1//===--- DumpAST.cpp - Serialize clang AST to LSP -------------------------===//
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#include "DumpAST.h"
10#include "Protocol.h"
11#include "SourceCode.h"
12#include "support/Logger.h"
13#include "clang/AST/ASTTypeTraits.h"
14#include "clang/AST/Expr.h"
15#include "clang/AST/ExprCXX.h"
16#include "clang/AST/NestedNameSpecifier.h"
17#include "clang/AST/PrettyPrinter.h"
18#include "clang/AST/RecursiveASTVisitor.h"
19#include "clang/AST/TextNodeDumper.h"
20#include "clang/AST/Type.h"
21#include "clang/AST/TypeLoc.h"
22#include "clang/Basic/Specifiers.h"
23#include "clang/Tooling/Syntax/Tokens.h"
24#include "llvm/ADT/StringRef.h"
25#include "llvm/Support/raw_ostream.h"
26#include <optional>
27
28namespace clang {
29namespace clangd {
30namespace {
31
32using llvm::raw_ostream;
33template <typename Print> std::string toString(const Print &C) {
34 std::string Result;
35 llvm::raw_string_ostream OS(Result);
36 C(OS);
37 return std::move(OS.str());
38}
39
40bool isInjectedClassName(Decl *D) {
41 if (const auto *CRD = llvm::dyn_cast<CXXRecordDecl>(D))
42 return CRD->isInjectedClassName();
43 return false;
44}
45
46class DumpVisitor : public RecursiveASTVisitor<DumpVisitor> {
47 using Base = RecursiveASTVisitor<DumpVisitor>;
48
49 const syntax::TokenBuffer &Tokens;
50 const ASTContext &Ctx;
51
52 // Pointers are into 'children' vector.
53 // They remain valid because while a node is on the stack we only add
54 // descendants, not siblings.
55 std::vector<ASTNode *> Stack;
56
57 // Generic logic used to handle traversal of all node kinds.
58
59 template <typename T>
60 bool traverseNodePre(llvm::StringRef Role, const T &Node) {
61 if (Stack.empty()) {
62 assert(Root.role.empty());
63 Stack.push_back(&Root);
64 } else {
65 Stack.back()->children.emplace_back();
66 Stack.push_back(&Stack.back()->children.back());
67 }
68 auto &N = *Stack.back();
69 N.role = Role.str();
70 N.kind = getKind(Node);
71 N.detail = getDetail(Node);
72 N.range = getRange(Node);
73 N.arcana = getArcana(Node);
74 return true;
75 }
76 bool traverseNodePost() {
77 assert(!Stack.empty());
78 Stack.pop_back();
79 return true;
80 }
81 template <typename T, typename Callable>
82 bool traverseNode(llvm::StringRef Role, const T &Node, const Callable &Body) {
83 traverseNodePre(Role, Node);
84 Body();
85 return traverseNodePost();
86 }
87
88 // Range: most nodes have getSourceRange(), with a couple of exceptions.
89 // We only return it if it's valid at both ends and there are no macros.
90
91 template <typename T> std::optional<Range> getRange(const T &Node) {
92 SourceRange SR = getSourceRange(Node);
93 auto Spelled = Tokens.spelledForExpanded(Tokens.expandedTokens(SR));
94 if (!Spelled)
95 return std::nullopt;
96 return halfOpenToRange(
97 Tokens.sourceManager(),
98 CharSourceRange::getCharRange(Spelled->front().location(),
99 Spelled->back().endLocation()));
100 }
101 template <typename T, typename = decltype(std::declval<T>().getSourceRange())>
102 SourceRange getSourceRange(const T &Node) {
103 return Node.getSourceRange();
104 }
105 template <typename T,
106 typename = decltype(std::declval<T *>()->getSourceRange())>
107 SourceRange getSourceRange(const T *Node) {
108 return Node->getSourceRange();
109 }
110 // TemplateName doesn't have a real Loc node type.
111 SourceRange getSourceRange(const TemplateName &Node) { return SourceRange(); }
112 // Attr just uses a weird method name. Maybe we should fix it instead?
113 SourceRange getSourceRange(const Attr *Node) { return Node->getRange(); }
114
115 // Kind is usually the class name, without the suffix ("Type" etc).
116 // Where there's a set of variants instead, we use the 'Kind' enum values.
117
118 std::string getKind(const Decl *D) { return D->getDeclKindName(); }
119 std::string getKind(const Stmt *S) {
120 std::string Result = S->getStmtClassName();
121 if (llvm::StringRef(Result).endswith("Stmt") ||
122 llvm::StringRef(Result).endswith("Expr"))
123 Result.resize(Result.size() - 4);
124 return Result;
125 }
126 std::string getKind(const TypeLoc &TL) {
127 std::string Result;
128 if (TL.getTypeLocClass() == TypeLoc::Qualified)
129 return "Qualified";
130 return TL.getType()->getTypeClassName();
131 }
132 std::string getKind(const TemplateArgumentLoc &TAL) {
133 switch (TAL.getArgument().getKind()) {
134#define TEMPLATE_ARGUMENT_KIND(X) \
135 case TemplateArgument::X: \
136 return #X
137 TEMPLATE_ARGUMENT_KIND(Null);
138 TEMPLATE_ARGUMENT_KIND(NullPtr);
139 TEMPLATE_ARGUMENT_KIND(Expression);
140 TEMPLATE_ARGUMENT_KIND(Integral);
141 TEMPLATE_ARGUMENT_KIND(Pack);
142 TEMPLATE_ARGUMENT_KIND(Type);
143 TEMPLATE_ARGUMENT_KIND(Declaration);
144 TEMPLATE_ARGUMENT_KIND(Template);
145 TEMPLATE_ARGUMENT_KIND(TemplateExpansion);
146#undef TEMPLATE_ARGUMENT_KIND
147 }
148 llvm_unreachable("Unhandled ArgKind enum");
149 }
150 std::string getKind(const NestedNameSpecifierLoc &NNSL) {
151 assert(NNSL.getNestedNameSpecifier());
152 switch (NNSL.getNestedNameSpecifier()->getKind()) {
153#define NNS_KIND(X) \
154 case NestedNameSpecifier::X: \
155 return #X
156 NNS_KIND(Identifier);
157 NNS_KIND(Namespace);
158 NNS_KIND(TypeSpec);
159 NNS_KIND(TypeSpecWithTemplate);
160 NNS_KIND(Global);
161 NNS_KIND(Super);
162 NNS_KIND(NamespaceAlias);
163#undef NNS_KIND
164 }
165 llvm_unreachable("Unhandled SpecifierKind enum");
166 }
167 std::string getKind(const CXXCtorInitializer *CCI) {
168 if (CCI->isBaseInitializer())
169 return "BaseInitializer";
170 if (CCI->isDelegatingInitializer())
171 return "DelegatingInitializer";
172 if (CCI->isAnyMemberInitializer())
173 return "MemberInitializer";
174 llvm_unreachable("Unhandled CXXCtorInitializer type");
175 }
176 std::string getKind(const TemplateName &TN) {
177 switch (TN.getKind()) {
178#define TEMPLATE_KIND(X) \
179 case TemplateName::X: \
180 return #X;
181 TEMPLATE_KIND(Template);
182 TEMPLATE_KIND(OverloadedTemplate);
183 TEMPLATE_KIND(AssumedTemplate);
184 TEMPLATE_KIND(QualifiedTemplate);
185 TEMPLATE_KIND(DependentTemplate);
186 TEMPLATE_KIND(SubstTemplateTemplateParm);
187 TEMPLATE_KIND(SubstTemplateTemplateParmPack);
188 TEMPLATE_KIND(UsingTemplate);
189#undef TEMPLATE_KIND
190 }
191 llvm_unreachable("Unhandled NameKind enum");
192 }
193 std::string getKind(const Attr *A) {
194 switch (A->getKind()) {
195#define ATTR(X) \
196 case attr::X: \
197 return #X;
198#include "clang/Basic/AttrList.inc"
199#undef ATTR
200 }
201 llvm_unreachable("Unhandled attr::Kind enum");
202 }
203 std::string getKind(const CXXBaseSpecifier &CBS) {
204 // There aren't really any variants of CXXBaseSpecifier.
205 // To avoid special cases in the API/UI, use public/private as the kind.
206 return getAccessSpelling(CBS.getAccessSpecifier()).str();
207 }
208
209 // Detail is the single most important fact about the node.
210 // Often this is the name, sometimes a "kind" enum like operators or casts.
211 // We should avoid unbounded text, like dumping parameter lists.
212
213 std::string getDetail(const Decl *D) {
214 const auto *ND = dyn_cast<NamedDecl>(D);
215 if (!ND || llvm::isa_and_nonnull<CXXConstructorDecl>(ND->getAsFunction()) ||
216 isa<CXXDestructorDecl>(ND))
217 return "";
218 std::string Name = toString([&](raw_ostream &OS) { ND->printName(OS); });
219 if (Name.empty())
220 return "(anonymous)";
221 return Name;
222 }
223 std::string getDetail(const Stmt *S) {
224 if (const auto *DRE = dyn_cast<DeclRefExpr>(S))
225 return DRE->getNameInfo().getAsString();
226 if (const auto *DSDRE = dyn_cast<DependentScopeDeclRefExpr>(S))
227 return DSDRE->getNameInfo().getAsString();
228 if (const auto *ME = dyn_cast<MemberExpr>(S))
229 return ME->getMemberNameInfo().getAsString();
230 if (const auto *CE = dyn_cast<CastExpr>(S))
231 return CE->getCastKindName();
232 if (const auto *BO = dyn_cast<BinaryOperator>(S))
233 return BO->getOpcodeStr().str();
234 if (const auto *UO = dyn_cast<UnaryOperator>(S))
235 return UnaryOperator::getOpcodeStr(UO->getOpcode()).str();
236 if (const auto *CCO = dyn_cast<CXXConstructExpr>(S))
237 return CCO->getConstructor()->getNameAsString();
238 if (const auto *CTE = dyn_cast<CXXThisExpr>(S)) {
239 bool Const = CTE->getType()->getPointeeType().isLocalConstQualified();
240 if (CTE->isImplicit())
241 return Const ? "const, implicit" : "implicit";
242 if (Const)
243 return "const";
244 return "";
245 }
246 if (isa<IntegerLiteral, FloatingLiteral, FixedPointLiteral,
247 CharacterLiteral, ImaginaryLiteral, CXXBoolLiteralExpr>(S))
248 return toString([&](raw_ostream &OS) {
249 S->printPretty(OS, nullptr, Ctx.getPrintingPolicy());
250 });
251 if (const auto *MTE = dyn_cast<MaterializeTemporaryExpr>(S))
252 return MTE->isBoundToLvalueReference() ? "lvalue" : "rvalue";
253 return "";
254 }
255 std::string getDetail(const TypeLoc &TL) {
256 if (TL.getType().hasLocalQualifiers())
257 return TL.getType().getLocalQualifiers().getAsString(
258 Ctx.getPrintingPolicy());
259 if (const auto *TT = dyn_cast<TagType>(TL.getTypePtr()))
260 return getDetail(TT->getDecl());
261 if (const auto *DT = dyn_cast<DeducedType>(TL.getTypePtr()))
262 if (DT->isDeduced())
263 return DT->getDeducedType().getAsString(Ctx.getPrintingPolicy());
264 if (const auto *BT = dyn_cast<BuiltinType>(TL.getTypePtr()))
265 return BT->getName(Ctx.getPrintingPolicy()).str();
266 if (const auto *TTPT = dyn_cast<TemplateTypeParmType>(TL.getTypePtr()))
267 return getDetail(TTPT->getDecl());
268 if (const auto *TT = dyn_cast<TypedefType>(TL.getTypePtr()))
269 return getDetail(TT->getDecl());
270 return "";
271 }
272 std::string getDetail(const NestedNameSpecifierLoc &NNSL) {
273 const auto &NNS = *NNSL.getNestedNameSpecifier();
274 switch (NNS.getKind()) {
275 case NestedNameSpecifier::Identifier:
276 return NNS.getAsIdentifier()->getName().str() + "::";
277 case NestedNameSpecifier::Namespace:
278 return NNS.getAsNamespace()->getNameAsString() + "::";
279 case NestedNameSpecifier::NamespaceAlias:
280 return NNS.getAsNamespaceAlias()->getNameAsString() + "::";
281 default:
282 return "";
283 }
284 }
285 std::string getDetail(const CXXCtorInitializer *CCI) {
286 if (FieldDecl *FD = CCI->getAnyMember())
287 return getDetail(FD);
288 if (TypeLoc TL = CCI->getBaseClassLoc())
289 return getDetail(TL);
290 return "";
291 }
292 std::string getDetail(const TemplateArgumentLoc &TAL) {
293 if (TAL.getArgument().getKind() == TemplateArgument::Integral)
294 return toString(TAL.getArgument().getAsIntegral(), 10);
295 return "";
296 }
297 std::string getDetail(const TemplateName &TN) {
298 return toString([&](raw_ostream &OS) {
299 TN.print(OS, Ctx.getPrintingPolicy(), TemplateName::Qualified::None);
300 });
301 }
302 std::string getDetail(const Attr *A) {
303 return A->getAttrName() ? A->getNormalizedFullName() : A->getSpelling();
304 }
305 std::string getDetail(const CXXBaseSpecifier &CBS) {
306 return CBS.isVirtual() ? "virtual" : "";
307 }
308
309 /// Arcana is produced by TextNodeDumper, for the types it supports.
310
311 template <typename Dump> std::string dump(const Dump &D) {
312 return toString([&](raw_ostream &OS) {
313 TextNodeDumper Dumper(OS, Ctx, /*ShowColors=*/false);
314 D(Dumper);
315 });
316 }
317 template <typename T> std::string getArcana(const T &N) {
318 return dump([&](TextNodeDumper &D) { D.Visit(N); });
319 }
320 std::string getArcana(const NestedNameSpecifierLoc &NNS) { return ""; }
321 std::string getArcana(const TemplateName &NNS) { return ""; }
322 std::string getArcana(const CXXBaseSpecifier &CBS) { return ""; }
323 std::string getArcana(const TemplateArgumentLoc &TAL) {
324 return dump([&](TextNodeDumper &D) {
325 D.Visit(TAL.getArgument(), TAL.getSourceRange());
326 });
327 }
328 std::string getArcana(const TypeLoc &TL) {
329 return dump([&](TextNodeDumper &D) { D.Visit(TL.getType()); });
330 }
331
332public:
333 ASTNode Root;
334 DumpVisitor(const syntax::TokenBuffer &Tokens, const ASTContext &Ctx)
335 : Tokens(Tokens), Ctx(Ctx) {}
336
337 // Override traversal to record the nodes we care about.
338 // Generally, these are nodes with position information (TypeLoc, not Type).
339
340 bool TraverseDecl(Decl *D) {
341 return !D || isInjectedClassName(D) ||
342 traverseNode("declaration", D, [&] { Base::TraverseDecl(D); });
343 }
344 bool TraverseTypeLoc(TypeLoc TL) {
345 return !TL || traverseNode("type", TL, [&] { Base::TraverseTypeLoc(TL); });
346 }
347 bool TraverseTemplateName(const TemplateName &TN) {
348 return traverseNode("template name", TN,
349 [&] { Base::TraverseTemplateName(TN); });
350 }
351 bool TraverseTemplateArgumentLoc(const TemplateArgumentLoc &TAL) {
352 return traverseNode("template argument", TAL,
353 [&] { Base::TraverseTemplateArgumentLoc(TAL); });
354 }
355 bool TraverseNestedNameSpecifierLoc(NestedNameSpecifierLoc NNSL) {
356 return !NNSL || traverseNode("specifier", NNSL, [&] {
357 Base::TraverseNestedNameSpecifierLoc(NNSL);
358 });
359 }
360 bool TraverseConstructorInitializer(CXXCtorInitializer *CCI) {
361 return !CCI || traverseNode("constructor initializer", CCI, [&] {
362 Base::TraverseConstructorInitializer(CCI);
363 });
364 }
365 bool TraverseAttr(Attr *A) {
366 return !A || traverseNode("attribute", A, [&] { Base::TraverseAttr(A); });
367 }
368 bool TraverseCXXBaseSpecifier(const CXXBaseSpecifier &CBS) {
369 return traverseNode("base", CBS,
370 [&] { Base::TraverseCXXBaseSpecifier(CBS); });
371 }
372 // Stmt is the same, but this form allows the data recursion optimization.
373 bool dataTraverseStmtPre(Stmt *S) {
374 return S && traverseNodePre(isa<Expr>(S) ? "expression" : "statement", S);
375 }
376 bool dataTraverseStmtPost(Stmt *X) { return traverseNodePost(); }
377
378 // QualifiedTypeLoc is handled strangely in RecursiveASTVisitor: the derived
379 // TraverseTypeLoc is not called for the inner UnqualTypeLoc.
380 // This means we'd never see 'int' in 'const int'! Work around that here.
381 // (The reason for the behavior is to avoid traversing the nested Type twice,
382 // but we ignore TraverseType anyway).
383 bool TraverseQualifiedTypeLoc(QualifiedTypeLoc QTL) {
384 return TraverseTypeLoc(QTL.getUnqualifiedLoc());
385 }
386 // Uninteresting parts of the AST that don't have locations within them.
387 bool TraverseNestedNameSpecifier(NestedNameSpecifier *) { return true; }
388 bool TraverseType(QualType) { return true; }
389
390 // OpaqueValueExpr blocks traversal, we must explicitly traverse it.
391 bool TraverseOpaqueValueExpr(OpaqueValueExpr *E) {
392 return TraverseStmt(E->getSourceExpr());
393 }
394 // We only want to traverse the *syntactic form* to understand the selection.
395 bool TraversePseudoObjectExpr(PseudoObjectExpr *E) {
396 return TraverseStmt(E->getSyntacticForm());
397 }
398};
399
400} // namespace
401
402ASTNode dumpAST(const DynTypedNode &N, const syntax::TokenBuffer &Tokens,
403 const ASTContext &Ctx) {
404 DumpVisitor V(Tokens, Ctx);
405 // DynTypedNode only works with const, RecursiveASTVisitor only non-const :-(
406 if (const auto *D = N.get<Decl>())
407 V.TraverseDecl(const_cast<Decl *>(D));
408 else if (const auto *S = N.get<Stmt>())
409 V.TraverseStmt(const_cast<Stmt *>(S));
410 else if (const auto *NNSL = N.get<NestedNameSpecifierLoc>())
411 V.TraverseNestedNameSpecifierLoc(
412 *const_cast<NestedNameSpecifierLoc *>(NNSL));
413 else if (const auto *NNS = N.get<NestedNameSpecifier>())
414 V.TraverseNestedNameSpecifier(const_cast<NestedNameSpecifier *>(NNS));
415 else if (const auto *TL = N.get<TypeLoc>())
416 V.TraverseTypeLoc(*const_cast<TypeLoc *>(TL));
417 else if (const auto *QT = N.get<QualType>())
418 V.TraverseType(*const_cast<QualType *>(QT));
419 else if (const auto *CCI = N.get<CXXCtorInitializer>())
420 V.TraverseConstructorInitializer(const_cast<CXXCtorInitializer *>(CCI));
421 else if (const auto *TAL = N.get<TemplateArgumentLoc>())
422 V.TraverseTemplateArgumentLoc(*const_cast<TemplateArgumentLoc *>(TAL));
423 else if (const auto *CBS = N.get<CXXBaseSpecifier>())
424 V.TraverseCXXBaseSpecifier(*const_cast<CXXBaseSpecifier *>(CBS));
425 else
426 elog("dumpAST: unhandled DynTypedNode kind {0}",
427 N.getNodeKind().asStringRef());
428 return std::move(V.Root);
429}
430
431} // namespace clangd
432} // namespace clang
433