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/DiagnosticLex.h"
17#include "clang/Basic/LangOptions.h"
18#include "clang/Basic/OperatorKinds.h"
19#include "clang/Basic/Specifiers.h"
20#include "clang/Basic/TargetBuiltins.h"
21#include "clang/Basic/TokenKinds.h"
22#include "llvm/ADT/DenseMapInfo.h"
23#include "llvm/ADT/Enum.h"
24#include "llvm/ADT/FoldingSet.h"
25#include "llvm/ADT/StringMap.h"
26#include "llvm/ADT/StringRef.h"
27#include "llvm/Support/Allocator.h"
28#include "llvm/Support/raw_ostream.h"
29#include <cassert>
30#include <cstdio>
31#include <cstring>
32#include <string>
33
34using 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.
39static_assert(2 * LargestBuiltinID < (2 << (InterestingIdentifierBits - 1)),
40 "Insufficient ObjCOrBuiltinID Bits");
41
42//===----------------------------------------------------------------------===//
43// IdentifierTable Implementation
44//===----------------------------------------------------------------------===//
45
46IdentifierIterator::~IdentifierIterator() = default;
47
48IdentifierInfoLookup::~IdentifierInfoLookup() = default;
49
50namespace {
51
52/// A simple identifier lookup iterator that represents an
53/// empty sequence of identifiers.
54class EmptyLookupIterator : public IdentifierIterator {
55public:
56 StringRef Next() override { return StringRef(); }
57};
58
59} // namespace
60
61IdentifierIterator *IdentifierInfoLookup::getIdentifiers() {
62 return new EmptyLookupIterator();
63}
64
65IdentifierTable::IdentifierTable(IdentifierInfoLookup *ExternalLookup)
66 : HashTable(8192), // Start with space for 8K identifiers.
67 ExternalLookup(ExternalLookup) {}
68
69IdentifierTable::IdentifierTable(const LangOptions &LangOpts,
70 IdentifierInfoLookup *ExternalLookup)
71 : IdentifierTable(ExternalLookup) {
72 // Populate the identifier table with info about keywords for the current
73 // language.
74 AddKeywords(LangOpts);
75}
76
77//===----------------------------------------------------------------------===//
78// Language Keyword Implementation
79//===----------------------------------------------------------------------===//
80
81// This works on a single TokenKey flag and checks the LangOpts to get the
82// KeywordStatus based exclusively on this flag, so that it can be merged in
83// getKeywordStatus. Most should be enabled/disabled, but some might imply
84// 'future' versions, or extensions. Returns 'unknown' unless this is KNOWN to
85// be disabled, and the calling function makes it 'disabled' if no other flag
86// changes it. This is necessary for the KEYNOCXX and KEYNOOPENCL flags.
87static KeywordStatus getKeywordStatusHelper(const LangOptions &LangOpts,
88 TokenKey Flag) {
89 // Flag is a single bit version of TokenKey (that is, not
90 // KEYALL/KEYALLCXX/etc), so we can check with == throughout this function.
91 assert((Flag & ~(Flag - 1)) == Flag && "Multiple bits set?");
92
93 switch (Flag) {
94 case KEYC99:
95 if (LangOpts.C99)
96 return KS_Enabled;
97 return !LangOpts.CPlusPlus ? KS_Future : KS_Unknown;
98 case KEYC23:
99 if (LangOpts.C23)
100 return KS_Enabled;
101 return !LangOpts.CPlusPlus ? KS_Future : KS_Unknown;
102 case KEYCXX:
103 return LangOpts.CPlusPlus ? KS_Enabled : KS_Unknown;
104 case KEYCXX11:
105 if (LangOpts.CPlusPlus11)
106 return KS_Enabled;
107 return LangOpts.CPlusPlus ? KS_Future : KS_Unknown;
108 case KEYCXX20:
109 if (LangOpts.CPlusPlus20)
110 return KS_Enabled;
111 return LangOpts.CPlusPlus ? KS_Future : KS_Unknown;
112 case KEYGNU:
113 return LangOpts.GNUKeywords ? KS_Extension : KS_Unknown;
114 case KEYMS:
115 return LangOpts.MicrosoftExt ? KS_Extension : KS_Unknown;
116 case BOOLSUPPORT:
117 if (LangOpts.Bool) return KS_Enabled;
118 return !LangOpts.CPlusPlus ? KS_Future : KS_Unknown;
119 case KEYALTIVEC:
120 return LangOpts.AltiVec ? KS_Enabled : KS_Unknown;
121 case KEYBORLAND:
122 return LangOpts.Borland ? KS_Extension : KS_Unknown;
123 case KEYOPENCLC:
124 return LangOpts.OpenCL && !LangOpts.OpenCLCPlusPlus ? KS_Enabled
125 : KS_Unknown;
126 case WCHARSUPPORT:
127 return LangOpts.WChar ? KS_Enabled : KS_Unknown;
128 case HALFSUPPORT:
129 return LangOpts.Half ? KS_Enabled : KS_Unknown;
130 case CHAR8SUPPORT:
131 if (LangOpts.Char8) return KS_Enabled;
132 if (LangOpts.CPlusPlus20) return KS_Unknown;
133 if (LangOpts.CPlusPlus) return KS_Future;
134 return KS_Unknown;
135 case KEYOBJC:
136 // We treat bridge casts as objective-C keywords so we can warn on them
137 // in non-arc mode.
138 return LangOpts.ObjC ? KS_Enabled : KS_Unknown;
139 case KEYZVECTOR:
140 return LangOpts.ZVector ? KS_Enabled : KS_Unknown;
141 case KEYCOROUTINES:
142 return LangOpts.Coroutines ? KS_Enabled : KS_Unknown;
143 case KEYMODULES:
144 return KS_Unknown;
145 case KEYOPENCLCXX:
146 return LangOpts.OpenCLCPlusPlus ? KS_Enabled : KS_Unknown;
147 case KEYMSCOMPAT:
148 return LangOpts.MSVCCompat ? KS_Enabled : KS_Unknown;
149 case KEYSYCL:
150 return LangOpts.isSYCL() ? KS_Enabled : KS_Unknown;
151 case KEYCUDA:
152 return LangOpts.CUDA ? KS_Enabled : KS_Unknown;
153 case KEYZOS:
154 return LangOpts.ZOSExt ? KS_Enabled : KS_Unknown;
155 case KEYHLSL:
156 return LangOpts.HLSL ? KS_Enabled : KS_Unknown;
157 case KEYNOCXX:
158 // This is enabled in all non-C++ modes, but might be enabled for other
159 // reasons as well.
160 return LangOpts.CPlusPlus ? KS_Unknown : KS_Enabled;
161 case KEYNOOPENCL:
162 case KEYNOMS18:
163 case KEYNOHLSL:
164 // The disable behavior for this is handled in getKeywordStatus.
165 return KS_Unknown;
166 case KEYFIXEDPOINT:
167 return LangOpts.FixedPoint ? KS_Enabled : KS_Disabled;
168 case KEYDEFERTS:
169 return LangOpts.DeferTS ? KS_Enabled : KS_Disabled;
170 default:
171 llvm_unreachable("Unknown KeywordStatus flag");
172 }
173}
174
175KeywordStatus clang::getKeywordStatus(const LangOptions &LangOpts,
176 unsigned Flags) {
177 // KEYALL means always enabled, so special case this one.
178 if (Flags == KEYALL) return KS_Enabled;
179 // These are tests that need to 'always win', as they are special in that they
180 // disable based on certain conditions.
181 if (LangOpts.OpenCL && (Flags & KEYNOOPENCL)) return KS_Disabled;
182 if (LangOpts.HLSL && (Flags & KEYNOHLSL))
183 return KS_Disabled;
184 if (LangOpts.MSVCCompat && (Flags & KEYNOMS18) &&
185 !LangOpts.isCompatibleWithMSVC(MajorVersion: LangOptions::MSVC2015))
186 return KS_Disabled;
187 KeywordStatus CurStatus = KS_Unknown;
188
189 while (Flags != 0) {
190 unsigned CurFlag = Flags & ~(Flags - 1);
191 Flags = Flags & ~CurFlag;
192 CurStatus = std::max(
193 a: CurStatus,
194 b: getKeywordStatusHelper(LangOpts, Flag: static_cast<TokenKey>(CurFlag)));
195 }
196
197 if (CurStatus == KS_Unknown)
198 return KS_Disabled;
199 return CurStatus;
200}
201
202static bool IsKeywordInCpp(unsigned Flags) {
203 return (Flags & (KEYCXX | KEYCXX11 | KEYCXX20 | BOOLSUPPORT | WCHARSUPPORT |
204 CHAR8SUPPORT)) != 0;
205}
206
207static void MarkIdentifierAsKeywordInCpp(IdentifierTable &Table,
208 StringRef Name) {
209 IdentifierInfo &II = Table.get(Name, TokenCode: tok::identifier);
210 II.setIsKeywordInCPlusPlus();
211 II.setHandleIdentifierCase();
212}
213
214/// AddKeyword - This method is used to associate a token ID with specific
215/// identifiers because they are language keywords. This causes the lexer to
216/// automatically map matching identifiers to specialized token codes.
217static void AddKeyword(StringRef Keyword,
218 tok::TokenKind TokenCode, unsigned Flags,
219 const LangOptions &LangOpts, IdentifierTable &Table) {
220 KeywordStatus AddResult = getKeywordStatus(LangOpts, Flags);
221
222 // Don't add this keyword if disabled in this language and isn't otherwise
223 // special.
224 if (AddResult == KS_Disabled) {
225 // We do not consider any identifiers to be C++ keywords when in
226 // Objective-C because @ effectively introduces a custom grammar where C++
227 // keywords can be used (and similar for selectors). We could enable this
228 // for Objective-C, but it would require more logic to ensure we do not
229 // issue compatibility diagnostics in these cases.
230 if (!LangOpts.ObjC && IsKeywordInCpp(Flags))
231 MarkIdentifierAsKeywordInCpp(Table, Name: Keyword);
232 return;
233 }
234
235 IdentifierInfo &Info =
236 Table.get(Name: Keyword, TokenCode: AddResult == KS_Future ? tok::identifier : TokenCode);
237 Info.setIsExtensionToken(AddResult == KS_Extension);
238 Info.setIsFutureCompatKeyword(AddResult == KS_Future);
239}
240
241/// AddCXXOperatorKeyword - Register a C++ operator keyword alternative
242/// representations.
243static void AddCXXOperatorKeyword(StringRef Keyword,
244 tok::TokenKind TokenCode,
245 IdentifierTable &Table) {
246 IdentifierInfo &Info = Table.get(Name: Keyword, TokenCode);
247 Info.setIsCPlusPlusOperatorKeyword();
248}
249
250/// AddObjCKeyword - Register an Objective-C \@keyword like "class" "selector"
251/// or "property".
252static void AddObjCKeyword(StringRef Name,
253 tok::ObjCKeywordKind ObjCID,
254 IdentifierTable &Table) {
255 Table.get(Name).setObjCKeywordID(ObjCID);
256}
257
258static void AddNotableIdentifier(StringRef Name,
259 tok::NotableIdentifierKind BTID,
260 IdentifierTable &Table) {
261 // Don't add 'not_notable' identifier.
262 if (BTID != tok::not_notable) {
263 IdentifierInfo &Info = Table.get(Name, TokenCode: tok::identifier);
264 Info.setNotableIdentifierID(BTID);
265 }
266}
267
268/// AddKeywords - Add all keywords to the symbol table.
269///
270void IdentifierTable::AddKeywords(const LangOptions &LangOpts) {
271 // Add keywords and tokens for the current language.
272#define KEYWORD(NAME, FLAGS) \
273 AddKeyword(StringRef(#NAME), tok::kw_ ## NAME, \
274 FLAGS, LangOpts, *this);
275#define ALIAS(NAME, TOK, FLAGS) \
276 AddKeyword(StringRef(NAME), tok::kw_ ## TOK, \
277 FLAGS, LangOpts, *this);
278#define CXX_KEYWORD_OPERATOR(NAME, ALIAS) \
279 if (LangOpts.CXXOperatorNames) \
280 AddCXXOperatorKeyword(StringRef(#NAME), tok::ALIAS, *this); \
281 else \
282 MarkIdentifierAsKeywordInCpp(*this, StringRef(#NAME));
283#define OBJC_AT_KEYWORD(NAME) \
284 if (LangOpts.ObjC) \
285 AddObjCKeyword(StringRef(#NAME), tok::objc_##NAME, *this);
286#define NOTABLE_IDENTIFIER(NAME) \
287 AddNotableIdentifier(StringRef(#NAME), tok::NAME, *this);
288
289#define TESTING_KEYWORD(NAME, FLAGS)
290#include "clang/Basic/TokenKinds.def"
291
292 if (LangOpts.ParseUnknownAnytype)
293 AddKeyword(Keyword: "__unknown_anytype", TokenCode: tok::kw___unknown_anytype, Flags: KEYALL,
294 LangOpts, Table&: *this);
295
296 if (LangOpts.DeclSpecKeyword)
297 AddKeyword(Keyword: "__declspec", TokenCode: tok::kw___declspec, Flags: KEYALL, LangOpts, Table&: *this);
298
299 if (LangOpts.IEEE128)
300 AddKeyword(Keyword: "__ieee128", TokenCode: tok::kw___float128, Flags: KEYALL, LangOpts, Table&: *this);
301
302 // Add the 'import' and 'module' contextual keywords.
303 get(Name: "import").setKeywordImport(true);
304 get(Name: "module").setModuleKeyword(true);
305 get(Name: "__preprocessed_import").setKeywordImport(true);
306 get(Name: "__preprocessed_module").setModuleKeyword(true);
307}
308
309/// Checks if the specified token kind represents a keyword in the
310/// specified language.
311/// \returns Status of the keyword in the language.
312static KeywordStatus getTokenKwStatus(const LangOptions &LangOpts,
313 tok::TokenKind K) {
314 switch (K) {
315#define KEYWORD(NAME, FLAGS) \
316 case tok::kw_##NAME: return getKeywordStatus(LangOpts, FLAGS);
317#include "clang/Basic/TokenKinds.def"
318 default: return KS_Disabled;
319 }
320}
321
322/// Returns true if the identifier represents a keyword in the
323/// specified language.
324bool IdentifierInfo::isKeyword(const LangOptions &LangOpts) const {
325 switch (getTokenKwStatus(LangOpts, K: getTokenID())) {
326 case KS_Enabled:
327 case KS_Extension:
328 return true;
329 default:
330 return false;
331 }
332}
333
334/// Returns true if the identifier represents a C++ keyword in the
335/// specified language.
336bool IdentifierInfo::isCPlusPlusKeyword(const LangOptions &LangOpts) const {
337 if (!LangOpts.CPlusPlus || !isKeyword(LangOpts))
338 return false;
339 // This is a C++ keyword if this identifier is not a keyword when checked
340 // using LangOptions without C++ support.
341 LangOptions LangOptsNoCPP = LangOpts;
342 LangOptsNoCPP.CPlusPlus = false;
343 LangOptsNoCPP.CPlusPlus11 = false;
344 LangOptsNoCPP.CPlusPlus20 = false;
345 return !isKeyword(LangOpts: LangOptsNoCPP);
346}
347
348ReservedIdentifierStatus
349IdentifierInfo::isReserved(const LangOptions &LangOpts) const {
350 StringRef Name = getName();
351
352 // '_' is a reserved identifier, but its use is so common (e.g. to store
353 // ignored values) that we don't warn on it.
354 if (Name.size() <= 1)
355 return ReservedIdentifierStatus::NotReserved;
356
357 // [lex.name] p3
358 if (Name[0] == '_') {
359
360 // Each name that begins with an underscore followed by an uppercase letter
361 // or another underscore is reserved.
362 if (Name[1] == '_')
363 return ReservedIdentifierStatus::StartsWithDoubleUnderscore;
364
365 if ('A' <= Name[1] && Name[1] <= 'Z')
366 return ReservedIdentifierStatus::
367 StartsWithUnderscoreFollowedByCapitalLetter;
368
369 // This is a bit misleading: it actually means it's only reserved if we're
370 // at global scope because it starts with an underscore.
371 return ReservedIdentifierStatus::StartsWithUnderscoreAtGlobalScope;
372 }
373
374 // Each name that contains a double underscore (__) is reserved.
375 if (LangOpts.CPlusPlus && Name.contains(Other: "__"))
376 return ReservedIdentifierStatus::ContainsDoubleUnderscore;
377
378 return ReservedIdentifierStatus::NotReserved;
379}
380
381ReservedLiteralSuffixIdStatus
382IdentifierInfo::isReservedLiteralSuffixId() const {
383 StringRef Name = getName();
384
385 // Note: the diag::warn_deprecated_literal_operator_id diagnostic depends on
386 // this being the first check we do, so if this order changes, we have to fix
387 // that as well.
388 if (Name[0] != '_')
389 return ReservedLiteralSuffixIdStatus::NotStartsWithUnderscore;
390
391 if (Name.contains(Other: "__"))
392 return ReservedLiteralSuffixIdStatus::ContainsDoubleUnderscore;
393
394 return ReservedLiteralSuffixIdStatus::NotReserved;
395}
396
397StringRef IdentifierInfo::deuglifiedName() const {
398 StringRef Name = getName();
399 if (Name.size() >= 2 && Name.front() == '_' &&
400 (Name[1] == '_' || (Name[1] >= 'A' && Name[1] <= 'Z')))
401 return Name.ltrim(Char: '_');
402 return Name;
403}
404
405tok::PPKeywordKind IdentifierInfo::getPPKeywordID() const {
406 // We use a perfect hash function here involving the length of the keyword,
407 // the first and third character. For preprocessor ID's there are no
408 // collisions (if there were, the switch below would complain about duplicate
409 // case values). Note that this depends on 'if' being null terminated.
410
411#define HASH(LEN, FIRST, THIRD) \
412 (LEN << 6) + (((FIRST - 'a') - (THIRD - 'a')) & 63)
413#define CASE(LEN, FIRST, THIRD, NAME) \
414 case HASH(LEN, FIRST, THIRD): \
415 return memcmp(Name, #NAME, LEN) ? tok::pp_not_keyword : tok::pp_ ## NAME
416
417 unsigned Len = getLength();
418 if (Len < 2) return tok::pp_not_keyword;
419 const char *Name = getNameStart();
420
421 if (Name[0] == '_' && isImportKeyword())
422 return tok::pp___preprocessed_import;
423 if (Name[0] == '_' && isModuleKeyword())
424 return tok::pp___preprocessed_module;
425
426 // clang-format off
427 switch (HASH(Len, Name[0], Name[2])) {
428 default: return tok::pp_not_keyword;
429 CASE( 2, 'i', '\0', if);
430 CASE( 4, 'e', 'i', elif);
431 CASE( 4, 'e', 's', else);
432 CASE( 4, 'l', 'n', line);
433 CASE( 4, 's', 'c', sccs);
434 CASE( 5, 'e', 'b', embed);
435 CASE( 5, 'e', 'd', endif);
436 CASE( 5, 'e', 'r', error);
437 CASE( 5, 'i', 'e', ident);
438 CASE( 5, 'i', 'd', ifdef);
439 CASE( 5, 'u', 'd', undef);
440
441 CASE( 6, 'a', 's', assert);
442 CASE( 6, 'd', 'f', define);
443 CASE( 6, 'i', 'n', ifndef);
444 CASE( 6, 'i', 'p', import);
445 CASE( 6, 'm', 'd', module);
446 CASE( 6, 'p', 'a', pragma);
447
448 CASE( 7, 'd', 'f', defined);
449 CASE( 7, 'e', 'i', elifdef);
450 CASE( 7, 'i', 'c', include);
451 CASE( 7, 'w', 'r', warning);
452
453 CASE( 8, 'e', 'i', elifndef);
454 CASE( 8, 'u', 'a', unassert);
455 CASE(12, 'i', 'c', include_next);
456
457 CASE(14, '_', 'p', __public_macro);
458
459 CASE(15, '_', 'p', __private_macro);
460
461 CASE(16, '_', 'i', __include_macros);
462#undef CASE
463#undef HASH
464 }
465 // clang-format on
466}
467
468//===----------------------------------------------------------------------===//
469// Stats Implementation
470//===----------------------------------------------------------------------===//
471
472/// PrintStats - Print statistics about how well the identifier table is doing
473/// at hashing identifiers.
474void IdentifierTable::PrintStats() const {
475 unsigned NumBuckets = HashTable.getNumBuckets();
476 unsigned NumIdentifiers = HashTable.getNumItems();
477 unsigned NumEmptyBuckets = NumBuckets-NumIdentifiers;
478 unsigned AverageIdentifierSize = 0;
479 unsigned MaxIdentifierLength = 0;
480
481 // TODO: Figure out maximum times an identifier had to probe for -stats.
482 for (llvm::StringMap<IdentifierInfo*, llvm::BumpPtrAllocator>::const_iterator
483 I = HashTable.begin(), E = HashTable.end(); I != E; ++I) {
484 unsigned IdLen = I->getKeyLength();
485 AverageIdentifierSize += IdLen;
486 if (MaxIdentifierLength < IdLen)
487 MaxIdentifierLength = IdLen;
488 }
489
490 fprintf(stderr, format: "\n*** Identifier Table Stats:\n");
491 fprintf(stderr, format: "# Identifiers: %d\n", NumIdentifiers);
492 fprintf(stderr, format: "# Empty Buckets: %d\n", NumEmptyBuckets);
493 fprintf(stderr, format: "Hash density (#identifiers per bucket): %f\n",
494 NumIdentifiers/(double)NumBuckets);
495 fprintf(stderr, format: "Ave identifier length: %f\n",
496 (AverageIdentifierSize/(double)NumIdentifiers));
497 fprintf(stderr, format: "Max identifier length: %d\n", MaxIdentifierLength);
498
499 // Compute statistics about the memory allocated for identifiers.
500 HashTable.getAllocator().PrintStats();
501}
502
503//===----------------------------------------------------------------------===//
504// SelectorTable Implementation
505//===----------------------------------------------------------------------===//
506
507unsigned llvm::DenseMapInfo<clang::Selector>::getHashValue(clang::Selector S) {
508 return DenseMapInfo<void*>::getHashValue(PtrVal: S.getAsOpaquePtr());
509}
510
511bool Selector::isKeywordSelector(ArrayRef<StringRef> Names) const {
512 assert(!Names.empty() && "must have >= 1 selector slots");
513 if (getNumArgs() != Names.size())
514 return false;
515 for (unsigned I = 0, E = Names.size(); I != E; ++I) {
516 if (getNameForSlot(argIndex: I) != Names[I])
517 return false;
518 }
519 return true;
520}
521
522bool Selector::isUnarySelector(StringRef Name) const {
523 return isUnarySelector() && getNameForSlot(argIndex: 0) == Name;
524}
525
526unsigned Selector::getNumArgs() const {
527 unsigned IIF = getIdentifierInfoFlag();
528 if (IIF <= ZeroArg)
529 return 0;
530 if (IIF == OneArg)
531 return 1;
532 // We point to a MultiKeywordSelector.
533 MultiKeywordSelector *SI = getMultiKeywordSelector();
534 return SI->getNumArgs();
535}
536
537const IdentifierInfo *
538Selector::getIdentifierInfoForSlot(unsigned argIndex) const {
539 if (getIdentifierInfoFlag() < MultiArg) {
540 assert(argIndex == 0 && "illegal keyword index");
541 return getAsIdentifierInfo();
542 }
543
544 // We point to a MultiKeywordSelector.
545 MultiKeywordSelector *SI = getMultiKeywordSelector();
546 return SI->getIdentifierInfoForSlot(i: argIndex);
547}
548
549StringRef Selector::getNameForSlot(unsigned int argIndex) const {
550 const IdentifierInfo *II = getIdentifierInfoForSlot(argIndex);
551 return II ? II->getName() : StringRef();
552}
553
554std::string MultiKeywordSelector::getName() const {
555 SmallString<256> Str;
556 llvm::raw_svector_ostream OS(Str);
557 for (keyword_iterator I = keyword_begin(), E = keyword_end(); I != E; ++I) {
558 if (*I)
559 OS << (*I)->getName();
560 OS << ':';
561 }
562
563 return std::string(OS.str());
564}
565
566std::string Selector::getAsString() const {
567 if (isNull())
568 return "<null selector>";
569
570 if (getIdentifierInfoFlag() < MultiArg) {
571 const IdentifierInfo *II = getAsIdentifierInfo();
572
573 if (getNumArgs() == 0) {
574 assert(II && "If the number of arguments is 0 then II is guaranteed to "
575 "not be null.");
576 return std::string(II->getName());
577 }
578
579 if (!II)
580 return ":";
581
582 return II->getName().str() + ":";
583 }
584
585 // We have a multiple keyword selector.
586 return getMultiKeywordSelector()->getName();
587}
588
589void Selector::print(llvm::raw_ostream &OS) const {
590 OS << getAsString();
591}
592
593LLVM_DUMP_METHOD void Selector::dump() const { print(OS&: llvm::errs()); }
594
595/// Interpreting the given string using the normal CamelCase
596/// conventions, determine whether the given string starts with the
597/// given "word", which is assumed to end in a lowercase letter.
598static bool startsWithWord(StringRef name, StringRef word) {
599 if (name.size() < word.size()) return false;
600 return ((name.size() == word.size() || !isLowercase(c: name[word.size()])) &&
601 name.starts_with(Prefix: word));
602}
603
604ObjCMethodFamily Selector::getMethodFamilyImpl(Selector sel) {
605 const IdentifierInfo *first = sel.getIdentifierInfoForSlot(argIndex: 0);
606 if (!first) return OMF_None;
607
608 StringRef name = first->getName();
609 if (sel.isUnarySelector()) {
610 if (name == "autorelease") return OMF_autorelease;
611 if (name == "dealloc") return OMF_dealloc;
612 if (name == "finalize") return OMF_finalize;
613 if (name == "release") return OMF_release;
614 if (name == "retain") return OMF_retain;
615 if (name == "retainCount") return OMF_retainCount;
616 if (name == "self") return OMF_self;
617 if (name == "initialize") return OMF_initialize;
618 }
619
620 if (name == "performSelector" || name == "performSelectorInBackground" ||
621 name == "performSelectorOnMainThread")
622 return OMF_performSelector;
623
624 // The other method families may begin with a prefix of underscores.
625 name = name.ltrim(Char: '_');
626
627 if (name.empty()) return OMF_None;
628 switch (name.front()) {
629 case 'a':
630 if (startsWithWord(name, word: "alloc")) return OMF_alloc;
631 break;
632 case 'c':
633 if (startsWithWord(name, word: "copy")) return OMF_copy;
634 break;
635 case 'i':
636 if (startsWithWord(name, word: "init")) return OMF_init;
637 break;
638 case 'm':
639 if (startsWithWord(name, word: "mutableCopy")) return OMF_mutableCopy;
640 break;
641 case 'n':
642 if (startsWithWord(name, word: "new")) return OMF_new;
643 break;
644 default:
645 break;
646 }
647
648 return OMF_None;
649}
650
651ObjCInstanceTypeFamily Selector::getInstTypeMethodFamily(Selector sel) {
652 const IdentifierInfo *first = sel.getIdentifierInfoForSlot(argIndex: 0);
653 if (!first) return OIT_None;
654
655 StringRef name = first->getName();
656
657 if (name.empty()) return OIT_None;
658 switch (name.front()) {
659 case 'a':
660 if (startsWithWord(name, word: "array")) return OIT_Array;
661 break;
662 case 'd':
663 if (startsWithWord(name, word: "default")) return OIT_ReturnsSelf;
664 if (startsWithWord(name, word: "dictionary")) return OIT_Dictionary;
665 break;
666 case 's':
667 if (startsWithWord(name, word: "shared")) return OIT_ReturnsSelf;
668 if (startsWithWord(name, word: "standard")) return OIT_Singleton;
669 break;
670 case 'i':
671 if (startsWithWord(name, word: "init")) return OIT_Init;
672 break;
673 default:
674 break;
675 }
676 return OIT_None;
677}
678
679ObjCStringFormatFamily Selector::getStringFormatFamilyImpl(Selector sel) {
680 const IdentifierInfo *first = sel.getIdentifierInfoForSlot(argIndex: 0);
681 if (!first) return SFF_None;
682
683 StringRef name = first->getName();
684
685 switch (name.front()) {
686 case 'a':
687 if (name == "appendFormat") return SFF_NSString;
688 break;
689
690 case 'i':
691 if (name == "initWithFormat") return SFF_NSString;
692 break;
693
694 case 'l':
695 if (name == "localizedStringWithFormat") return SFF_NSString;
696 break;
697
698 case 's':
699 if (name == "stringByAppendingFormat" ||
700 name == "stringWithFormat") return SFF_NSString;
701 break;
702 }
703 return SFF_None;
704}
705
706namespace {
707
708struct SelectorTableImpl {
709 llvm::UniquingSet<MultiKeywordSelector> Table;
710 llvm::BumpPtrAllocator Allocator;
711};
712
713} // namespace
714
715static SelectorTableImpl &getSelectorTableImpl(void *P) {
716 return *static_cast<SelectorTableImpl*>(P);
717}
718
719SmallString<64>
720SelectorTable::constructSetterName(StringRef Name) {
721 SmallString<64> SetterName("set");
722 SetterName += Name;
723 SetterName[3] = toUppercase(c: SetterName[3]);
724 return SetterName;
725}
726
727Selector
728SelectorTable::constructSetterSelector(IdentifierTable &Idents,
729 SelectorTable &SelTable,
730 const IdentifierInfo *Name) {
731 IdentifierInfo *SetterName =
732 &Idents.get(Name: constructSetterName(Name: Name->getName()));
733 return SelTable.getUnarySelector(ID: SetterName);
734}
735
736std::string SelectorTable::getPropertyNameFromSetterSelector(Selector Sel) {
737 StringRef Name = Sel.getNameForSlot(argIndex: 0);
738 assert(Name.starts_with("set") && "invalid setter name");
739 return (Twine(toLowercase(c: Name[3])) + Name.drop_front(N: 4)).str();
740}
741
742size_t SelectorTable::getTotalMemory() const {
743 SelectorTableImpl &SelTabImpl = getSelectorTableImpl(P: Impl);
744 return SelTabImpl.Allocator.getTotalMemory();
745}
746
747Selector SelectorTable::getSelector(unsigned nKeys,
748 const IdentifierInfo **IIV) {
749 if (nKeys < 2)
750 return Selector(IIV[0], nKeys);
751
752 SelectorTableImpl &SelTabImpl = getSelectorTableImpl(P: Impl);
753
754 llvm::FoldingSetInsertToken InsertToken;
755 if (MultiKeywordSelector *SI =
756 SelTabImpl.Table.lookup(Key: ArrayRef(IIV, nKeys), Token&: InsertToken))
757 return Selector(SI);
758
759 // MultiKeywordSelector objects are not allocated with new because they have a
760 // variable size array (for parameter types) at the end of them.
761 unsigned Size = sizeof(MultiKeywordSelector) + nKeys*sizeof(IdentifierInfo *);
762 MultiKeywordSelector *SI =
763 (MultiKeywordSelector *)SelTabImpl.Allocator.Allocate(
764 Size, Alignment: alignof(MultiKeywordSelector));
765 new (SI) MultiKeywordSelector(nKeys, IIV);
766 SelTabImpl.Table.insert(N: SI, Token: InsertToken);
767 return Selector(SI);
768}
769
770SelectorTable::SelectorTable() {
771 Impl = new SelectorTableImpl();
772}
773
774SelectorTable::~SelectorTable() {
775 delete &getSelectorTableImpl(P: Impl);
776}
777
778const char *clang::getOperatorSpelling(OverloadedOperatorKind Operator) {
779 switch (Operator) {
780 case OO_None:
781 case NUM_OVERLOADED_OPERATORS:
782 return nullptr;
783
784#define OVERLOADED_OPERATOR(Name,Spelling,Token,Unary,Binary,MemberOnly) \
785 case OO_##Name: return Spelling;
786#include "clang/Basic/OperatorKinds.def"
787 }
788
789 llvm_unreachable("Invalid OverloadedOperatorKind!");
790}
791
792StringRef clang::getNullabilitySpelling(NullabilityKind kind,
793 bool isContextSensitive) {
794 switch (kind) {
795 case NullabilityKind::NonNull:
796 return isContextSensitive ? "nonnull" : "_Nonnull";
797
798 case NullabilityKind::Nullable:
799 return isContextSensitive ? "nullable" : "_Nullable";
800
801 case NullabilityKind::NullableResult:
802 assert(!isContextSensitive &&
803 "_Nullable_result isn't supported as context-sensitive keyword");
804 return "_Nullable_result";
805
806 case NullabilityKind::Unspecified:
807 return isContextSensitive ? "null_unspecified" : "_Null_unspecified";
808 }
809 llvm_unreachable("Unknown nullability kind.");
810}
811
812llvm::raw_ostream &clang::operator<<(llvm::raw_ostream &OS,
813 NullabilityKind NK) {
814 switch (NK) {
815 case NullabilityKind::NonNull:
816 return OS << "NonNull";
817 case NullabilityKind::Nullable:
818 return OS << "Nullable";
819 case NullabilityKind::NullableResult:
820 return OS << "NullableResult";
821 case NullabilityKind::Unspecified:
822 return OS << "Unspecified";
823 }
824 llvm_unreachable("Unknown nullability kind.");
825}
826
827diag::kind
828IdentifierTable::getFutureCompatDiagKind(const IdentifierInfo &II,
829 const LangOptions &LangOpts) {
830 assert(II.isFutureCompatKeyword() && "diagnostic should not be needed");
831
832 constexpr llvm::EnumStringDef<uint32_t> FlagsDefs[] = {
833#define KEYWORD(NAME, FLAGS) {{#NAME}, FLAGS},
834#include "clang/Basic/TokenKinds.def"
835#undef KEYWORD
836 };
837 static constexpr auto FlagsData = BUILD_ENUM_STRINGS(FlagsDefs);
838 const auto *Entry = llvm::EnumStrings(FlagsData).fromString(Str: II.getName());
839 assert(Entry && "future compat keyword should be in keyword list");
840 uint32_t Flags = Entry->value();
841
842 if (LangOpts.CPlusPlus) {
843 if ((Flags & KEYCXX11) == KEYCXX11)
844 return diag::warn_cxx11_keyword;
845
846 // char8_t is not modeled as a CXX20_KEYWORD because it's not
847 // unconditionally enabled in C++20 mode. (It can be disabled
848 // by -fno-char8_t.)
849 if (((Flags & KEYCXX20) == KEYCXX20) ||
850 ((Flags & CHAR8SUPPORT) == CHAR8SUPPORT))
851 return diag::warn_cxx20_keyword;
852 } else {
853 if ((Flags & KEYC99) == KEYC99)
854 return diag::warn_c99_keyword;
855 if ((Flags & KEYC23) == KEYC23)
856 return diag::warn_c23_keyword;
857 }
858
859 llvm_unreachable(
860 "Keyword not known to come from a newer Standard or proposed Standard");
861}
862