1//===--- Headers.h - Include headers -----------------------------*- C++-*-===//
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#ifndef LLVM_CLANG_TOOLS_EXTRA_CLANGD_HEADERS_H
10#define LLVM_CLANG_TOOLS_EXTRA_CLANGD_HEADERS_H
11
12#include "Protocol.h"
13#include "SourceCode.h"
14#include "index/Symbol.h"
15#include "support/Path.h"
16#include "clang/Basic/FileEntry.h"
17#include "clang/Basic/TokenKinds.h"
18#include "clang/Format/Format.h"
19#include "clang/Frontend/CompilerInstance.h"
20#include "clang/Lex/HeaderSearch.h"
21#include "clang/Lex/Preprocessor.h"
22#include "clang/Tooling/Inclusions/HeaderIncludes.h"
23#include "clang/Tooling/Inclusions/StandardLibrary.h"
24#include "llvm/ADT/ArrayRef.h"
25#include "llvm/ADT/DenseSet.h"
26#include "llvm/ADT/StringRef.h"
27#include "llvm/ADT/StringSet.h"
28#include "llvm/Support/Error.h"
29#include "llvm/Support/FileSystem/UniqueID.h"
30#include <optional>
31#include <string>
32
33namespace clang {
34namespace clangd {
35
36/// Returns true if \p Include is literal include like "path" or <path>.
37bool isLiteralInclude(llvm::StringRef Include);
38
39/// Represents a header file to be #include'd.
40struct HeaderFile {
41 std::string File;
42 /// If this is true, `File` is a literal string quoted with <> or "" that
43 /// can be #included directly; otherwise, `File` is an absolute file path.
44 bool Verbatim;
45
46 bool valid() const;
47};
48
49/// A header and directives as stored in a Symbol.
50struct SymbolInclude {
51 /// The header to include. This is either a URI or a verbatim include which is
52 /// quoted with <> or "".
53 llvm::StringRef Header;
54 /// The include directive(s) that can be used, e.g. #import and/or #include.
55 Symbol::IncludeDirective Directive;
56};
57
58/// Creates a `HeaderFile` from \p Header which can be either a URI or a literal
59/// include.
60llvm::Expected<HeaderFile> toHeaderFile(llvm::StringRef Header,
61 llvm::StringRef HintPath);
62
63// Returns include headers for \p Sym sorted by popularity. If two headers are
64// equally popular, prefer the shorter one.
65llvm::SmallVector<SymbolInclude, 1> getRankedIncludes(const Symbol &Sym);
66
67// An #include directive that we found in the main file.
68struct Inclusion {
69 tok::PPKeywordKind Directive; // Directive used for inclusion, e.g. import
70 std::string Written; // Inclusion name as written e.g. <vector>.
71 Path Resolved; // Resolved path of included file. Empty if not resolved.
72 unsigned HashOffset = 0; // Byte offset from start of file to #.
73 int HashLine = 0; // Line number containing the directive, 0-indexed.
74 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User;
75 std::optional<unsigned> HeaderID;
76};
77llvm::raw_ostream &operator<<(llvm::raw_ostream &, const Inclusion &);
78bool operator==(const Inclusion &LHS, const Inclusion &RHS);
79
80// Contains information about one file in the build graph and its direct
81// dependencies. Doesn't own the strings it references (IncludeGraph is
82// self-contained).
83struct IncludeGraphNode {
84 enum class SourceFlag : uint8_t {
85 None = 0,
86 // Whether current file is a main file rather than a header.
87 IsTU = 1 << 0,
88 // Whether current file had any uncompilable errors during indexing.
89 HadErrors = 1 << 1,
90 };
91
92 SourceFlag Flags = SourceFlag::None;
93 llvm::StringRef URI;
94 FileDigest Digest{{0}};
95 std::vector<llvm::StringRef> DirectIncludes;
96};
97// FileURI and FileInclusions are references to keys of the map containing
98// them.
99// Important: The graph generated by those callbacks might contain cycles, self
100// edges and multi edges.
101using IncludeGraph = llvm::StringMap<IncludeGraphNode>;
102
103inline IncludeGraphNode::SourceFlag operator|(IncludeGraphNode::SourceFlag A,
104 IncludeGraphNode::SourceFlag B) {
105 return static_cast<IncludeGraphNode::SourceFlag>(static_cast<uint8_t>(A) |
106 static_cast<uint8_t>(B));
107}
108
109inline bool operator&(IncludeGraphNode::SourceFlag A,
110 IncludeGraphNode::SourceFlag B) {
111 return static_cast<uint8_t>(A) & static_cast<uint8_t>(B);
112}
113
114inline IncludeGraphNode::SourceFlag &
115operator|=(IncludeGraphNode::SourceFlag &A, IncludeGraphNode::SourceFlag B) {
116 return A = A | B;
117}
118
119// Information captured about the inclusion graph in a translation unit.
120// This includes detailed information about the direct #includes, and summary
121// information about all transitive includes.
122//
123// It should be built incrementally with collectIncludeStructureCallback().
124// When we build the preamble, we capture and store its include structure along
125// with the preamble data. When we use the preamble, we can copy its
126// IncludeStructure and use another collectIncludeStructureCallback() to fill
127// in any non-preamble inclusions.
128class IncludeStructure {
129public:
130 IncludeStructure() {
131 // Reserve HeaderID = 0 for the main file.
132 RealPathNames.emplace_back();
133 }
134
135 // Inserts a PPCallback and CommentHandler that visits all includes in the
136 // main file and populates the structure. It will also scan for IWYU pragmas
137 // in comments.
138 void collect(const CompilerInstance &CI);
139
140 // HeaderID identifies file in the include graph. It corresponds to a
141 // FileEntry rather than a FileID, but stays stable across preamble & main
142 // file builds.
143 enum class HeaderID : unsigned {};
144
145 std::optional<HeaderID> getID(const FileEntry *Entry) const;
146 HeaderID getOrCreateID(FileEntryRef Entry);
147
148 StringRef getRealPath(HeaderID ID) const {
149 assert(static_cast<unsigned>(ID) <= RealPathNames.size());
150 return RealPathNames[static_cast<unsigned>(ID)];
151 }
152
153 // Return all transitively reachable files.
154 llvm::ArrayRef<std::string> allHeaders() const { return RealPathNames; }
155
156 // Returns includes inside the main file with the given spelling.
157 // Spelling should include brackets or quotes, e.g. <foo>.
158 llvm::SmallVector<const Inclusion *>
159 mainFileIncludesWithSpelling(llvm::StringRef Spelling) const;
160
161 // Return all transitively reachable files, and their minimum include depth.
162 // All transitive includes (absolute paths), with their minimum include depth.
163 // Root --> 0, #included file --> 1, etc.
164 // Root is the ID of the header being visited first.
165 llvm::DenseMap<HeaderID, unsigned>
166 includeDepth(HeaderID Root = MainFileID) const;
167
168 // Maps HeaderID to the ids of the files included from it.
169 llvm::DenseMap<HeaderID, SmallVector<HeaderID>> IncludeChildren;
170
171 llvm::DenseMap<tooling::stdlib::Header, llvm::SmallVector<HeaderID>>
172 StdlibHeaders;
173
174 std::vector<Inclusion> MainFileIncludes;
175
176 // We reserve HeaderID(0) for the main file and will manually check for that
177 // in getID and getOrCreateID because the UniqueID is not stable when the
178 // content of the main file changes.
179 static const HeaderID MainFileID = HeaderID(0u);
180
181 class RecordHeaders;
182
183private:
184 // MainFileEntry will be used to check if the queried file is the main file
185 // or not.
186 const FileEntry *MainFileEntry = nullptr;
187
188 std::vector<std::string> RealPathNames; // In HeaderID order.
189 // FileEntry::UniqueID is mapped to the internal representation (HeaderID).
190 // Identifying files in a way that persists from preamble build to subsequent
191 // builds is surprisingly hard. FileID is unavailable in InclusionDirective(),
192 // and RealPathName and UniqueID are not preserved in
193 // the preamble.
194 llvm::DenseMap<llvm::sys::fs::UniqueID, HeaderID> UIDToIndex;
195
196 // Maps written includes to indices in MainFileInclude for easier lookup by
197 // spelling.
198 llvm::StringMap<llvm::SmallVector<unsigned>> MainFileIncludesBySpelling;
199};
200
201// Calculates insertion edit for including a new header in a file.
202class IncludeInserter {
203public:
204 // If \p HeaderSearchInfo is nullptr (e.g. when compile command is
205 // infeasible), this will only try to insert verbatim headers, and
206 // include path of non-verbatim header will not be shortened.
207 IncludeInserter(StringRef FileName, StringRef Code,
208 const format::FormatStyle &Style, StringRef BuildDir,
209 HeaderSearch *HeaderSearchInfo)
210 : FileName(FileName), Code(Code), BuildDir(BuildDir),
211 HeaderSearchInfo(HeaderSearchInfo),
212 Inserter(FileName, Code, Style.IncludeStyle) {}
213
214 void addExisting(const Inclusion &Inc);
215
216 /// Checks whether to add an #include of the header into \p File.
217 /// An #include will not be added if:
218 /// - Either \p DeclaringHeader or \p InsertedHeader is already (directly)
219 /// in \p Inclusions (including those included via different paths).
220 /// - \p DeclaringHeader or \p InsertedHeader is the same as \p File.
221 ///
222 /// \param DeclaringHeader is path of the original header corresponding to \p
223 /// InsertedHeader e.g. the header that declares a symbol.
224 /// \param InsertedHeader The preferred header to be inserted. This could be
225 /// the same as DeclaringHeader but must be provided.
226 bool shouldInsertInclude(PathRef DeclaringHeader,
227 const HeaderFile &InsertedHeader) const;
228
229 /// Determines the preferred way to #include a file, taking into account the
230 /// search path. Usually this will prefer a shorter representation like
231 /// 'Foo/Bar.h' over a longer one like 'Baz/include/Foo/Bar.h'.
232 ///
233 /// \param InsertedHeader The preferred header to be inserted.
234 ///
235 /// \param IncludingFile is the absolute path of the file that InsertedHeader
236 /// will be inserted.
237 ///
238 /// \return A quoted "path" or <path> to be included, or std::nullopt if it
239 /// couldn't be shortened.
240 std::optional<std::string>
241 calculateIncludePath(const HeaderFile &InsertedHeader,
242 llvm::StringRef IncludingFile) const;
243
244 /// Calculates an edit that inserts \p VerbatimHeader into code. If the header
245 /// is already included, this returns std::nullopt.
246 std::optional<TextEdit> insert(llvm::StringRef VerbatimHeader,
247 tooling::IncludeDirective Directive) const;
248
249private:
250 StringRef FileName;
251 StringRef Code;
252 StringRef BuildDir;
253 HeaderSearch *HeaderSearchInfo = nullptr;
254 llvm::StringSet<> IncludedHeaders; // Both written and resolved.
255 tooling::HeaderIncludes Inserter; // Computers insertion replacement.
256};
257
258} // namespace clangd
259} // namespace clang
260
261namespace llvm {
262
263// Support HeaderIDs as DenseMap keys.
264template <> struct DenseMapInfo<clang::clangd::IncludeStructure::HeaderID> {
265 static inline clang::clangd::IncludeStructure::HeaderID getEmptyKey() {
266 return static_cast<clang::clangd::IncludeStructure::HeaderID>(-1);
267 }
268
269 static inline clang::clangd::IncludeStructure::HeaderID getTombstoneKey() {
270 return static_cast<clang::clangd::IncludeStructure::HeaderID>(-2);
271 }
272
273 static unsigned
274 getHashValue(const clang::clangd::IncludeStructure::HeaderID &Tag) {
275 return hash_value(static_cast<unsigned>(Tag));
276 }
277
278 static bool isEqual(const clang::clangd::IncludeStructure::HeaderID &LHS,
279 const clang::clangd::IncludeStructure::HeaderID &RHS) {
280 return LHS == RHS;
281 }
282};
283
284} // namespace llvm
285
286#endif // LLVM_CLANG_TOOLS_EXTRA_CLANGD_HEADERS_H
287