blob: c8502c580996a237dab2f02b751cdca15816bb29 [file] [log] [blame]
Sam McCallb536a2a2017-12-19 12:23:48 +00001//===--- SourceCode.h - Manipulating source code as strings -----*- C++ -*-===//
2//
Chandler Carruth2946cd72019-01-19 08:50:56 +00003// 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
Sam McCallb536a2a2017-12-19 12:23:48 +00006//
7//===----------------------------------------------------------------------===//
8#include "SourceCode.h"
9
Sam McCall9fb22b22019-05-06 10:25:10 +000010#include "FuzzyMatch.h"
Kadir Cetinkaya538c2752020-05-14 12:26:47 +020011#include "Preamble.h"
Sam McCalla69698f2019-03-27 17:47:49 +000012#include "Protocol.h"
Kadir Cetinkaya5b270932019-09-09 12:28:44 +000013#include "refactor/Tweak.h"
Sam McCallad97ccf2020-04-28 17:49:17 +020014#include "support/Context.h"
15#include "support/Logger.h"
Marc-Andre Laperle1be69702018-07-05 19:35:01 +000016#include "clang/AST/ASTContext.h"
Shaurya Gupta0d26d6f2019-07-12 11:42:31 +000017#include "clang/Basic/LangOptions.h"
18#include "clang/Basic/SourceLocation.h"
Marc-Andre Laperle63a10982018-02-21 02:39:08 +000019#include "clang/Basic/SourceManager.h"
Sam McCallc316b222019-04-26 07:45:49 +000020#include "clang/Basic/TokenKinds.h"
Haojian Wu509efe52019-11-13 16:30:07 +010021#include "clang/Driver/Types.h"
Sam McCallc316b222019-04-26 07:45:49 +000022#include "clang/Format/Format.h"
Marc-Andre Laperle1be69702018-07-05 19:35:01 +000023#include "clang/Lex/Lexer.h"
Haojian Wu9d34f452019-07-01 09:26:48 +000024#include "clang/Lex/Preprocessor.h"
Kadir Cetinkayad62e3ed2019-09-25 11:35:38 +020025#include "clang/Lex/Token.h"
Kadir Cetinkaya5b270932019-09-09 12:28:44 +000026#include "clang/Tooling/Core/Replacement.h"
Kadir Cetinkaya98bb0942020-02-27 15:10:54 +010027#include "clang/Tooling/Syntax/Tokens.h"
Kadir Cetinkaya5b270932019-09-09 12:28:44 +000028#include "llvm/ADT/ArrayRef.h"
Ilya Biryukov43998782019-01-31 21:30:05 +000029#include "llvm/ADT/None.h"
Kadir Cetinkayad62e3ed2019-09-25 11:35:38 +020030#include "llvm/ADT/STLExtras.h"
Sam McCallc316b222019-04-26 07:45:49 +000031#include "llvm/ADT/StringExtras.h"
Kadir Cetinkaya5b270932019-09-09 12:28:44 +000032#include "llvm/ADT/StringMap.h"
Ilya Biryukov43998782019-01-31 21:30:05 +000033#include "llvm/ADT/StringRef.h"
Sam McCall9fb22b22019-05-06 10:25:10 +000034#include "llvm/Support/Compiler.h"
Simon Marchi766338a2018-03-21 14:36:46 +000035#include "llvm/Support/Errc.h"
36#include "llvm/Support/Error.h"
Sam McCall8b25d222019-03-28 14:37:51 +000037#include "llvm/Support/ErrorHandling.h"
Kadir Cetinkaya5b270932019-09-09 12:28:44 +000038#include "llvm/Support/LineIterator.h"
39#include "llvm/Support/MemoryBuffer.h"
Marc-Andre Laperle1be69702018-07-05 19:35:01 +000040#include "llvm/Support/Path.h"
Kadir Cetinkaya5b270932019-09-09 12:28:44 +000041#include "llvm/Support/SHA1.h"
42#include "llvm/Support/VirtualFileSystem.h"
Sam McCall674d8a92019-07-08 11:33:17 +000043#include "llvm/Support/xxhash.h"
Sam McCallc316b222019-04-26 07:45:49 +000044#include <algorithm>
Kadir Cetinkayad62e3ed2019-09-25 11:35:38 +020045#include <cstddef>
46#include <string>
47#include <vector>
Marc-Andre Laperle63a10982018-02-21 02:39:08 +000048
Sam McCallb536a2a2017-12-19 12:23:48 +000049namespace clang {
50namespace clangd {
Sam McCallb536a2a2017-12-19 12:23:48 +000051
Sam McCalla4962cc2018-04-27 11:59:28 +000052// Here be dragons. LSP positions use columns measured in *UTF-16 code units*!
53// Clangd uses UTF-8 and byte-offsets internally, so conversion is nontrivial.
54
55// Iterates over unicode codepoints in the (UTF-8) string. For each,
56// invokes CB(UTF-8 length, UTF-16 length), and breaks if it returns true.
57// Returns true if CB returned true, false if we hit the end of string.
Sam McCallf2c8f6e2020-06-10 06:40:00 +020058//
59// If the string is not valid UTF-8, we log this error and "decode" the
60// text in some arbitrary way. This is pretty sad, but this tends to happen deep
61// within indexing of headers where clang misdetected the encoding, and
62// propagating the error all the way back up is (probably?) not be worth it.
Sam McCalla4962cc2018-04-27 11:59:28 +000063template <typename Callback>
Ilya Biryukovf2001aa2019-01-07 15:45:19 +000064static bool iterateCodepoints(llvm::StringRef U8, const Callback &CB) {
Sam McCallf2c8f6e2020-06-10 06:40:00 +020065 bool LoggedInvalid = false;
Sam McCall8b25d222019-03-28 14:37:51 +000066 // A codepoint takes two UTF-16 code unit if it's astral (outside BMP).
67 // Astral codepoints are encoded as 4 bytes in UTF-8, starting with 11110xxx.
Sam McCalla4962cc2018-04-27 11:59:28 +000068 for (size_t I = 0; I < U8.size();) {
69 unsigned char C = static_cast<unsigned char>(U8[I]);
70 if (LLVM_LIKELY(!(C & 0x80))) { // ASCII character.
71 if (CB(1, 1))
72 return true;
73 ++I;
74 continue;
75 }
76 // This convenient property of UTF-8 holds for all non-ASCII characters.
Ilya Biryukovf2001aa2019-01-07 15:45:19 +000077 size_t UTF8Length = llvm::countLeadingOnes(C);
Sam McCalla4962cc2018-04-27 11:59:28 +000078 // 0xxx is ASCII, handled above. 10xxx is a trailing byte, invalid here.
Sam McCallf2c8f6e2020-06-10 06:40:00 +020079 // 11111xxx is not valid UTF-8 at all, maybe some ISO-8859-*.
80 if (LLVM_UNLIKELY(UTF8Length < 2 || UTF8Length > 4)) {
81 if (!LoggedInvalid) {
82 elog("File has invalid UTF-8 near offset {0}: {1}", I, llvm::toHex(U8));
83 LoggedInvalid = true;
84 }
85 // We can't give a correct result, but avoid returning something wild.
86 // Pretend this is a valid ASCII byte, for lack of better options.
87 // (Too late to get ISO-8859-* right, we've skipped some bytes already).
88 if (CB(1, 1))
89 return true;
90 ++I;
91 continue;
92 }
Sam McCalla4962cc2018-04-27 11:59:28 +000093 I += UTF8Length; // Skip over all trailing bytes.
94 // A codepoint takes two UTF-16 code unit if it's astral (outside BMP).
95 // Astral codepoints are encoded as 4 bytes in UTF-8 (11110xxx ...)
96 if (CB(UTF8Length, UTF8Length == 4 ? 2 : 1))
97 return true;
98 }
99 return false;
100}
101
Sam McCall8b25d222019-03-28 14:37:51 +0000102// Returns the byte offset into the string that is an offset of \p Units in
103// the specified encoding.
104// Conceptually, this converts to the encoding, truncates to CodeUnits,
105// converts back to UTF-8, and returns the length in bytes.
106static size_t measureUnits(llvm::StringRef U8, int Units, OffsetEncoding Enc,
107 bool &Valid) {
108 Valid = Units >= 0;
109 if (Units <= 0)
110 return 0;
Sam McCalla4962cc2018-04-27 11:59:28 +0000111 size_t Result = 0;
Sam McCall8b25d222019-03-28 14:37:51 +0000112 switch (Enc) {
113 case OffsetEncoding::UTF8:
114 Result = Units;
115 break;
116 case OffsetEncoding::UTF16:
117 Valid = iterateCodepoints(U8, [&](int U8Len, int U16Len) {
118 Result += U8Len;
119 Units -= U16Len;
120 return Units <= 0;
121 });
122 if (Units < 0) // Offset in the middle of a surrogate pair.
123 Valid = false;
124 break;
125 case OffsetEncoding::UTF32:
126 Valid = iterateCodepoints(U8, [&](int U8Len, int U16Len) {
127 Result += U8Len;
128 Units--;
129 return Units <= 0;
130 });
131 break;
132 case OffsetEncoding::UnsupportedEncoding:
133 llvm_unreachable("unsupported encoding");
134 }
Sam McCalla4962cc2018-04-27 11:59:28 +0000135 // Don't return an out-of-range index if we overran.
Sam McCall8b25d222019-03-28 14:37:51 +0000136 if (Result > U8.size()) {
137 Valid = false;
138 return U8.size();
139 }
140 return Result;
Sam McCalla4962cc2018-04-27 11:59:28 +0000141}
142
Sam McCalla69698f2019-03-27 17:47:49 +0000143Key<OffsetEncoding> kCurrentOffsetEncoding;
Sam McCall8b25d222019-03-28 14:37:51 +0000144static OffsetEncoding lspEncoding() {
Sam McCalla69698f2019-03-27 17:47:49 +0000145 auto *Enc = Context::current().get(kCurrentOffsetEncoding);
Sam McCall8b25d222019-03-28 14:37:51 +0000146 return Enc ? *Enc : OffsetEncoding::UTF16;
Sam McCalla69698f2019-03-27 17:47:49 +0000147}
148
Sam McCalla4962cc2018-04-27 11:59:28 +0000149// Like most strings in clangd, the input is UTF-8 encoded.
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000150size_t lspLength(llvm::StringRef Code) {
Sam McCalla4962cc2018-04-27 11:59:28 +0000151 size_t Count = 0;
Sam McCall8b25d222019-03-28 14:37:51 +0000152 switch (lspEncoding()) {
153 case OffsetEncoding::UTF8:
154 Count = Code.size();
155 break;
156 case OffsetEncoding::UTF16:
157 iterateCodepoints(Code, [&](int U8Len, int U16Len) {
158 Count += U16Len;
159 return false;
160 });
161 break;
162 case OffsetEncoding::UTF32:
163 iterateCodepoints(Code, [&](int U8Len, int U16Len) {
164 ++Count;
165 return false;
166 });
167 break;
168 case OffsetEncoding::UnsupportedEncoding:
169 llvm_unreachable("unsupported encoding");
170 }
Sam McCalla4962cc2018-04-27 11:59:28 +0000171 return Count;
172}
173
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000174llvm::Expected<size_t> positionToOffset(llvm::StringRef Code, Position P,
175 bool AllowColumnsBeyondLineLength) {
Sam McCallb536a2a2017-12-19 12:23:48 +0000176 if (P.line < 0)
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000177 return llvm::make_error<llvm::StringError>(
178 llvm::formatv("Line value can't be negative ({0})", P.line),
179 llvm::errc::invalid_argument);
Simon Marchi766338a2018-03-21 14:36:46 +0000180 if (P.character < 0)
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000181 return llvm::make_error<llvm::StringError>(
182 llvm::formatv("Character value can't be negative ({0})", P.character),
183 llvm::errc::invalid_argument);
Sam McCallb536a2a2017-12-19 12:23:48 +0000184 size_t StartOfLine = 0;
185 for (int I = 0; I != P.line; ++I) {
186 size_t NextNL = Code.find('\n', StartOfLine);
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000187 if (NextNL == llvm::StringRef::npos)
188 return llvm::make_error<llvm::StringError>(
189 llvm::formatv("Line value is out of range ({0})", P.line),
190 llvm::errc::invalid_argument);
Sam McCallb536a2a2017-12-19 12:23:48 +0000191 StartOfLine = NextNL + 1;
192 }
Sam McCalla69698f2019-03-27 17:47:49 +0000193 StringRef Line =
194 Code.substr(StartOfLine).take_until([](char C) { return C == '\n'; });
Simon Marchi766338a2018-03-21 14:36:46 +0000195
Sam McCall8b25d222019-03-28 14:37:51 +0000196 // P.character may be in UTF-16, transcode if necessary.
Sam McCalla4962cc2018-04-27 11:59:28 +0000197 bool Valid;
Sam McCall8b25d222019-03-28 14:37:51 +0000198 size_t ByteInLine = measureUnits(Line, P.character, lspEncoding(), Valid);
Sam McCalla4962cc2018-04-27 11:59:28 +0000199 if (!Valid && !AllowColumnsBeyondLineLength)
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000200 return llvm::make_error<llvm::StringError>(
Sam McCall8b25d222019-03-28 14:37:51 +0000201 llvm::formatv("{0} offset {1} is invalid for line {2}", lspEncoding(),
202 P.character, P.line),
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000203 llvm::errc::invalid_argument);
Sam McCall8b25d222019-03-28 14:37:51 +0000204 return StartOfLine + ByteInLine;
Sam McCallb536a2a2017-12-19 12:23:48 +0000205}
206
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000207Position offsetToPosition(llvm::StringRef Code, size_t Offset) {
Sam McCallb536a2a2017-12-19 12:23:48 +0000208 Offset = std::min(Code.size(), Offset);
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000209 llvm::StringRef Before = Code.substr(0, Offset);
Sam McCallb536a2a2017-12-19 12:23:48 +0000210 int Lines = Before.count('\n');
211 size_t PrevNL = Before.rfind('\n');
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000212 size_t StartOfLine = (PrevNL == llvm::StringRef::npos) ? 0 : (PrevNL + 1);
Ilya Biryukov7beea3a2018-02-14 10:52:04 +0000213 Position Pos;
214 Pos.line = Lines;
Sam McCall71891122018-10-23 11:51:53 +0000215 Pos.character = lspLength(Before.substr(StartOfLine));
Ilya Biryukov7beea3a2018-02-14 10:52:04 +0000216 return Pos;
Sam McCallb536a2a2017-12-19 12:23:48 +0000217}
218
Marc-Andre Laperle63a10982018-02-21 02:39:08 +0000219Position sourceLocToPosition(const SourceManager &SM, SourceLocation Loc) {
Sam McCalla4962cc2018-04-27 11:59:28 +0000220 // We use the SourceManager's line tables, but its column number is in bytes.
221 FileID FID;
222 unsigned Offset;
223 std::tie(FID, Offset) = SM.getDecomposedSpellingLoc(Loc);
Marc-Andre Laperle63a10982018-02-21 02:39:08 +0000224 Position P;
Sam McCalla4962cc2018-04-27 11:59:28 +0000225 P.line = static_cast<int>(SM.getLineNumber(FID, Offset)) - 1;
226 bool Invalid = false;
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000227 llvm::StringRef Code = SM.getBufferData(FID, &Invalid);
Sam McCalla4962cc2018-04-27 11:59:28 +0000228 if (!Invalid) {
229 auto ColumnInBytes = SM.getColumnNumber(FID, Offset) - 1;
230 auto LineSoFar = Code.substr(Offset - ColumnInBytes, ColumnInBytes);
Sam McCall71891122018-10-23 11:51:53 +0000231 P.character = lspLength(LineSoFar);
Sam McCalla4962cc2018-04-27 11:59:28 +0000232 }
Marc-Andre Laperle63a10982018-02-21 02:39:08 +0000233 return P;
234}
235
Sam McCall95738072019-08-06 20:25:59 +0000236bool isSpelledInSource(SourceLocation Loc, const SourceManager &SM) {
237 if (Loc.isMacroID()) {
238 std::string PrintLoc = SM.getSpellingLoc(Loc).printToString(SM);
239 if (llvm::StringRef(PrintLoc).startswith("<scratch") ||
240 llvm::StringRef(PrintLoc).startswith("<command line>"))
241 return false;
242 }
243 return true;
244}
245
Ilya Biryukov43998782019-01-31 21:30:05 +0000246bool isValidFileRange(const SourceManager &Mgr, SourceRange R) {
247 if (!R.getBegin().isValid() || !R.getEnd().isValid())
248 return false;
249
250 FileID BeginFID;
251 size_t BeginOffset = 0;
252 std::tie(BeginFID, BeginOffset) = Mgr.getDecomposedLoc(R.getBegin());
253
254 FileID EndFID;
255 size_t EndOffset = 0;
256 std::tie(EndFID, EndOffset) = Mgr.getDecomposedLoc(R.getEnd());
257
258 return BeginFID.isValid() && BeginFID == EndFID && BeginOffset <= EndOffset;
259}
260
Sam McCallc791d852019-08-27 08:44:06 +0000261SourceLocation includeHashLoc(FileID IncludedFile, const SourceManager &SM) {
262 assert(SM.getLocForEndOfFile(IncludedFile).isFileID());
263 FileID IncludingFile;
264 unsigned Offset;
265 std::tie(IncludingFile, Offset) =
266 SM.getDecomposedExpansionLoc(SM.getIncludeLoc(IncludedFile));
267 bool Invalid = false;
268 llvm::StringRef Buf = SM.getBufferData(IncludingFile, &Invalid);
269 if (Invalid)
270 return SourceLocation();
271 // Now buf is "...\n#include <foo>\n..."
272 // and Offset points here: ^
273 // Rewind to the preceding # on the line.
274 assert(Offset < Buf.size());
275 for (;; --Offset) {
276 if (Buf[Offset] == '#')
277 return SM.getComposedLoc(IncludingFile, Offset);
278 if (Buf[Offset] == '\n' || Offset == 0) // no hash, what's going on?
279 return SourceLocation();
280 }
281}
282
Shaurya Gupta0d26d6f2019-07-12 11:42:31 +0000283static unsigned getTokenLengthAtLoc(SourceLocation Loc, const SourceManager &SM,
284 const LangOptions &LangOpts) {
285 Token TheTok;
286 if (Lexer::getRawToken(Loc, TheTok, SM, LangOpts))
287 return 0;
288 // FIXME: Here we check whether the token at the location is a greatergreater
289 // (>>) token and consider it as a single greater (>). This is to get it
290 // working for templates but it isn't correct for the right shift operator. We
291 // can avoid this by using half open char ranges in getFileRange() but getting
292 // token ending is not well supported in macroIDs.
293 if (TheTok.is(tok::greatergreater))
294 return 1;
295 return TheTok.getLength();
296}
297
298// Returns location of the last character of the token at a given loc
299static SourceLocation getLocForTokenEnd(SourceLocation BeginLoc,
300 const SourceManager &SM,
301 const LangOptions &LangOpts) {
302 unsigned Len = getTokenLengthAtLoc(BeginLoc, SM, LangOpts);
303 return BeginLoc.getLocWithOffset(Len ? Len - 1 : 0);
304}
305
306// Returns location of the starting of the token at a given EndLoc
307static SourceLocation getLocForTokenBegin(SourceLocation EndLoc,
308 const SourceManager &SM,
309 const LangOptions &LangOpts) {
310 return EndLoc.getLocWithOffset(
311 -(signed)getTokenLengthAtLoc(EndLoc, SM, LangOpts));
312}
313
314// Converts a char source range to a token range.
315static SourceRange toTokenRange(CharSourceRange Range, const SourceManager &SM,
316 const LangOptions &LangOpts) {
317 if (!Range.isTokenRange())
318 Range.setEnd(getLocForTokenBegin(Range.getEnd(), SM, LangOpts));
319 return Range.getAsRange();
320}
321// Returns the union of two token ranges.
322// To find the maximum of the Ends of the ranges, we compare the location of the
323// last character of the token.
324static SourceRange unionTokenRange(SourceRange R1, SourceRange R2,
325 const SourceManager &SM,
326 const LangOptions &LangOpts) {
Sam McCallc791d852019-08-27 08:44:06 +0000327 SourceLocation Begin =
328 SM.isBeforeInTranslationUnit(R1.getBegin(), R2.getBegin())
329 ? R1.getBegin()
330 : R2.getBegin();
331 SourceLocation End =
332 SM.isBeforeInTranslationUnit(getLocForTokenEnd(R1.getEnd(), SM, LangOpts),
333 getLocForTokenEnd(R2.getEnd(), SM, LangOpts))
334 ? R2.getEnd()
335 : R1.getEnd();
336 return SourceRange(Begin, End);
Shaurya Gupta0d26d6f2019-07-12 11:42:31 +0000337}
338
Sam McCallc791d852019-08-27 08:44:06 +0000339// Given a range whose endpoints may be in different expansions or files,
340// tries to find a range within a common file by following up the expansion and
341// include location in each.
342static SourceRange rangeInCommonFile(SourceRange R, const SourceManager &SM,
343 const LangOptions &LangOpts) {
344 // Fast path for most common cases.
345 if (SM.isWrittenInSameFile(R.getBegin(), R.getEnd()))
346 return R;
347 // Record the stack of expansion locations for the beginning, keyed by FileID.
348 llvm::DenseMap<FileID, SourceLocation> BeginExpansions;
349 for (SourceLocation Begin = R.getBegin(); Begin.isValid();
350 Begin = Begin.isFileID()
351 ? includeHashLoc(SM.getFileID(Begin), SM)
352 : SM.getImmediateExpansionRange(Begin).getBegin()) {
353 BeginExpansions[SM.getFileID(Begin)] = Begin;
354 }
355 // Move up the stack of expansion locations for the end until we find the
356 // location in BeginExpansions with that has the same file id.
357 for (SourceLocation End = R.getEnd(); End.isValid();
358 End = End.isFileID() ? includeHashLoc(SM.getFileID(End), SM)
359 : toTokenRange(SM.getImmediateExpansionRange(End),
360 SM, LangOpts)
361 .getEnd()) {
362 auto It = BeginExpansions.find(SM.getFileID(End));
363 if (It != BeginExpansions.end()) {
364 if (SM.getFileOffset(It->second) > SM.getFileOffset(End))
365 return SourceLocation();
366 return {It->second, End};
367 }
368 }
369 return SourceRange();
Shaurya Gupta8fbb6ce2019-08-06 17:01:12 +0000370}
371
372// Find an expansion range (not necessarily immediate) the ends of which are in
373// the same file id.
374static SourceRange
375getExpansionTokenRangeInSameFile(SourceLocation Loc, const SourceManager &SM,
376 const LangOptions &LangOpts) {
Sam McCallc791d852019-08-27 08:44:06 +0000377 return rangeInCommonFile(
378 toTokenRange(SM.getImmediateExpansionRange(Loc), SM, LangOpts), SM,
379 LangOpts);
Shaurya Gupta8fbb6ce2019-08-06 17:01:12 +0000380}
Sam McCallc791d852019-08-27 08:44:06 +0000381
Shaurya Gupta8fbb6ce2019-08-06 17:01:12 +0000382// Returns the file range for a given Location as a Token Range
Shaurya Gupta0d26d6f2019-07-12 11:42:31 +0000383// This is quite similar to getFileLoc in SourceManager as both use
384// getImmediateExpansionRange and getImmediateSpellingLoc (for macro IDs).
385// However:
386// - We want to maintain the full range information as we move from one file to
387// the next. getFileLoc only uses the BeginLoc of getImmediateExpansionRange.
Shaurya Gupta8fbb6ce2019-08-06 17:01:12 +0000388// - We want to split '>>' tokens as the lexer parses the '>>' in nested
389// template instantiations as a '>>' instead of two '>'s.
Shaurya Gupta0d26d6f2019-07-12 11:42:31 +0000390// There is also getExpansionRange but it simply calls
391// getImmediateExpansionRange on the begin and ends separately which is wrong.
392static SourceRange getTokenFileRange(SourceLocation Loc,
393 const SourceManager &SM,
394 const LangOptions &LangOpts) {
395 SourceRange FileRange = Loc;
396 while (!FileRange.getBegin().isFileID()) {
Shaurya Gupta0d26d6f2019-07-12 11:42:31 +0000397 if (SM.isMacroArgExpansion(FileRange.getBegin())) {
Shaurya Gupta8fbb6ce2019-08-06 17:01:12 +0000398 FileRange = unionTokenRange(
399 SM.getImmediateSpellingLoc(FileRange.getBegin()),
400 SM.getImmediateSpellingLoc(FileRange.getEnd()), SM, LangOpts);
Sam McCallc791d852019-08-27 08:44:06 +0000401 assert(SM.isWrittenInSameFile(FileRange.getBegin(), FileRange.getEnd()));
Shaurya Gupta0d26d6f2019-07-12 11:42:31 +0000402 } else {
Shaurya Gupta8fbb6ce2019-08-06 17:01:12 +0000403 SourceRange ExpansionRangeForBegin =
404 getExpansionTokenRangeInSameFile(FileRange.getBegin(), SM, LangOpts);
405 SourceRange ExpansionRangeForEnd =
406 getExpansionTokenRangeInSameFile(FileRange.getEnd(), SM, LangOpts);
Sam McCallc791d852019-08-27 08:44:06 +0000407 if (ExpansionRangeForBegin.isInvalid() ||
408 ExpansionRangeForEnd.isInvalid())
409 return SourceRange();
410 assert(SM.isWrittenInSameFile(ExpansionRangeForBegin.getBegin(),
411 ExpansionRangeForEnd.getBegin()) &&
Shaurya Gupta8fbb6ce2019-08-06 17:01:12 +0000412 "Both Expansion ranges should be in same file.");
Shaurya Gupta0d26d6f2019-07-12 11:42:31 +0000413 FileRange = unionTokenRange(ExpansionRangeForBegin, ExpansionRangeForEnd,
414 SM, LangOpts);
415 }
416 }
417 return FileRange;
418}
419
Haojian Wu6ae86ea2019-07-19 08:33:39 +0000420bool isInsideMainFile(SourceLocation Loc, const SourceManager &SM) {
Sam McCall2be45692020-03-02 16:54:56 +0100421 if (!Loc.isValid())
422 return false;
423 FileID FID = SM.getFileID(SM.getExpansionLoc(Loc));
424 return FID == SM.getMainFileID() || FID == SM.getPreambleFileID();
Haojian Wu6ae86ea2019-07-19 08:33:39 +0000425}
426
Shaurya Gupta0d26d6f2019-07-12 11:42:31 +0000427llvm::Optional<SourceRange> toHalfOpenFileRange(const SourceManager &SM,
Ilya Biryukov43998782019-01-31 21:30:05 +0000428 const LangOptions &LangOpts,
429 SourceRange R) {
Shaurya Gupta0d26d6f2019-07-12 11:42:31 +0000430 SourceRange R1 = getTokenFileRange(R.getBegin(), SM, LangOpts);
431 if (!isValidFileRange(SM, R1))
Ilya Biryukov43998782019-01-31 21:30:05 +0000432 return llvm::None;
Ilya Biryukov43998782019-01-31 21:30:05 +0000433
Shaurya Gupta0d26d6f2019-07-12 11:42:31 +0000434 SourceRange R2 = getTokenFileRange(R.getEnd(), SM, LangOpts);
435 if (!isValidFileRange(SM, R2))
Ilya Biryukov43998782019-01-31 21:30:05 +0000436 return llvm::None;
Shaurya Gupta0d26d6f2019-07-12 11:42:31 +0000437
Sam McCallc791d852019-08-27 08:44:06 +0000438 SourceRange Result =
439 rangeInCommonFile(unionTokenRange(R1, R2, SM, LangOpts), SM, LangOpts);
Shaurya Gupta0d26d6f2019-07-12 11:42:31 +0000440 unsigned TokLen = getTokenLengthAtLoc(Result.getEnd(), SM, LangOpts);
441 // Convert from closed token range to half-open (char) range
442 Result.setEnd(Result.getEnd().getLocWithOffset(TokLen));
443 if (!isValidFileRange(SM, Result))
444 return llvm::None;
445
Ilya Biryukov43998782019-01-31 21:30:05 +0000446 return Result;
447}
448
449llvm::StringRef toSourceCode(const SourceManager &SM, SourceRange R) {
450 assert(isValidFileRange(SM, R));
451 bool Invalid = false;
452 auto *Buf = SM.getBuffer(SM.getFileID(R.getBegin()), &Invalid);
453 assert(!Invalid);
454
455 size_t BeginOffset = SM.getFileOffset(R.getBegin());
456 size_t EndOffset = SM.getFileOffset(R.getEnd());
457 return Buf->getBuffer().substr(BeginOffset, EndOffset - BeginOffset);
458}
459
Ilya Biryukovcce67a32019-01-29 14:17:36 +0000460llvm::Expected<SourceLocation> sourceLocationInMainFile(const SourceManager &SM,
461 Position P) {
462 llvm::StringRef Code = SM.getBuffer(SM.getMainFileID())->getBuffer();
463 auto Offset =
464 positionToOffset(Code, P, /*AllowColumnBeyondLineLength=*/false);
465 if (!Offset)
466 return Offset.takeError();
467 return SM.getLocForStartOfFile(SM.getMainFileID()).getLocWithOffset(*Offset);
468}
469
Ilya Biryukov71028b82018-03-12 15:28:22 +0000470Range halfOpenToRange(const SourceManager &SM, CharSourceRange R) {
471 // Clang is 1-based, LSP uses 0-based indexes.
472 Position Begin = sourceLocToPosition(SM, R.getBegin());
473 Position End = sourceLocToPosition(SM, R.getEnd());
474
475 return {Begin, End};
476}
477
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000478std::pair<size_t, size_t> offsetToClangLineColumn(llvm::StringRef Code,
Sam McCalla4962cc2018-04-27 11:59:28 +0000479 size_t Offset) {
480 Offset = std::min(Code.size(), Offset);
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000481 llvm::StringRef Before = Code.substr(0, Offset);
Sam McCalla4962cc2018-04-27 11:59:28 +0000482 int Lines = Before.count('\n');
483 size_t PrevNL = Before.rfind('\n');
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000484 size_t StartOfLine = (PrevNL == llvm::StringRef::npos) ? 0 : (PrevNL + 1);
Sam McCalla4962cc2018-04-27 11:59:28 +0000485 return {Lines + 1, Offset - StartOfLine + 1};
486}
487
Ilya Biryukov43998782019-01-31 21:30:05 +0000488std::pair<StringRef, StringRef> splitQualifiedName(StringRef QName) {
Marc-Andre Laperleb387b6e2018-04-23 20:00:52 +0000489 size_t Pos = QName.rfind("::");
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000490 if (Pos == llvm::StringRef::npos)
491 return {llvm::StringRef(), QName};
Marc-Andre Laperleb387b6e2018-04-23 20:00:52 +0000492 return {QName.substr(0, Pos + 2), QName.substr(Pos + 2)};
493}
494
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000495TextEdit replacementToEdit(llvm::StringRef Code,
496 const tooling::Replacement &R) {
Eric Liu9133ecd2018-05-11 12:12:08 +0000497 Range ReplacementRange = {
498 offsetToPosition(Code, R.getOffset()),
499 offsetToPosition(Code, R.getOffset() + R.getLength())};
Benjamin Krameradcd0262020-01-28 20:23:46 +0100500 return {ReplacementRange, std::string(R.getReplacementText())};
Eric Liu9133ecd2018-05-11 12:12:08 +0000501}
502
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000503std::vector<TextEdit> replacementsToEdits(llvm::StringRef Code,
Eric Liu9133ecd2018-05-11 12:12:08 +0000504 const tooling::Replacements &Repls) {
505 std::vector<TextEdit> Edits;
506 for (const auto &R : Repls)
507 Edits.push_back(replacementToEdit(Code, R));
508 return Edits;
509}
510
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000511llvm::Optional<std::string> getCanonicalPath(const FileEntry *F,
512 const SourceManager &SourceMgr) {
Kadir Cetinkayadd677932018-12-19 10:46:21 +0000513 if (!F)
514 return None;
Simon Marchi25f1f732018-08-10 22:27:53 +0000515
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000516 llvm::SmallString<128> FilePath = F->getName();
517 if (!llvm::sys::path::is_absolute(FilePath)) {
Kadir Cetinkayadd677932018-12-19 10:46:21 +0000518 if (auto EC =
Duncan P. N. Exon Smithdb8a7422019-03-26 22:32:06 +0000519 SourceMgr.getFileManager().getVirtualFileSystem().makeAbsolute(
Kadir Cetinkayadd677932018-12-19 10:46:21 +0000520 FilePath)) {
521 elog("Could not turn relative path '{0}' to absolute: {1}", FilePath,
522 EC.message());
Sam McCallc008af62018-10-20 15:30:37 +0000523 return None;
Marc-Andre Laperle1be69702018-07-05 19:35:01 +0000524 }
525 }
Simon Marchi25f1f732018-08-10 22:27:53 +0000526
Kadir Cetinkayadd677932018-12-19 10:46:21 +0000527 // Handle the symbolic link path case where the current working directory
Kadir Cetinkaya5b270932019-09-09 12:28:44 +0000528 // (getCurrentWorkingDirectory) is a symlink. We always want to the real
Kadir Cetinkayadd677932018-12-19 10:46:21 +0000529 // file path (instead of the symlink path) for the C++ symbols.
530 //
531 // Consider the following example:
532 //
533 // src dir: /project/src/foo.h
534 // current working directory (symlink): /tmp/build -> /project/src/
535 //
536 // The file path of Symbol is "/project/src/foo.h" instead of
537 // "/tmp/build/foo.h"
Harlan Haskinsa02f8572019-08-01 21:32:01 +0000538 if (auto Dir = SourceMgr.getFileManager().getDirectory(
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000539 llvm::sys::path::parent_path(FilePath))) {
540 llvm::SmallString<128> RealPath;
Harlan Haskinsa02f8572019-08-01 21:32:01 +0000541 llvm::StringRef DirName = SourceMgr.getFileManager().getCanonicalName(*Dir);
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000542 llvm::sys::path::append(RealPath, DirName,
543 llvm::sys::path::filename(FilePath));
Kadir Cetinkayadd677932018-12-19 10:46:21 +0000544 return RealPath.str().str();
Simon Marchi25f1f732018-08-10 22:27:53 +0000545 }
546
Kadir Cetinkayadd677932018-12-19 10:46:21 +0000547 return FilePath.str().str();
Marc-Andre Laperle1be69702018-07-05 19:35:01 +0000548}
549
Kadir Cetinkaya2f84d912018-08-08 08:59:29 +0000550TextEdit toTextEdit(const FixItHint &FixIt, const SourceManager &M,
551 const LangOptions &L) {
552 TextEdit Result;
553 Result.range =
554 halfOpenToRange(M, Lexer::makeFileCharRange(FixIt.RemoveRange, M, L));
555 Result.newText = FixIt.CodeToInsert;
556 return Result;
557}
558
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000559FileDigest digest(llvm::StringRef Content) {
Sam McCall674d8a92019-07-08 11:33:17 +0000560 uint64_t Hash{llvm::xxHash64(Content)};
561 FileDigest Result;
562 for (unsigned I = 0; I < Result.size(); ++I) {
563 Result[I] = uint8_t(Hash);
564 Hash >>= 8;
565 }
566 return Result;
Kadir Cetinkayad08eab42018-11-27 16:08:53 +0000567}
568
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000569llvm::Optional<FileDigest> digestFile(const SourceManager &SM, FileID FID) {
Kadir Cetinkayad08eab42018-11-27 16:08:53 +0000570 bool Invalid = false;
Ilya Biryukovf2001aa2019-01-07 15:45:19 +0000571 llvm::StringRef Content = SM.getBufferData(FID, &Invalid);
Kadir Cetinkayad08eab42018-11-27 16:08:53 +0000572 if (Invalid)
573 return None;
574 return digest(Content);
575}
576
Eric Liudd662772019-01-28 14:01:55 +0000577format::FormatStyle getFormatStyleForFile(llvm::StringRef File,
578 llvm::StringRef Content,
579 llvm::vfs::FileSystem *FS) {
580 auto Style = format::getStyle(format::DefaultFormatStyle, File,
581 format::DefaultFallbackStyle, Content, FS);
582 if (!Style) {
583 log("getStyle() failed for file {0}: {1}. Fallback is LLVM style.", File,
584 Style.takeError());
Sam McCall6fe20a42020-05-03 22:13:49 +0200585 return format::getLLVMStyle();
Eric Liudd662772019-01-28 14:01:55 +0000586 }
587 return *Style;
588}
589
Haojian Wu12e194c2019-02-06 15:24:50 +0000590llvm::Expected<tooling::Replacements>
591cleanupAndFormat(StringRef Code, const tooling::Replacements &Replaces,
592 const format::FormatStyle &Style) {
593 auto CleanReplaces = cleanupAroundReplacements(Code, Replaces, Style);
594 if (!CleanReplaces)
595 return CleanReplaces;
596 return formatReplacements(Code, std::move(*CleanReplaces), Style);
597}
598
Haojian Wuc5e4cf42019-11-07 10:53:19 +0100599static void
600lex(llvm::StringRef Code, const LangOptions &LangOpts,
Kadir Cetinkaya98bb0942020-02-27 15:10:54 +0100601 llvm::function_ref<void(const syntax::Token &, const SourceManager &SM)>
Haojian Wuc5e4cf42019-11-07 10:53:19 +0100602 Action) {
Sam McCallc316b222019-04-26 07:45:49 +0000603 // FIXME: InMemoryFileAdapter crashes unless the buffer is null terminated!
604 std::string NullTerminatedCode = Code.str();
605 SourceManagerForFile FileSM("dummy.cpp", NullTerminatedCode);
Eric Liu00d99bd2019-04-11 09:36:36 +0000606 auto &SM = FileSM.get();
Kadir Cetinkaya98bb0942020-02-27 15:10:54 +0100607 for (const auto &Tok : syntax::tokenize(SM.getMainFileID(), SM, LangOpts))
Haojian Wu7ea4c6f2019-10-30 13:21:47 +0100608 Action(Tok, SM);
Sam McCallc316b222019-04-26 07:45:49 +0000609}
610
611llvm::StringMap<unsigned> collectIdentifiers(llvm::StringRef Content,
612 const format::FormatStyle &Style) {
Eric Liu00d99bd2019-04-11 09:36:36 +0000613 llvm::StringMap<unsigned> Identifiers;
Haojian Wu7ea4c6f2019-10-30 13:21:47 +0100614 auto LangOpt = format::getFormattingLangOpts(Style);
Kadir Cetinkaya98bb0942020-02-27 15:10:54 +0100615 lex(Content, LangOpt, [&](const syntax::Token &Tok, const SourceManager &SM) {
616 if (Tok.kind() == tok::identifier)
617 ++Identifiers[Tok.text(SM)];
618 // FIXME: Should this function really return keywords too ?
619 else if (const auto *Keyword = tok::getKeywordSpelling(Tok.kind()))
620 ++Identifiers[Keyword];
Sam McCallc316b222019-04-26 07:45:49 +0000621 });
Eric Liu00d99bd2019-04-11 09:36:36 +0000622 return Identifiers;
623}
624
Haojian Wu7ea4c6f2019-10-30 13:21:47 +0100625std::vector<Range> collectIdentifierRanges(llvm::StringRef Identifier,
626 llvm::StringRef Content,
627 const LangOptions &LangOpts) {
628 std::vector<Range> Ranges;
Kadir Cetinkaya98bb0942020-02-27 15:10:54 +0100629 lex(Content, LangOpts,
630 [&](const syntax::Token &Tok, const SourceManager &SM) {
631 if (Tok.kind() != tok::identifier || Tok.text(SM) != Identifier)
632 return;
Kadir Cetinkaya37550392020-03-01 16:05:12 +0100633 Ranges.push_back(halfOpenToRange(SM, Tok.range(SM).toCharRange(SM)));
Kadir Cetinkaya98bb0942020-02-27 15:10:54 +0100634 });
Haojian Wu7ea4c6f2019-10-30 13:21:47 +0100635 return Ranges;
636}
637
Sam McCallc316b222019-04-26 07:45:49 +0000638namespace {
Kadir Cetinkayad62e3ed2019-09-25 11:35:38 +0200639struct NamespaceEvent {
640 enum {
641 BeginNamespace, // namespace <ns> {. Payload is resolved <ns>.
642 EndNamespace, // } // namespace <ns>. Payload is resolved *outer*
643 // namespace.
644 UsingDirective // using namespace <ns>. Payload is unresolved <ns>.
645 } Trigger;
646 std::string Payload;
647 Position Pos;
Sam McCallc316b222019-04-26 07:45:49 +0000648};
649// Scans C++ source code for constructs that change the visible namespaces.
Nathan Ridge445195b2020-03-05 19:03:26 -0500650void parseNamespaceEvents(llvm::StringRef Code, const LangOptions &LangOpts,
Kadir Cetinkayad62e3ed2019-09-25 11:35:38 +0200651 llvm::function_ref<void(NamespaceEvent)> Callback) {
Sam McCallc316b222019-04-26 07:45:49 +0000652
653 // Stack of enclosing namespaces, e.g. {"clang", "clangd"}
654 std::vector<std::string> Enclosing; // Contains e.g. "clang", "clangd"
655 // Stack counts open braces. true if the brace opened a namespace.
656 std::vector<bool> BraceStack;
657
658 enum {
659 Default,
660 Namespace, // just saw 'namespace'
661 NamespaceName, // just saw 'namespace' NSName
662 Using, // just saw 'using'
663 UsingNamespace, // just saw 'using namespace'
664 UsingNamespaceName, // just saw 'using namespace' NSName
665 } State = Default;
666 std::string NSName;
667
Kadir Cetinkayad62e3ed2019-09-25 11:35:38 +0200668 NamespaceEvent Event;
Nathan Ridge445195b2020-03-05 19:03:26 -0500669 lex(Code, LangOpts, [&](const syntax::Token &Tok, const SourceManager &SM) {
670 Event.Pos = sourceLocToPosition(SM, Tok.location());
671 switch (Tok.kind()) {
672 case tok::kw_using:
673 State = State == Default ? Using : Default;
674 break;
675 case tok::kw_namespace:
676 switch (State) {
677 case Using:
678 State = UsingNamespace;
679 break;
680 case Default:
681 State = Namespace;
682 break;
683 default:
684 State = Default;
685 break;
686 }
687 break;
688 case tok::identifier:
689 switch (State) {
690 case UsingNamespace:
691 NSName.clear();
692 LLVM_FALLTHROUGH;
693 case UsingNamespaceName:
694 NSName.append(Tok.text(SM).str());
695 State = UsingNamespaceName;
696 break;
697 case Namespace:
698 NSName.clear();
699 LLVM_FALLTHROUGH;
700 case NamespaceName:
701 NSName.append(Tok.text(SM).str());
702 State = NamespaceName;
703 break;
704 case Using:
705 case Default:
706 State = Default;
707 break;
708 }
709 break;
710 case tok::coloncolon:
711 // This can come at the beginning or in the middle of a namespace
712 // name.
713 switch (State) {
714 case UsingNamespace:
715 NSName.clear();
716 LLVM_FALLTHROUGH;
717 case UsingNamespaceName:
718 NSName.append("::");
719 State = UsingNamespaceName;
720 break;
721 case NamespaceName:
722 NSName.append("::");
723 State = NamespaceName;
724 break;
725 case Namespace: // Not legal here.
726 case Using:
727 case Default:
728 State = Default;
729 break;
730 }
731 break;
732 case tok::l_brace:
733 // Record which { started a namespace, so we know when } ends one.
734 if (State == NamespaceName) {
735 // Parsed: namespace <name> {
736 BraceStack.push_back(true);
737 Enclosing.push_back(NSName);
738 Event.Trigger = NamespaceEvent::BeginNamespace;
739 Event.Payload = llvm::join(Enclosing, "::");
740 Callback(Event);
741 } else {
742 // This case includes anonymous namespaces (State = Namespace).
743 // For our purposes, they're not namespaces and we ignore them.
744 BraceStack.push_back(false);
745 }
746 State = Default;
747 break;
748 case tok::r_brace:
749 // If braces are unmatched, we're going to be confused, but don't
750 // crash.
751 if (!BraceStack.empty()) {
752 if (BraceStack.back()) {
753 // Parsed: } // namespace
754 Enclosing.pop_back();
755 Event.Trigger = NamespaceEvent::EndNamespace;
756 Event.Payload = llvm::join(Enclosing, "::");
757 Callback(Event);
Sam McCallc316b222019-04-26 07:45:49 +0000758 }
Nathan Ridge445195b2020-03-05 19:03:26 -0500759 BraceStack.pop_back();
760 }
761 break;
762 case tok::semi:
763 if (State == UsingNamespaceName) {
764 // Parsed: using namespace <name> ;
765 Event.Trigger = NamespaceEvent::UsingDirective;
766 Event.Payload = std::move(NSName);
767 Callback(Event);
768 }
769 State = Default;
770 break;
771 default:
772 State = Default;
773 break;
774 }
775 });
Sam McCallc316b222019-04-26 07:45:49 +0000776}
777
778// Returns the prefix namespaces of NS: {"" ... NS}.
779llvm::SmallVector<llvm::StringRef, 8> ancestorNamespaces(llvm::StringRef NS) {
780 llvm::SmallVector<llvm::StringRef, 8> Results;
781 Results.push_back(NS.take_front(0));
782 NS.split(Results, "::", /*MaxSplit=*/-1, /*KeepEmpty=*/false);
783 for (llvm::StringRef &R : Results)
784 R = NS.take_front(R.end() - NS.begin());
785 return Results;
786}
787
788} // namespace
789
790std::vector<std::string> visibleNamespaces(llvm::StringRef Code,
Nathan Ridge445195b2020-03-05 19:03:26 -0500791 const LangOptions &LangOpts) {
Sam McCallc316b222019-04-26 07:45:49 +0000792 std::string Current;
793 // Map from namespace to (resolved) namespaces introduced via using directive.
794 llvm::StringMap<llvm::StringSet<>> UsingDirectives;
795
Nathan Ridge445195b2020-03-05 19:03:26 -0500796 parseNamespaceEvents(Code, LangOpts, [&](NamespaceEvent Event) {
Kadir Cetinkayad62e3ed2019-09-25 11:35:38 +0200797 llvm::StringRef NS = Event.Payload;
798 switch (Event.Trigger) {
799 case NamespaceEvent::BeginNamespace:
800 case NamespaceEvent::EndNamespace:
801 Current = std::move(Event.Payload);
802 break;
803 case NamespaceEvent::UsingDirective:
804 if (NS.consume_front("::"))
805 UsingDirectives[Current].insert(NS);
806 else {
807 for (llvm::StringRef Enclosing : ancestorNamespaces(Current)) {
808 if (Enclosing.empty())
809 UsingDirectives[Current].insert(NS);
810 else
811 UsingDirectives[Current].insert((Enclosing + "::" + NS).str());
812 }
813 }
814 break;
815 }
816 });
Sam McCallc316b222019-04-26 07:45:49 +0000817
818 std::vector<std::string> Found;
819 for (llvm::StringRef Enclosing : ancestorNamespaces(Current)) {
Benjamin Krameradcd0262020-01-28 20:23:46 +0100820 Found.push_back(std::string(Enclosing));
Sam McCallc316b222019-04-26 07:45:49 +0000821 auto It = UsingDirectives.find(Enclosing);
822 if (It != UsingDirectives.end())
Kadir Cetinkayad62e3ed2019-09-25 11:35:38 +0200823 for (const auto &Used : It->second)
Benjamin Krameradcd0262020-01-28 20:23:46 +0100824 Found.push_back(std::string(Used.getKey()));
Sam McCallc316b222019-04-26 07:45:49 +0000825 }
826
Sam McCallc316b222019-04-26 07:45:49 +0000827 llvm::sort(Found, [&](const std::string &LHS, const std::string &RHS) {
828 if (Current == RHS)
829 return false;
830 if (Current == LHS)
831 return true;
832 return LHS < RHS;
833 });
834 Found.erase(std::unique(Found.begin(), Found.end()), Found.end());
835 return Found;
836}
837
Sam McCall9fb22b22019-05-06 10:25:10 +0000838llvm::StringSet<> collectWords(llvm::StringRef Content) {
839 // We assume short words are not significant.
840 // We may want to consider other stopwords, e.g. language keywords.
841 // (A very naive implementation showed no benefit, but lexing might do better)
842 static constexpr int MinWordLength = 4;
843
844 std::vector<CharRole> Roles(Content.size());
845 calculateRoles(Content, Roles);
846
847 llvm::StringSet<> Result;
848 llvm::SmallString<256> Word;
849 auto Flush = [&] {
850 if (Word.size() >= MinWordLength) {
851 for (char &C : Word)
852 C = llvm::toLower(C);
853 Result.insert(Word);
854 }
855 Word.clear();
856 };
857 for (unsigned I = 0; I < Content.size(); ++I) {
858 switch (Roles[I]) {
859 case Head:
860 Flush();
861 LLVM_FALLTHROUGH;
862 case Tail:
863 Word.push_back(Content[I]);
864 break;
865 case Unknown:
866 case Separator:
867 Flush();
868 break;
869 }
870 }
871 Flush();
872
873 return Result;
874}
875
Sam McCall3f1c2bf2020-03-02 22:45:25 +0100876static bool isLikelyIdentifier(llvm::StringRef Word, llvm::StringRef Before,
877 llvm::StringRef After) {
878 // `foo` is an identifier.
879 if (Before.endswith("`") && After.startswith("`"))
880 return true;
881 // In foo::bar, both foo and bar are identifiers.
882 if (Before.endswith("::") || After.startswith("::"))
883 return true;
884 // Doxygen tags like \c foo indicate identifiers.
885 // Don't search too far back.
886 // This duplicates clang's doxygen parser, revisit if it gets complicated.
887 Before = Before.take_back(100); // Don't search too far back.
888 auto Pos = Before.find_last_of("\\@");
889 if (Pos != llvm::StringRef::npos) {
890 llvm::StringRef Tag = Before.substr(Pos + 1).rtrim(' ');
891 if (Tag == "p" || Tag == "c" || Tag == "class" || Tag == "tparam" ||
892 Tag == "param" || Tag == "param[in]" || Tag == "param[out]" ||
893 Tag == "param[in,out]" || Tag == "retval" || Tag == "throw" ||
894 Tag == "throws" || Tag == "link")
895 return true;
896 }
897
898 // Word contains underscore.
899 // This handles things like snake_case and MACRO_CASE.
900 if (Word.contains('_')) {
901 return true;
902 }
903 // Word contains capital letter other than at beginning.
904 // This handles things like lowerCamel and UpperCamel.
905 // The check for also containing a lowercase letter is to rule out
906 // initialisms like "HTTP".
907 bool HasLower = Word.find_if(clang::isLowercase) != StringRef::npos;
908 bool HasUpper = Word.substr(1).find_if(clang::isUppercase) != StringRef::npos;
909 if (HasLower && HasUpper) {
910 return true;
911 }
912 // FIXME: consider mid-sentence Capitalization?
913 return false;
914}
915
916llvm::Optional<SpelledWord> SpelledWord::touching(SourceLocation SpelledLoc,
917 const syntax::TokenBuffer &TB,
918 const LangOptions &LangOpts) {
919 const auto &SM = TB.sourceManager();
920 auto Touching = syntax::spelledTokensTouching(SpelledLoc, TB);
921 for (const auto &T : Touching) {
922 // If the token is an identifier or a keyword, don't use any heuristics.
923 if (tok::isAnyIdentifier(T.kind()) || tok::getKeywordSpelling(T.kind())) {
924 SpelledWord Result;
925 Result.Location = T.location();
926 Result.Text = T.text(SM);
927 Result.LikelyIdentifier = tok::isAnyIdentifier(T.kind());
928 Result.PartOfSpelledToken = &T;
929 Result.SpelledToken = &T;
930 auto Expanded =
931 TB.expandedTokens(SM.getMacroArgExpandedLocation(T.location()));
932 if (Expanded.size() == 1 && Expanded.front().text(SM) == Result.Text)
933 Result.ExpandedToken = &Expanded.front();
934 return Result;
935 }
936 }
937 FileID File;
938 unsigned Offset;
939 std::tie(File, Offset) = SM.getDecomposedLoc(SpelledLoc);
940 bool Invalid = false;
941 llvm::StringRef Code = SM.getBufferData(File, &Invalid);
942 if (Invalid)
943 return llvm::None;
944 unsigned B = Offset, E = Offset;
945 while (B > 0 && isIdentifierBody(Code[B - 1]))
946 --B;
947 while (E < Code.size() && isIdentifierBody(Code[E]))
948 ++E;
949 if (B == E)
950 return llvm::None;
951
952 SpelledWord Result;
953 Result.Location = SM.getComposedLoc(File, B);
954 Result.Text = Code.slice(B, E);
955 Result.LikelyIdentifier =
956 isLikelyIdentifier(Result.Text, Code.substr(0, B), Code.substr(E)) &&
957 // should not be a keyword
958 tok::isAnyIdentifier(
959 IdentifierTable(LangOpts).get(Result.Text).getTokenID());
960 for (const auto &T : Touching)
961 if (T.location() <= Result.Location)
962 Result.PartOfSpelledToken = &T;
963 return Result;
964}
965
Kadir Cetinkaya3ae2fc72020-02-28 09:25:40 +0100966llvm::Optional<DefinedMacro> locateMacroAt(const syntax::Token &SpelledTok,
Haojian Wu9d34f452019-07-01 09:26:48 +0000967 Preprocessor &PP) {
Kadir Cetinkaya3ae2fc72020-02-28 09:25:40 +0100968 SourceLocation Loc = SpelledTok.location();
Kadir Cetinkayac24c89d2020-02-27 16:02:44 +0100969 assert(Loc.isFileID());
Haojian Wu9d34f452019-07-01 09:26:48 +0000970 const auto &SM = PP.getSourceManager();
Kadir Cetinkaya3ae2fc72020-02-28 09:25:40 +0100971 IdentifierInfo *IdentifierInfo = PP.getIdentifierInfo(SpelledTok.text(SM));
Haojian Wu9d34f452019-07-01 09:26:48 +0000972 if (!IdentifierInfo || !IdentifierInfo->hadMacroDefinition())
973 return None;
974
Haojian Wu9d34f452019-07-01 09:26:48 +0000975 // Get the definition just before the searched location so that a macro
Kadir Cetinkayac24c89d2020-02-27 16:02:44 +0100976 // referenced in a '#undef MACRO' can still be found. Note that we only do
977 // that if Loc is not pointing at start of file.
978 if (SM.getLocForStartOfFile(SM.getFileID(Loc)) != Loc)
979 Loc = Loc.getLocWithOffset(-1);
980 MacroDefinition MacroDef = PP.getMacroDefinitionAtLoc(IdentifierInfo, Loc);
Haojian Wu9d34f452019-07-01 09:26:48 +0000981 if (auto *MI = MacroDef.getMacroInfo())
Kadir Cetinkaya538c2752020-05-14 12:26:47 +0200982 return DefinedMacro{
983 IdentifierInfo->getName(), MI,
984 translatePreamblePatchLocation(MI->getDefinitionLoc(), SM)};
Haojian Wu9d34f452019-07-01 09:26:48 +0000985 return None;
986}
987
Kadir Cetinkaya5b270932019-09-09 12:28:44 +0000988llvm::Expected<std::string> Edit::apply() const {
989 return tooling::applyAllReplacements(InitialCode, Replacements);
990}
991
992std::vector<TextEdit> Edit::asTextEdits() const {
993 return replacementsToEdits(InitialCode, Replacements);
994}
995
996bool Edit::canApplyTo(llvm::StringRef Code) const {
997 // Create line iterators, since line numbers are important while applying our
998 // edit we cannot skip blank lines.
999 auto LHS = llvm::MemoryBuffer::getMemBuffer(Code);
1000 llvm::line_iterator LHSIt(*LHS, /*SkipBlanks=*/false);
1001
1002 auto RHS = llvm::MemoryBuffer::getMemBuffer(InitialCode);
1003 llvm::line_iterator RHSIt(*RHS, /*SkipBlanks=*/false);
1004
1005 // Compare the InitialCode we prepared the edit for with the Code we received
1006 // line by line to make sure there are no differences.
1007 // FIXME: This check is too conservative now, it should be enough to only
1008 // check lines around the replacements contained inside the Edit.
1009 while (!LHSIt.is_at_eof() && !RHSIt.is_at_eof()) {
1010 if (*LHSIt != *RHSIt)
1011 return false;
1012 ++LHSIt;
1013 ++RHSIt;
1014 }
1015
1016 // After we reach EOF for any of the files we make sure the other one doesn't
1017 // contain any additional content except empty lines, they should not
1018 // interfere with the edit we produced.
1019 while (!LHSIt.is_at_eof()) {
1020 if (!LHSIt->empty())
1021 return false;
1022 ++LHSIt;
1023 }
1024 while (!RHSIt.is_at_eof()) {
1025 if (!RHSIt->empty())
1026 return false;
1027 ++RHSIt;
1028 }
1029 return true;
1030}
1031
1032llvm::Error reformatEdit(Edit &E, const format::FormatStyle &Style) {
1033 if (auto NewEdits = cleanupAndFormat(E.InitialCode, E.Replacements, Style))
1034 E.Replacements = std::move(*NewEdits);
1035 else
1036 return NewEdits.takeError();
1037 return llvm::Error::success();
1038}
1039
Kadir Cetinkayad62e3ed2019-09-25 11:35:38 +02001040EligibleRegion getEligiblePoints(llvm::StringRef Code,
1041 llvm::StringRef FullyQualifiedName,
Nathan Ridge445195b2020-03-05 19:03:26 -05001042 const LangOptions &LangOpts) {
Kadir Cetinkayad62e3ed2019-09-25 11:35:38 +02001043 EligibleRegion ER;
1044 // Start with global namespace.
1045 std::vector<std::string> Enclosing = {""};
1046 // FIXME: In addition to namespaces try to generate events for function
1047 // definitions as well. One might use a closing parantheses(")" followed by an
1048 // opening brace "{" to trigger the start.
Nathan Ridge445195b2020-03-05 19:03:26 -05001049 parseNamespaceEvents(Code, LangOpts, [&](NamespaceEvent Event) {
Kadir Cetinkayad62e3ed2019-09-25 11:35:38 +02001050 // Using Directives only introduces declarations to current scope, they do
1051 // not change the current namespace, so skip them.
1052 if (Event.Trigger == NamespaceEvent::UsingDirective)
1053 return;
1054 // Do not qualify the global namespace.
1055 if (!Event.Payload.empty())
1056 Event.Payload.append("::");
1057
1058 std::string CurrentNamespace;
1059 if (Event.Trigger == NamespaceEvent::BeginNamespace) {
1060 Enclosing.emplace_back(std::move(Event.Payload));
1061 CurrentNamespace = Enclosing.back();
1062 // parseNameSpaceEvents reports the beginning position of a token; we want
1063 // to insert after '{', so increment by one.
1064 ++Event.Pos.character;
1065 } else {
1066 // Event.Payload points to outer namespace when exiting a scope, so use
1067 // the namespace we've last entered instead.
1068 CurrentNamespace = std::move(Enclosing.back());
1069 Enclosing.pop_back();
1070 assert(Enclosing.back() == Event.Payload);
1071 }
1072
1073 // Ignore namespaces that are not a prefix of the target.
1074 if (!FullyQualifiedName.startswith(CurrentNamespace))
1075 return;
1076
1077 // Prefer the namespace that shares the longest prefix with target.
1078 if (CurrentNamespace.size() > ER.EnclosingNamespace.size()) {
1079 ER.EligiblePoints.clear();
1080 ER.EnclosingNamespace = CurrentNamespace;
1081 }
1082 if (CurrentNamespace.size() == ER.EnclosingNamespace.size())
1083 ER.EligiblePoints.emplace_back(std::move(Event.Pos));
1084 });
1085 // If there were no shared namespaces just return EOF.
1086 if (ER.EligiblePoints.empty()) {
1087 assert(ER.EnclosingNamespace.empty());
1088 ER.EligiblePoints.emplace_back(offsetToPosition(Code, Code.size()));
1089 }
1090 return ER;
1091}
1092
Haojian Wu509efe52019-11-13 16:30:07 +01001093bool isHeaderFile(llvm::StringRef FileName,
1094 llvm::Optional<LangOptions> LangOpts) {
1095 // Respect the langOpts, for non-file-extension cases, e.g. standard library
1096 // files.
1097 if (LangOpts && LangOpts->IsHeaderFile)
1098 return true;
1099 namespace types = clang::driver::types;
1100 auto Lang = types::lookupTypeForExtension(
1101 llvm::sys::path::extension(FileName).substr(1));
1102 return Lang != types::TY_INVALID && types::onlyPrecompileType(Lang);
1103}
1104
Haojian Wuf8865c02020-02-05 12:03:29 +01001105bool isProtoFile(SourceLocation Loc, const SourceManager &SM) {
1106 auto FileName = SM.getFilename(Loc);
1107 if (!FileName.endswith(".proto.h") && !FileName.endswith(".pb.h"))
1108 return false;
1109 auto FID = SM.getFileID(Loc);
1110 // All proto generated headers should start with this line.
1111 static const char *PROTO_HEADER_COMMENT =
1112 "// Generated by the protocol buffer compiler. DO NOT EDIT!";
1113 // Double check that this is an actual protobuf header.
1114 return SM.getBufferData(FID).startswith(PROTO_HEADER_COMMENT);
1115}
1116
Sam McCallb536a2a2017-12-19 12:23:48 +00001117} // namespace clangd
1118} // namespace clang