PrintStats() const {
384   unsigned NumBuckets = HashTable.getNumBuckets();
385   unsigned NumIdentifiers = HashTable.getNumItems();
386   unsigned NumEmptyBuckets = NumBuckets-NumIdentifiers;
387   unsigned AverageIdentifierSize = 0;
388   unsigned MaxIdentifierLength = 0;
389 
390   // TODO: Figure out maximum times an identifier had to probe for -stats.
391   for (llvm::StringMap<IdentifierInfo*, llvm::BumpPtrAllocator>::const_iterator
392        I = HashTable.begin(), E = HashTable.end(); I != E; ++I) {
393     unsigned IdLen = I->getKeyLength();
394     AverageIdentifierSize += IdLen;
395     if (MaxIdentifierLength < IdLen)
396       MaxIdentifierLength = IdLen;
397   }
398 
399   fprintf(stderr, "\n*** Identifier Table Stats:\n");
400   fprintf(stderr, "# Identifiers:   %d\n", NumIdentifiers);
401   fprintf(stderr, "# Empty Buckets: %d\n", NumEmptyBuckets);
402   fprintf(stderr, "Hash density (#identifiers per bucket): %f\n",
403           NumIdentifiers/(double)NumBuckets);
404   fprintf(stderr, "Ave identifier length: %f\n",
405           (AverageIdentifierSize/(double)NumIdentifiers));
406   fprintf(stderr, "Max identifier length: %d\n", MaxIdentifierLength);
407 
408   // Compute statistics about the memory allocated for identifiers.
409   HashTable.getAllocator().PrintStats();
410 }
411 
412 //===----------------------------------------------------------------------===//
413 // SelectorTable Implementation
414 //===----------------------------------------------------------------------===//
415 
416 unsigned llvm::DenseMapInfo<clang::Selector>::getHashValue(clang::Selector S) {
417   return DenseMapInfo<void*>::getHashValue(S.getAsOpaquePtr());
418 }
419 
420 namespace clang {
421 
422 /// One of these variable length records is kept for each
423 /// selector containing more than one keyword. We use a folding set
424 /// to unique aggregate names (keyword selectors in ObjC parlance). Access to
425 /// this class is provided strictly through Selector.
426 class alignas(IdentifierInfoAlignment) MultiKeywordSelector
427     : public detail::DeclarationNameExtra,
428       public llvm::FoldingSetNode {
429   MultiKeywordSelector(unsigned nKeys) : DeclarationNameExtra(nKeys) {}
430 
431 public:
432   // Constructor for keyword selectors.
433   MultiKeywordSelector(unsigned nKeys, IdentifierInfo **IIV)
434       : DeclarationNameExtra(nKeys) {
435     assert((nKeys > 1) && "not a multi-keyword selector");
436 
437     // Fill in the trailing keyword array.
438     IdentifierInfo **KeyInfo = reinterpret_cast<IdentifierInfo **>(this + 1);
439     for (unsigned i = 0; i != nKeys; ++i)
440       KeyInfo[i] = IIV[i];
441   }
442 
443   // getName - Derive the full selector name and return it.
444   std::string getName() const;
445 
446   using DeclarationNameExtra::getNumArgs;
447 
448   using keyword_iterator = IdentifierInfo *const *;
449 
450   keyword_iterator keyword_begin() const {
451     return reinterpret_cast<keyword_iterator>(this + 1);
452   }
453 
454   keyword_iterator keyword_end() const {
455     return keyword_begin() + getNumArgs();
456   }
457 
458   IdentifierInfo *getIdentifierInfoForSlot(unsigned i) const {
459     assert(i < getNumArgs() && "getIdentifierInfoForSlot(): illegal index");
460     return keyword_begin()[i];
461   }
462 
463   static void Profile(llvm::FoldingSetNodeID &ID, keyword_iterator ArgTys,
464                       unsigned NumArgs) {
465     ID.AddInteger(NumArgs);
466     for (unsigned i = 0; i != NumArgs; ++i)
467       ID.AddPointer(ArgTys[i]);
468   }
469 
470   void Profile(llvm::FoldingSetNodeID &ID) {
471     Profile(ID, keyword_begin(), getNumArgs());
472   }
473 };
474 
475 } // namespace clang.
476 
477 bool Selector::isKeywordSelector(ArrayRef<StringRef> Names) const {
478   assert(!Names.empty() && "must have >= 1 selector slots");
479   if (getNumArgs() != Names.size())
480     return false;
481   for (unsigned I = 0, E = Names.size(); I != E; ++I) {
482     if (getNameForSlot(I) != Names[I])
483       return false;
484   }
485   return true;
486 }
487 
488 bool Selector::isUnarySelector(StringRef Name) const {
489   return isUnarySelector() && getNameForSlot(0) == Name;
490 }
491 
492 unsigned Selector::getNumArgs() const {
493   unsigned IIF = getIdentifierInfoFlag();
494   if (IIF <= ZeroArg)
495     return 0;
496   if (IIF == OneArg)
497     return 1;
498   // We point to a MultiKeywordSelector.
499   MultiKeywordSelector *SI = getMultiKeywordSelector();
500   return SI->getNumArgs();
501 }
502 
503 IdentifierInfo *Selector::getIdentifierInfoForSlot(unsigned argIndex) const {
504   if (getIdentifierInfoFlag() < MultiArg) {
505     assert(argIndex == 0 && "illegal keyword index");
506     return getAsIdentifierInfo();
507   }
508 
509   // We point to a MultiKeywordSelector.
510   MultiKeywordSelector *SI = getMultiKeywordSelector();
511   return SI->getIdentifierInfoForSlot(argIndex);
512 }
513 
514 StringRef Selector::getNameForSlot(unsigned int argIndex) const {
515   IdentifierInfo *II = getIdentifierInfoForSlot(argIndex);
516   return II ? II->getName() : StringRef();
517 }
518 
519 std::string MultiKeywordSelector::getName() const {
520   SmallString<256> Str;
521   llvm::raw_svector_ostream OS(Str);
522   for (keyword_iterator I = keyword_begin(), E = keyword_end(); I != E; ++I) {
523     if (*I)
524       OS << (*I)->getName();
525     OS << ':';
526   }
527 
528   return std::string(OS.str());
529 }
530 
531 std::string Selector::getAsString() const {
532   if (InfoPtr == 0)
533     return "<null selector>";
534 
535   if (getIdentifierInfoFlag() < MultiArg) {
536     IdentifierInfo *II = getAsIdentifierInfo();
537 
538     if (getNumArgs() == 0) {
539       assert(II && "If the number of arguments is 0 then II is guaranteed to "
540                    "not be null.");
541       return std::string(II->getName());
542     }
543 
544     if (!II)
545       return ":";
546 
547     return II->getName().str() + ":";
548   }
549 
550   // We have a multiple keyword selector.
551   return getMultiKeywordSelector()->getName();
552 }
553 
554 void Selector::print(llvm::raw_ostream &OS) const {
555   OS << getAsString();
556 }
557 
558 LLVM_DUMP_METHOD void Selector::dump() const { print(llvm::errs()); }
559 
560 /// Interpreting the given string using the normal CamelCase
561 /// conventions, determine whether the given string starts with the
562 /// given "word", which is assumed to end in a lowercase letter.
563 static bool startsWithWord(StringRef name, StringRef word) {
564   if (name.size() < word.size()) return false;
565   return ((name.size() == word.size() || !isLowercase(name[word.size()])) &&
566           name.startswith(word));
567 }
568 
569 ObjCMethodFamily Selector::getMethodFamilyImpl(Selector sel) {
570   IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
571   if (!first) return OMF_None;
572 
573   StringRef name = first->getName();
574   if (sel.isUnarySelector()) {
575     if (name == "autorelease") return OMF_autorelease;
576     if (name == "dealloc") return OMF_dealloc;
577     if (name == "finalize") return OMF_finalize;
578     if (name == "release") return OMF_release;
579     if (name == "retain") return OMF_retain;
580     if (name == "retainCount") return OMF_retainCount;
581     if (name == "self") return OMF_self;
582     if (name == "initialize") return OMF_initialize;
583   }
584 
585   if (name == "performSelector" || name == "performSelectorInBackground" ||
586       name == "performSelectorOnMainThread")
587     return OMF_performSelector;
588 
589   // The other method families may begin with a prefix of underscores.
590   while (!name.empty() && name.front() == '_')
591     name = name.substr(1);
592 
593   if (name.empty()) return OMF_None;
594   switch (name.front()) {
595   case 'a':
596     if (startsWithWord(name, "alloc")) return OMF_alloc;
597     break;
598   case 'c':
599     if (startsWithWord(name, "copy")) return OMF_copy;
600     break;
601   case 'i':
602     if (startsWithWord(name, "init")) return OMF_init;
603     break;
604   case 'm':
605     if (startsWithWord(name, "mutableCopy")) return OMF_mutableCopy;
606     break;
607   case 'n':
608     if (startsWithWord(name, "new")) return OMF_new;
609     break;
610   default:
611     break;
612   }
613 
614   return OMF_None;
615 }
616 
617 ObjCInstanceTypeFamily Selector::getInstTypeMethodFamily(Selector sel) {
618   IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
619   if (!first) return OIT_None;
620 
621   StringRef name = first->getName();
622 
623   if (name.empty()) return OIT_None;
624   switch (name.front()) {
625     case 'a':
626       if (startsWithWord(name, "array")) return OIT_Array;
627       break;
628     case 'd':
629       if (startsWithWord(name, "default")) return OIT_ReturnsSelf;
630       if (startsWithWord(name, "dictionary")) return OIT_Dictionary;
631       break;
632     case 's':
633       if (startsWithWord(name, "shared")) return OIT_ReturnsSelf;
634       if (startsWithWord(name, "standard")) return OIT_Singleton;
635       break;
636     case 'i':
637       if (startsWithWord(name, "init")) return OIT_Init;
638       break;
639     default:
640       break;
641   }
642   return OIT_None;
643 }
644 
645 ObjCStringFormatFamily Selector::getStringFormatFamilyImpl(Selector sel) {
646   IdentifierInfo *first = sel.getIdentifierInfoForSlot(0);
647   if (!first) return SFF_None;
648 
649   StringRef name = first->getName();
650 
651   switch (name.front()) {
652     case 'a':
653       if (name == "appendFormat") return SFF_NSString;
654       break;
655 
656     case 'i':
657       if (name == "initWithFormat") return SFF_NSString;
658       break;
659 
660     case 'l':
661       if (name == "localizedStringWithFormat") return SFF_NSString;
662       break;
663 
664     case 's':
665       if (name == "stringByAppendingFormat" ||
666           name == "stringWithFormat") return SFF_NSString;
667       break;
668   }
669   return SFF_None;
670 }
671 
672 namespace {
673 
674 struct SelectorTableImpl {
675   llvm::FoldingSet<MultiKeywordSelector> Table;
676   llvm::BumpPtrAllocator Allocator;
677 };
678 
679 } // namespace
680 
681 static SelectorTableImpl &getSelectorTableImpl(void *P) {
682   return *static_cast<SelectorTableImpl*>(P);
683 }
684 
685 SmallString<64>
686 SelectorTable::constructSetterName(StringRef Name) {
687   SmallString<64> SetterName("set");
688   SetterName += Name;
689   SetterName[3] = toUppercase(SetterName[3]);
690   return SetterName;
691 }
692 
693 Selector
694 SelectorTable::constructSetterSelector(IdentifierTable &Idents,
695                                        SelectorTable &SelTable,
696                                        const IdentifierInfo *Name) {
697   IdentifierInfo *SetterName =
698     &Idents.get(constructSetterName(Name->getName()));
699   return SelTable.getUnarySelector(SetterName);
700 }
701 
702 std::string SelectorTable::getPropertyNameFromSetterSelector(Selector Sel) {
703   StringRef Name = Sel.getNameForSlot(0);
704   assert(Name.startswith("set") && "invalid setter name");
705   return (Twine(toLowercase(Name[3])) + Name.drop_front(4)).str();
706 }
707 
708 size_t SelectorTable::getTotalMemory() const {
709   SelectorTableImpl &SelTabImpl = getSelectorTableImpl(Impl);
710   return SelTabImpl.Allocator.getTotalMemory();
711 }
712 
713 Selector SelectorTable::getSelector(unsigned nKeys, IdentifierInfo **IIV) {
714   if (nKeys < 2)
715     return Selector(IIV[0], nKeys);
716 
717   SelectorTableImpl &SelTabImpl = getSelectorTableImpl(Impl);
718 
719   // Unique selector, to guarantee there is one per name.
720   llvm::FoldingSetNodeID ID;
721   MultiKeywordSelector::Profile(ID, IIV, nKeys);
722 
723   void *InsertPos = nullptr;
724   if (MultiKeywordSelector *SI =
725         SelTabImpl.Table.FindNodeOrInsertPos(ID, InsertPos))
726     return Selector(SI);
727 
728   // MultiKeywordSelector objects are not allocated with new because they have a
729   // variable size array (for parameter types) at the end of them.
730   unsigned Size = sizeof(MultiKeywordSelector) + nKeys*sizeof(IdentifierInfo *);
731   MultiKeywordSelector *SI =
732       (MultiKeywordSelector *)SelTabImpl.Allocator.Allocate(
733           Size, alignof(MultiKeywordSelector));
734   new (SI) MultiKeywordSelector(nKeys, IIV);
735   SelTabImpl.Table.InsertNode(SI, InsertPos);
736   return Selector(SI);
737 }
738 
739 SelectorTable::SelectorTable() {
740   Impl = new SelectorTableImpl();
741 }
742 
743 SelectorTable::~SelectorTable() {
744   delete &getSelectorTableImpl(Impl);
745 }
746 
747 const char *clang::getOperatorSpelling(OverloadedOperatorKind Operator) {
748   switch (Operator) {
749   case OO_None:
750   case NUM_OVERLOADED_OPERATORS:
751     return nullptr;
752 
753 #define OVERLOADED_OPERATOR(Name,Spelling,Token,Unary,Binary,MemberOnly) \
754   case OO_##Name: return Spelling;
755 #include "clang/Basic/OperatorKinds.def"
756   }
757 
758   llvm_unreachable("Invalid OverloadedOperatorKind!");
759 }
760 
761 StringRef clang::getNullabilitySpelling(NullabilityKind kind,
762                                         bool isContextSensitive) {
763   switch (kind) {
764   case NullabilityKind::NonNull:
765     return isContextSensitive ? "nonnull" : "_Nonnull";
766 
767   case NullabilityKind::Nullable:
768     return isContextSensitive ? "nullable" : "_Nullable";
769 
770   case NullabilityKind::NullableResult:
771     assert(!isContextSensitive &&
772            "_Nullable_result isn't supported as context-sensitive keyword");
773     return "_Nullable_result";
774 
775   case NullabilityKind::Unspecified:
776     return isContextSensitive ? "null_unspecified" : "_Null_unspecified";
777   }
778   llvm_unreachable("Unknown nullability kind.");
779 }
780