| 1 |
//===--- FormatTokenLexer.cpp - Lex FormatTokens -------------*- 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 |
/// \file |
| 10 |
/// This file implements FormatTokenLexer, which tokenizes a source file |
| 11 |
/// into a FormatToken stream suitable for ClangFormat. |
| 12 |
/// |
| 13 |
//===----------------------------------------------------------------------===// |
| 14 |
|
| 15 |
#include "FormatTokenLexer.h" |
| 16 |
#include "FormatToken.h" |
| 17 |
#include "clang/Basic/SourceLocation.h" |
| 18 |
#include "clang/Basic/SourceManager.h" |
| 19 |
#include "clang/Format/Format.h" |
| 20 |
#include "llvm/Support/Regex.h" |
| 21 |
|
| 22 |
namespace clang { |
| 23 |
namespace format { |
| 24 |
|
| 25 |
FormatTokenLexer::FormatTokenLexer( |
| 26 |
const SourceManager &SourceMgr, FileID ID, unsigned Column, |
| 27 |
const FormatStyle &Style, encoding::Encoding Encoding, |
| 28 |
llvm::SpecificBumpPtrAllocator<FormatToken> &Allocator, |
| 29 |
IdentifierTable &IdentTable) |
| 30 |
: FormatTok(nullptr), IsFirstToken(true), StateStack({LexerState::NORMAL}), |
| 31 |
Column(Column), TrailingWhitespace(0), |
| 32 |
LangOpts(getFormattingLangOpts(Style)), SourceMgr(SourceMgr), ID(ID), |
| 33 |
Style(Style), IdentTable(IdentTable), Keywords(IdentTable), |
| 34 |
Encoding(Encoding), Allocator(Allocator), FirstInLineIndex(0), |
| 35 |
FormattingDisabled(false), MacroBlockBeginRegex(Style.MacroBlockBegin), |
| 36 |
MacroBlockEndRegex(Style.MacroBlockEnd) { |
| 37 |
Lex.reset(new Lexer(ID, SourceMgr.getBufferOrFake(ID), SourceMgr, LangOpts)); |
| 38 |
Lex->SetKeepWhitespaceMode(true); |
| 39 |
|
| 40 |
for (const std::string &ForEachMacro : Style.ForEachMacros) { |
| 41 |
auto Identifier = &IdentTable.get(ForEachMacro); |
| 42 |
Macros.insert({Identifier, TT_ForEachMacro}); |
| 43 |
} |
| 44 |
for (const std::string &IfMacro : Style.IfMacros) { |
| 45 |
auto Identifier = &IdentTable.get(IfMacro); |
| 46 |
Macros.insert({Identifier, TT_IfMacro}); |
| 47 |
} |
| 48 |
for (const std::string &AttributeMacro : Style.AttributeMacros) { |
| 49 |
auto Identifier = &IdentTable.get(AttributeMacro); |
| 50 |
Macros.insert({Identifier, TT_AttributeMacro}); |
| 51 |
} |
| 52 |
for (const std::string &StatementMacro : Style.StatementMacros) { |
| 53 |
auto Identifier = &IdentTable.get(StatementMacro); |
| 54 |
Macros.insert({Identifier, TT_StatementMacro}); |
| 55 |
} |
| 56 |
for (const std::string &TypenameMacro : Style.TypenameMacros) { |
| 57 |
auto Identifier = &IdentTable.get(TypenameMacro); |
| 58 |
Macros.insert({Identifier, TT_TypenameMacro}); |
| 59 |
} |
| 60 |
for (const std::string &NamespaceMacro : Style.NamespaceMacros) { |
| 61 |
auto Identifier = &IdentTable.get(NamespaceMacro); |
| 62 |
Macros.insert({Identifier, TT_NamespaceMacro}); |
| 63 |
} |
| 64 |
for (const std::string &WhitespaceSensitiveMacro : |
| 65 |
Style.WhitespaceSensitiveMacros) { |
| 66 |
auto Identifier = &IdentTable.get(WhitespaceSensitiveMacro); |
| 67 |
Macros.insert({Identifier, TT_UntouchableMacroFunc}); |
| 68 |
} |
| 69 |
for (const std::string &StatementAttributeLikeMacro : |
| 70 |
Style.StatementAttributeLikeMacros) { |
| 71 |
auto Identifier = &IdentTable.get(StatementAttributeLikeMacro); |
| 72 |
Macros.insert({Identifier, TT_StatementAttributeLikeMacro}); |
| 73 |
} |
| 74 |
} |
| 75 |
|
| 76 |
ArrayRef<FormatToken *> FormatTokenLexer::lex() { |
| 77 |
assert(Tokens.empty()); |
| 78 |
assert(FirstInLineIndex == 0); |
| 79 |
do { |
| 80 |
Tokens.push_back(getNextToken()); |
| 81 |
if (Style.isJavaScript()) { |
| 82 |
tryParseJSRegexLiteral(); |
| 83 |
handleTemplateStrings(); |
| 84 |
} |
| 85 |
if (Style.Language == FormatStyle::LK_TextProto) |
| 86 |
tryParsePythonComment(); |
| 87 |
tryMergePreviousTokens(); |
| 88 |
if (Style.isCSharp()) { |
| 89 |
// This needs to come after tokens have been merged so that C# |
| 90 |
// string literals are correctly identified. |
| 91 |
handleCSharpVerbatimAndInterpolatedStrings(); |
| 92 |
} |
| 93 |
if (Tokens.back()->NewlinesBefore > 0 || Tokens.back()->IsMultiline) |
| 94 |
FirstInLineIndex = Tokens.size() - 1; |
| 95 |
} while (Tokens.back()->isNot(tok::eof)); |
| 96 |
return Tokens; |
| 97 |
} |
| 98 |
|
| 99 |
void FormatTokenLexer::tryMergePreviousTokens() { |
| 100 |
if (tryMerge_TMacro()) |
| 101 |
return; |
| 102 |
if (tryMergeConflictMarkers()) |
| 103 |
return; |
| 104 |
if (tryMergeLessLess()) |
| 105 |
return; |
| 106 |
if (tryMergeForEach()) |
| 107 |
return; |
| 108 |
if (Style.isCpp() && tryTransformTryUsageForC()) |
| 109 |
return; |
| 110 |
|
| 111 |
if (Style.isJavaScript() || Style.isCSharp()) { |
| 112 |
static const tok::TokenKind NullishCoalescingOperator[] = {tok::question, |
| 113 |
tok::question}; |
| 114 |
static const tok::TokenKind NullPropagatingOperator[] = {tok::question, |
| 115 |
tok::period}; |
| 116 |
static const tok::TokenKind FatArrow[] = {tok::equal, tok::greater}; |
| 117 |
|
| 118 |
if (tryMergeTokens(FatArrow, TT_FatArrow)) |
| 119 |
return; |
| 120 |
if (tryMergeTokens(NullishCoalescingOperator, TT_NullCoalescingOperator)) { |
| 121 |
// Treat like the "||" operator (as opposed to the ternary ?). |
| 122 |
Tokens.back()->Tok.setKind(tok::pipepipe); |
| 123 |
return; |
| 124 |
} |
| 125 |
if (tryMergeTokens(NullPropagatingOperator, TT_NullPropagatingOperator)) { |
| 126 |
// Treat like a regular "." access. |
| 127 |
Tokens.back()->Tok.setKind(tok::period); |
| 128 |
return; |
| 129 |
} |
| 130 |
if (tryMergeNullishCoalescingEqual()) |
| 131 |
return; |
| 132 |
} |
| 133 |
|
| 134 |
if (Style.isCSharp()) { |
| 135 |
static const tok::TokenKind CSharpNullConditionalLSquare[] = { |
| 136 |
tok::question, tok::l_square}; |
| 137 |
|
| 138 |
if (tryMergeCSharpKeywordVariables()) |
| 139 |
return; |
| 140 |
if (tryMergeCSharpStringLiteral()) |
| 141 |
return; |
| 142 |
if (tryTransformCSharpForEach()) |
| 143 |
return; |
| 144 |
if (tryMergeTokens(CSharpNullConditionalLSquare, |
| 145 |
TT_CSharpNullConditionalLSquare)) { |
| 146 |
// Treat like a regular "[" operator. |
| 147 |
Tokens.back()->Tok.setKind(tok::l_square); |
| 148 |
return; |
| 149 |
} |
| 150 |
} |
| 151 |
|
| 152 |
if (tryMergeNSStringLiteral()) |
| 153 |
return; |
| 154 |
|
| 155 |
if (Style.isJavaScript()) { |
| 156 |
static const tok::TokenKind JSIdentity[] = {tok::equalequal, tok::equal}; |
| 157 |
static const tok::TokenKind JSNotIdentity[] = {tok::exclaimequal, |
| 158 |
tok::equal}; |
| 159 |
static const tok::TokenKind JSShiftEqual[] = {tok::greater, tok::greater, |
| 160 |
tok::greaterequal}; |
| 161 |
static const tok::TokenKind JSExponentiation[] = {tok::star, tok::star}; |
| 162 |
static const tok::TokenKind JSExponentiationEqual[] = {tok::star, |
| 163 |
tok::starequal}; |
| 164 |
static const tok::TokenKind JSPipePipeEqual[] = {tok::pipepipe, tok::equal}; |
| 165 |
static const tok::TokenKind JSAndAndEqual[] = {tok::ampamp, tok::equal}; |
| 166 |
|
| 167 |
// FIXME: Investigate what token type gives the correct operator priority. |
| 168 |
if (tryMergeTokens(JSIdentity, TT_BinaryOperator)) |
| 169 |
return; |
| 170 |
if (tryMergeTokens(JSNotIdentity, TT_BinaryOperator)) |
| 171 |
return; |
| 172 |
if (tryMergeTokens(JSShiftEqual, TT_BinaryOperator)) |
| 173 |
return; |
| 174 |
if (tryMergeTokens(JSExponentiation, TT_JsExponentiation)) |
| 175 |
return; |
| 176 |
if (tryMergeTokens(JSExponentiationEqual, TT_JsExponentiationEqual)) { |
| 177 |
Tokens.back()->Tok.setKind(tok::starequal); |
| 178 |
return; |
| 179 |
} |
| 180 |
if (tryMergeTokens(JSAndAndEqual, TT_JsAndAndEqual) || |
| 181 |
tryMergeTokens(JSPipePipeEqual, TT_JsPipePipeEqual)) { |
| 182 |
// Treat like the "=" assignment operator. |
| 183 |
Tokens.back()->Tok.setKind(tok::equal); |
| 184 |
return; |
| 185 |
} |
| 186 |
if (tryMergeJSPrivateIdentifier()) |
| 187 |
return; |
| 188 |
} |
| 189 |
|
| 190 |
if (Style.Language == FormatStyle::LK_Java) { |
| 191 |
static const tok::TokenKind JavaRightLogicalShiftAssign[] = { |
| 192 |
tok::greater, tok::greater, tok::greaterequal}; |
| 193 |
if (tryMergeTokens(JavaRightLogicalShiftAssign, TT_BinaryOperator)) |
| 194 |
return; |
| 195 |
} |
| 196 |
|
| 197 |
if (Style.isVerilog()) { |
| 198 |
// Merge the number following a base like `'h?a0`. |
| 199 |
if (Tokens.size() >= 3 && Tokens.end()[-3]->is(TT_VerilogNumberBase) && |
| 200 |
Tokens.end()[-2]->is(tok::numeric_constant) && |
| 201 |
Tokens.back()->isOneOf(tok::numeric_constant, tok::identifier, |
| 202 |
tok::question) && |
| 203 |
tryMergeTokens(2, TT_Unknown)) { |
| 204 |
return; |
| 205 |
} |
| 206 |
// Part select. |
| 207 |
if (tryMergeTokensAny({{tok::minus, tok::colon}, {tok::plus, tok::colon}}, |
| 208 |
TT_BitFieldColon)) { |
| 209 |
return; |
| 210 |
} |
| 211 |
// Xnor. The combined token is treated as a caret which can also be either a |
| 212 |
// unary or binary operator. The actual type is determined in |
| 213 |
// TokenAnnotator. We also check the token length so we know it is not |
| 214 |
// already a merged token. |
| 215 |
if (Tokens.back()->TokenText.size() == 1 && |
| 216 |
tryMergeTokensAny({{tok::caret, tok::tilde}, {tok::tilde, tok::caret}}, |
| 217 |
TT_BinaryOperator)) { |
| 218 |
Tokens.back()->Tok.setKind(tok::caret); |
| 219 |
return; |
| 220 |
} |
| 221 |
// Signed shift and distribution weight. |
| 222 |
if (tryMergeTokens({tok::less, tok::less}, TT_BinaryOperator)) { |
| 223 |
Tokens.back()->Tok.setKind(tok::lessless); |
| 224 |
return; |
| 225 |
} |
| 226 |
if (tryMergeTokens({tok::greater, tok::greater}, TT_BinaryOperator)) { |
| 227 |
Tokens.back()->Tok.setKind(tok::greatergreater); |
| 228 |
return; |
| 229 |
} |
| 230 |
if (tryMergeTokensAny({{tok::lessless, tok::equal}, |
| 231 |
{tok::lessless, tok::lessequal}, |
| 232 |
{tok::greatergreater, tok::equal}, |
| 233 |
{tok::greatergreater, tok::greaterequal}, |
| 234 |
{tok::colon, tok::equal}, |
| 235 |
{tok::colon, tok::slash}}, |
| 236 |
TT_BinaryOperator)) { |
| 237 |
Tokens.back()->ForcedPrecedence = prec::Assignment; |
| 238 |
return; |
| 239 |
} |
| 240 |
// Exponentiation, signed shift, case equality, and wildcard equality. |
| 241 |
if (tryMergeTokensAny({{tok::star, tok::star}, |
| 242 |
{tok::lessless, tok::less}, |
| 243 |
{tok::greatergreater, tok::greater}, |
| 244 |
{tok::exclaimequal, tok::equal}, |
| 245 |
{tok::exclaimequal, tok::question}, |
| 246 |
{tok::equalequal, tok::equal}, |
| 247 |
{tok::equalequal, tok::question}}, |
| 248 |
TT_BinaryOperator)) { |
| 249 |
return; |
| 250 |
} |
| 251 |
// Module paths in specify blocks and implications in properties. |
| 252 |
if (tryMergeTokensAny({{tok::plusequal, tok::greater}, |
| 253 |
{tok::plus, tok::star, tok::greater}, |
| 254 |
{tok::minusequal, tok::greater}, |
| 255 |
{tok::minus, tok::star, tok::greater}, |
| 256 |
{tok::less, tok::arrow}, |
| 257 |
{tok::equal, tok::greater}, |
| 258 |
{tok::star, tok::greater}, |
| 259 |
{tok::pipeequal, tok::greater}, |
| 260 |
{tok::pipe, tok::arrow}, |
| 261 |
{tok::hash, tok::minus, tok::hash}, |
| 262 |
{tok::hash, tok::equal, tok::hash}}, |
| 263 |
TT_BinaryOperator)) { |
| 264 |
Tokens.back()->ForcedPrecedence = prec::Comma; |
| 265 |
return; |
| 266 |
} |
| 267 |
} |
| 268 |
} |
| 269 |
|
| 270 |
bool FormatTokenLexer::tryMergeNSStringLiteral() { |
| 271 |
if (Tokens.size() < 2) |
| 272 |
return false; |
| 273 |
auto &At = *(Tokens.end() - 2); |
| 274 |
auto &String = *(Tokens.end() - 1); |
| 275 |
if (!At->is(tok::at) || !String->is(tok::string_literal)) |
| 276 |
return false; |
| 277 |
At->Tok.setKind(tok::string_literal); |
| 278 |
At->TokenText = StringRef(At->TokenText.begin(), |
| 279 |
String->TokenText.end() - At->TokenText.begin()); |
| 280 |
At->ColumnWidth += String->ColumnWidth; |
| 281 |
At->setType(TT_ObjCStringLiteral); |
| 282 |
Tokens.erase(Tokens.end() - 1); |
| 283 |
return true; |
| 284 |
} |
| 285 |
|
| 286 |
bool FormatTokenLexer::tryMergeJSPrivateIdentifier() { |
| 287 |
// Merges #idenfier into a single identifier with the text #identifier |
| 288 |
// but the token tok::identifier. |
| 289 |
if (Tokens.size() < 2) |
| 290 |
return false; |
| 291 |
auto &Hash = *(Tokens.end() - 2); |
| 292 |
auto &Identifier = *(Tokens.end() - 1); |
| 293 |
if (!Hash->is(tok::hash) || !Identifier->is(tok::identifier)) |
| 294 |
return false; |
| 295 |
Hash->Tok.setKind(tok::identifier); |
| 296 |
Hash->TokenText = |
| 297 |
StringRef(Hash->TokenText.begin(), |
| 298 |
Identifier->TokenText.end() - Hash->TokenText.begin()); |
| 299 |
Hash->ColumnWidth += Identifier->ColumnWidth; |
| 300 |
Hash->setType(TT_JsPrivateIdentifier); |
| 301 |
Tokens.erase(Tokens.end() - 1); |
| 302 |
return true; |
| 303 |
} |
| 304 |
|
| 305 |
// Search for verbatim or interpolated string literals @"ABC" or |
| 306 |
// $"aaaaa{abc}aaaaa" i and mark the token as TT_CSharpStringLiteral, and to |
| 307 |
// prevent splitting of @, $ and ". |
| 308 |
// Merging of multiline verbatim strings with embedded '"' is handled in |
| 309 |
// handleCSharpVerbatimAndInterpolatedStrings with lower-level lexing. |
| 310 |
bool FormatTokenLexer::tryMergeCSharpStringLiteral() { |
| 311 |
if (Tokens.size() < 2) |
| 312 |
return false; |
| 313 |
|
| 314 |
// Look for @"aaaaaa" or $"aaaaaa". |
| 315 |
const auto String = *(Tokens.end() - 1); |
| 316 |
if (String->isNot(tok::string_literal)) |
| 317 |
return false; |
| 318 |
|
| 319 |
auto Prefix = *(Tokens.end() - 2); |
| 320 |
if (Prefix->isNot(tok::at) && Prefix->TokenText != "$") |
| 321 |
return false; |
| 322 |
|
| 323 |
if (Tokens.size() > 2) { |
| 324 |
const auto Tok = *(Tokens.end() - 3); |
| 325 |
if ((Tok->TokenText == "$" && Prefix->is(tok::at)) || |
| 326 |
(Tok->is(tok::at) && Prefix->TokenText == "$")) { |
| 327 |
// This looks like $@"aaa" or @$"aaa" so we need to combine all 3 tokens. |
| 328 |
Tok->ColumnWidth += Prefix->ColumnWidth; |
| 329 |
Tokens.erase(Tokens.end() - 2); |
| 330 |
Prefix = Tok; |
| 331 |
} |
| 332 |
} |
| 333 |
|
| 334 |
// Convert back into just a string_literal. |
| 335 |
Prefix->Tok.setKind(tok::string_literal); |
| 336 |
Prefix->TokenText = |
| 337 |
StringRef(Prefix->TokenText.begin(), |
| 338 |
String->TokenText.end() - Prefix->TokenText.begin()); |
| 339 |
Prefix->ColumnWidth += String->ColumnWidth; |
| 340 |
Prefix->setType(TT_CSharpStringLiteral); |
| 341 |
Tokens.erase(Tokens.end() - 1); |
| 342 |
return true; |
| 343 |
} |
| 344 |
|
| 345 |
// Valid C# attribute targets: |
| 346 |
// https://docs.microsoft.com/en-us/dotnet/csharp/programming-guide/concepts/attributes/#attribute-targets |
| 347 |
const llvm::StringSet<> FormatTokenLexer::CSharpAttributeTargets = { |
| 348 |
"assembly", "module", "field", "event", "method", |
| 349 |
"param", "property", "return", "type", |
| 350 |
}; |
| 351 |
|
| 352 |
bool FormatTokenLexer::tryMergeNullishCoalescingEqual() { |
| 353 |
if (Tokens.size() < 2) |
| 354 |
return false; |
| 355 |
auto &NullishCoalescing = *(Tokens.end() - 2); |
| 356 |
auto &Equal = *(Tokens.end() - 1); |
| 357 |
if (NullishCoalescing->getType() != TT_NullCoalescingOperator || |
| 358 |
!Equal->is(tok::equal)) { |
| 359 |
return false; |
| 360 |
} |
| 361 |
NullishCoalescing->Tok.setKind(tok::equal); // no '??=' in clang tokens. |
| 362 |
NullishCoalescing->TokenText = |
| 363 |
StringRef(NullishCoalescing->TokenText.begin(), |
| 364 |
Equal->TokenText.end() - NullishCoalescing->TokenText.begin()); |
| 365 |
NullishCoalescing->ColumnWidth += Equal->ColumnWidth; |
| 366 |
NullishCoalescing->setType(TT_NullCoalescingEqual); |
| 367 |
Tokens.erase(Tokens.end() - 1); |
| 368 |
return true; |
| 369 |
} |
| 370 |
|
| 371 |
bool FormatTokenLexer::tryMergeCSharpKeywordVariables() { |
| 372 |
if (Tokens.size() < 2) |
| 373 |
return false; |
| 374 |
const auto At = *(Tokens.end() - 2); |
| 375 |
if (At->isNot(tok::at)) |
| 376 |
return false; |
| 377 |
const auto Keyword = *(Tokens.end() - 1); |
| 378 |
if (Keyword->TokenText == "$") |
| 379 |
return false; |
| 380 |
if (!Keywords.isCSharpKeyword(*Keyword)) |
| 381 |
return false; |
| 382 |
|
| 383 |
At->Tok.setKind(tok::identifier); |
| 384 |
At->TokenText = StringRef(At->TokenText.begin(), |
| 385 |
Keyword->TokenText.end() - At->TokenText.begin()); |
| 386 |
At->ColumnWidth += Keyword->ColumnWidth; |
| 387 |
At->setType(Keyword->getType()); |
| 388 |
Tokens.erase(Tokens.end() - 1); |
| 389 |
return true; |
| 390 |
} |
| 391 |
|
| 392 |
// In C# transform identifier foreach into kw_foreach |
| 393 |
bool FormatTokenLexer::tryTransformCSharpForEach() { |
| 394 |
if (Tokens.size() < 1) |
| 395 |
return false; |
| 396 |
auto &Identifier = *(Tokens.end() - 1); |
| 397 |
if (!Identifier->is(tok::identifier)) |
| 398 |
return false; |
| 399 |
if (Identifier->TokenText != "foreach") |
| 400 |
return false; |
| 401 |
|
| 402 |
Identifier->setType(TT_ForEachMacro); |
| 403 |
Identifier->Tok.setKind(tok::kw_for); |
| 404 |
return true; |
| 405 |
} |
| 406 |
|
| 407 |
bool FormatTokenLexer::tryMergeForEach() { |
| 408 |
if (Tokens.size() < 2) |
| 409 |
return false; |
| 410 |
auto &For = *(Tokens.end() - 2); |
| 411 |
auto &Each = *(Tokens.end() - 1); |
| 412 |
if (!For->is(tok::kw_for)) |
| 413 |
return false; |
| 414 |
if (!Each->is(tok::identifier)) |
| 415 |
return false; |
| 416 |
if (Each->TokenText != "each") |
| 417 |
return false; |
| 418 |
|
| 419 |
For->setType(TT_ForEachMacro); |
| 420 |
For->Tok.setKind(tok::kw_for); |
| 421 |
|
| 422 |
For->TokenText = StringRef(For->TokenText.begin(), |
| 423 |
Each->TokenText.end() - For->TokenText.begin()); |
| 424 |
For->ColumnWidth += Each->ColumnWidth; |
| 425 |
Tokens.erase(Tokens.end() - 1); |
| 426 |
return true; |
| 427 |
} |
| 428 |
|
| 429 |
bool FormatTokenLexer::tryTransformTryUsageForC() { |
| 430 |
if (Tokens.size() < 2) |
| 431 |
return false; |
| 432 |
auto &Try = *(Tokens.end() - 2); |
| 433 |
if (!Try->is(tok::kw_try)) |
| 434 |
return false; |
| 435 |
auto &Next = *(Tokens.end() - 1); |
| 436 |
if (Next->isOneOf(tok::l_brace, tok::colon, tok::hash, tok::comment)) |
| 437 |
return false; |
| 438 |
|
| 439 |
if (Tokens.size() > 2) { |
| 440 |
auto &At = *(Tokens.end() - 3); |
| 441 |
if (At->is(tok::at)) |
| 442 |
return false; |
| 443 |
} |
| 444 |
|
| 445 |
Try->Tok.setKind(tok::identifier); |
| 446 |
return true; |
| 447 |
} |
| 448 |
|
| 449 |
bool FormatTokenLexer::tryMergeLessLess() { |
| 450 |
// Merge X,less,less,Y into X,lessless,Y unless X or Y is less. |
| 451 |
if (Tokens.size() < 3) |
| 452 |
return false; |
| 453 |
|
| 454 |
auto First = Tokens.end() - 3; |
| 455 |
if (First[0]->isNot(tok::less) || First[1]->isNot(tok::less)) |
| 456 |
return false; |
| 457 |
|
| 458 |
// Only merge if there currently is no whitespace between the two "<". |
| 459 |
if (First[1]->hasWhitespaceBefore()) |
| 460 |
return false; |
| 461 |
|
| 462 |
auto X = Tokens.size() > 3 ? First[-1] : nullptr; |
| 463 |
auto Y = First[2]; |
| 464 |
if ((X && X->is(tok::less)) || Y->is(tok::less)) |
| 465 |
return false; |
| 466 |
|
| 467 |
// Do not remove a whitespace between the two "<" e.g. "operator< <>". |
| 468 |
if (X && X->is(tok::kw_operator) && Y->is(tok::greater)) |
| 469 |
return false; |
| 470 |
|
| 471 |
First[0]->Tok.setKind(tok::lessless); |
| 472 |
First[0]->TokenText = "<<"; |
| 473 |
First[0]->ColumnWidth += 1; |
| 474 |
Tokens.erase(Tokens.end() - 2); |
| 475 |
return true; |
| 476 |
} |
| 477 |
|
| 478 |
bool FormatTokenLexer::tryMergeTokens(ArrayRef<tok::TokenKind> Kinds, |
| 479 |
TokenType NewType) { |
| 480 |
if (Tokens.size() < Kinds.size()) |
| 481 |
return false; |
| 482 |
|
| 483 |
SmallVectorImpl<FormatToken *>::const_iterator First = |
| 484 |
Tokens.end() - Kinds.size(); |
| 485 |
for (unsigned i = 0; i < Kinds.size(); ++i) |
| 486 |
if (!First[i]->is(Kinds[i])) |
| 487 |
return false; |
| 488 |
|
| 489 |
return tryMergeTokens(Kinds.size(), NewType); |
| 490 |
} |
| 491 |
|
| 492 |
bool FormatTokenLexer::tryMergeTokens(size_t Count, TokenType NewType) { |
| 493 |
if (Tokens.size() < Count) |
| 494 |
return false; |
| 495 |
|
| 496 |
SmallVectorImpl<FormatToken *>::const_iterator First = Tokens.end() - Count; |
| 497 |
unsigned AddLength = 0; |
| 498 |
for (size_t i = 1; i < Count; ++i) { |
| 499 |
// If there is whitespace separating the token and the previous one, |
| 500 |
// they should not be merged. |
| 501 |
if (First[i]->hasWhitespaceBefore()) |
| 502 |
return false; |
| 503 |
AddLength += First[i]->TokenText.size(); |
| 504 |
} |
| 505 |
|
| 506 |
Tokens.resize(Tokens.size() - Count + 1); |
| 507 |
First[0]->TokenText = StringRef(First[0]->TokenText.data(), |
| 508 |
First[0]->TokenText.size() + AddLength); |
| 509 |
First[0]->ColumnWidth += AddLength; |
| 510 |
First[0]->setType(NewType); |
| 511 |
return true; |
| 512 |
} |
| 513 |
|
| 514 |
bool FormatTokenLexer::tryMergeTokensAny( |
| 515 |
ArrayRef<ArrayRef<tok::TokenKind>> Kinds, TokenType NewType) { |
| 516 |
return llvm::any_of(Kinds, [this, NewType](ArrayRef<tok::TokenKind> Kinds) { |
| 517 |
return tryMergeTokens(Kinds, NewType); |
| 518 |
}); |
| 519 |
} |
| 520 |
|
| 521 |
// Returns \c true if \p Tok can only be followed by an operand in JavaScript. |
| 522 |
bool FormatTokenLexer::precedesOperand(FormatToken *Tok) { |
| 523 |
// NB: This is not entirely correct, as an r_paren can introduce an operand |
| 524 |
// location in e.g. `if (foo) /bar/.exec(...);`. That is a rare enough |
| 525 |
// corner case to not matter in practice, though. |
| 526 |
return Tok->isOneOf(tok::period, tok::l_paren, tok::comma, tok::l_brace, |
| 527 |
tok::r_brace, tok::l_square, tok::semi, tok::exclaim, |
| 528 |
tok::colon, tok::question, tok::tilde) || |
| 529 |
Tok->isOneOf(tok::kw_return, tok::kw_do, tok::kw_case, tok::kw_throw, |
| 530 |
tok::kw_else, tok::kw_new, tok::kw_delete, tok::kw_void, |
| 531 |
tok::kw_typeof, Keywords.kw_instanceof, Keywords.kw_in) || |
| 532 |
Tok->isBinaryOperator(); |
| 533 |
} |
| 534 |
|
| 535 |
bool FormatTokenLexer::canPrecedeRegexLiteral(FormatToken *Prev) { |
| 536 |
if (!Prev) |
| 537 |
return true; |
| 538 |
|
| 539 |
// Regex literals can only follow after prefix unary operators, not after |
| 540 |
// postfix unary operators. If the '++' is followed by a non-operand |
| 541 |
// introducing token, the slash here is the operand and not the start of a |
| 542 |
// regex. |
| 543 |
// `!` is an unary prefix operator, but also a post-fix operator that casts |
| 544 |
// away nullability, so the same check applies. |
| 545 |
if (Prev->isOneOf(tok::plusplus, tok::minusminus, tok::exclaim)) |
| 546 |
return Tokens.size() < 3 || precedesOperand(Tokens[Tokens.size() - 3]); |
| 547 |
|
| 548 |
// The previous token must introduce an operand location where regex |
| 549 |
// literals can occur. |
| 550 |
if (!precedesOperand(Prev)) |
| 551 |
return false; |
| 552 |
|
| 553 |
return true; |
| 554 |
} |
| 555 |
|
| 556 |
// Tries to parse a JavaScript Regex literal starting at the current token, |
| 557 |
// if that begins with a slash and is in a location where JavaScript allows |
| 558 |
// regex literals. Changes the current token to a regex literal and updates |
| 559 |
// its text if successful. |
| 560 |
void FormatTokenLexer::tryParseJSRegexLiteral() { |
| 561 |
FormatToken *RegexToken = Tokens.back(); |
| 562 |
if (!RegexToken->isOneOf(tok::slash, tok::slashequal)) |
| 563 |
return; |
| 564 |
|
| 565 |
FormatToken *Prev = nullptr; |
| 566 |
for (FormatToken *FT : llvm::drop_begin(llvm::reverse(Tokens))) { |
| 567 |
// NB: Because previous pointers are not initialized yet, this cannot use |
| 568 |
// Token.getPreviousNonComment. |
| 569 |
if (FT->isNot(tok::comment)) { |
| 570 |
Prev = FT; |
| 571 |
break; |
| 572 |
} |
| 573 |
} |
| 574 |
|
| 575 |
if (!canPrecedeRegexLiteral(Prev)) |
| 576 |
return; |
| 577 |
|
| 578 |
// 'Manually' lex ahead in the current file buffer. |
| 579 |
const char *Offset = Lex->getBufferLocation(); |
| 580 |
const char *RegexBegin = Offset - RegexToken->TokenText.size(); |
| 581 |
StringRef Buffer = Lex->getBuffer(); |
| 582 |
bool InCharacterClass = false; |
| 583 |
bool HaveClosingSlash = false; |
| 584 |
for (; !HaveClosingSlash && Offset != Buffer.end(); ++Offset) { |
| 585 |
// Regular expressions are terminated with a '/', which can only be |
| 586 |
// escaped using '\' or a character class between '[' and ']'. |
| 587 |
// See http://www.ecma-international.org/ecma-262/5.1/#sec-7.8.5. |
| 588 |
switch (*Offset) { |
| 589 |
case '\\': |
| 590 |
// Skip the escaped character. |
| 591 |
++Offset; |
| 592 |
break; |
| 593 |
case '[': |
| 594 |
InCharacterClass = true; |
| 595 |
break; |
| 596 |
case ']': |
| 597 |
InCharacterClass = false; |
| 598 |
break; |
| 599 |
case '/': |
| 600 |
if (!InCharacterClass) |
| 601 |
HaveClosingSlash = true; |
| 602 |
break; |
| 603 |
} |
| 604 |
} |
| 605 |
|
| 606 |
RegexToken->setType(TT_RegexLiteral); |
| 607 |
// Treat regex literals like other string_literals. |
| 608 |
RegexToken->Tok.setKind(tok::string_literal); |
| 609 |
RegexToken->TokenText = StringRef(RegexBegin, Offset - RegexBegin); |
| 610 |
RegexToken->ColumnWidth = RegexToken->TokenText.size(); |
| 611 |
|
| 612 |
resetLexer(SourceMgr.getFileOffset(Lex->getSourceLocation(Offset))); |
| 613 |
} |
| 614 |
|
| 615 |
static auto lexCSharpString(const char *Begin, const char *End, bool Verbatim, |
| 616 |
bool Interpolated) { |
| 617 |
auto Repeated = [&Begin, End]() { |
| 618 |
return Begin + 1 < End && Begin[1] == Begin[0]; |
| 619 |
}; |
| 620 |
|
| 621 |
// Look for a terminating '"' in the current file buffer. |
| 622 |
// Make no effort to format code within an interpolated or verbatim string. |
| 623 |
// |
| 624 |
// Interpolated strings could contain { } with " characters inside. |
| 625 |
// $"{x ?? "null"}" |
| 626 |
// should not be split into $"{x ?? ", null, "}" but should be treated as a |
| 627 |
// single string-literal. |
| 628 |
// |
| 629 |
// We opt not to try and format expressions inside {} within a C# |
| 630 |
// interpolated string. Formatting expressions within an interpolated string |
| 631 |
// would require similar work as that done for JavaScript template strings |
| 632 |
// in `handleTemplateStrings()`. |
| 633 |
for (int UnmatchedOpeningBraceCount = 0; Begin < End; ++Begin) { |
| 634 |
switch (*Begin) { |
| 635 |
case '\\': |
| 636 |
if (!Verbatim) |
| 637 |
++Begin; |
| 638 |
break; |
| 639 |
case '{': |
| 640 |
if (Interpolated) { |
| 641 |
// {{ inside an interpolated string is escaped, so skip it. |
| 642 |
if (Repeated()) |
| 643 |
++Begin; |
| 644 |
else |
| 645 |
++UnmatchedOpeningBraceCount; |
| 646 |
} |
| 647 |
break; |
| 648 |
case '}': |
| 649 |
if (Interpolated) { |
| 650 |
// }} inside an interpolated string is escaped, so skip it. |
| 651 |
if (Repeated()) |
| 652 |
++Begin; |
| 653 |
else if (UnmatchedOpeningBraceCount > 0) |
| 654 |
--UnmatchedOpeningBraceCount; |
| 655 |
else |
| 656 |
return End; |
| 657 |
} |
| 658 |
break; |
| 659 |
case '"': |
| 660 |
if (UnmatchedOpeningBraceCount > 0) |
| 661 |
break; |
| 662 |
// "" within a verbatim string is an escaped double quote: skip it. |
| 663 |
if (Verbatim && Repeated()) { |
| 664 |
++Begin; |
| 665 |
break; |
| 666 |
} |
| 667 |
return Begin; |
| 668 |
} |
| 669 |
} |
| 670 |
|
| 671 |
return End; |
| 672 |
} |
| 673 |
|
| 674 |
void FormatTokenLexer::handleCSharpVerbatimAndInterpolatedStrings() { |
| 675 |
FormatToken *CSharpStringLiteral = Tokens.back(); |
| 676 |
|
| 677 |
if (CSharpStringLiteral->isNot(TT_CSharpStringLiteral)) |
| 678 |
return; |
| 679 |
|
| 680 |
auto &TokenText = CSharpStringLiteral->TokenText; |
| 681 |
|
| 682 |
bool Verbatim = false; |
| 683 |
bool Interpolated = false; |
| 684 |
if (TokenText.startswith(R"($@")") || TokenText.startswith(R"(@$")")) { |
| 685 |
Verbatim = true; |
| 686 |
Interpolated = true; |
| 687 |
} else if (TokenText.startswith(R"(@")")) { |
| 688 |
Verbatim = true; |
| 689 |
} else if (TokenText.startswith(R"($")")) { |
| 690 |
Interpolated = true; |
| 691 |
} |
| 692 |
|
| 693 |
// Deal with multiline strings. |
| 694 |
if (!Verbatim && !Interpolated) |
| 695 |
return; |
| 696 |
|
| 697 |
const char *StrBegin = Lex->getBufferLocation() - TokenText.size(); |
| 698 |
const char *Offset = StrBegin; |
| 699 |
if (Verbatim && Interpolated) |
| 700 |
Offset += 3; |
| 701 |
else |
| 702 |
Offset += 2; |
| 703 |
|
| 704 |
const auto End = Lex->getBuffer().end(); |
| 705 |
Offset = lexCSharpString(Offset, End, Verbatim, Interpolated); |
| 706 |
|
| 707 |
// Make no attempt to format code properly if a verbatim string is |
| 708 |
// unterminated. |
| 709 |
if (Offset >= End) |
| 710 |
return; |
| 711 |
|
| 712 |
StringRef LiteralText(StrBegin, Offset - StrBegin + 1); |
| 713 |
TokenText = LiteralText; |
| 714 |
|
| 715 |
// Adjust width for potentially multiline string literals. |
| 716 |
size_t FirstBreak = LiteralText.find('\n'); |
| 717 |
StringRef FirstLineText = FirstBreak == StringRef::npos |
| 718 |
? LiteralText |
| 719 |
: LiteralText.substr(0, FirstBreak); |
| 720 |
CSharpStringLiteral->ColumnWidth = encoding::columnWidthWithTabs( |
| 721 |
FirstLineText, CSharpStringLiteral->OriginalColumn, Style.TabWidth, |
| 722 |
Encoding); |
| 723 |
size_t LastBreak = LiteralText.rfind('\n'); |
| 724 |
if (LastBreak != StringRef::npos) { |
| 725 |
CSharpStringLiteral->IsMultiline = true; |
| 726 |
unsigned StartColumn = 0; |
| 727 |
CSharpStringLiteral->LastLineColumnWidth = |
| 728 |
encoding::columnWidthWithTabs(LiteralText.substr(LastBreak + 1), |
| 729 |
StartColumn, Style.TabWidth, Encoding); |
| 730 |
} |
| 731 |
|
| 732 |
assert(Offset < End); |
| 733 |
resetLexer(SourceMgr.getFileOffset(Lex->getSourceLocation(Offset + 1))); |
| 734 |
} |
| 735 |
|
| 736 |
void FormatTokenLexer::handleTemplateStrings() { |
| 737 |
FormatToken *BacktickToken = Tokens.back(); |
| 738 |
|
| 739 |
if (BacktickToken->is(tok::l_brace)) { |
| 740 |
StateStack.push(LexerState::NORMAL); |
| 741 |
return; |
| 742 |
} |
| 743 |
if (BacktickToken->is(tok::r_brace)) { |
| 744 |
if (StateStack.size() == 1) |
| 745 |
return; |
| 746 |
StateStack.pop(); |
| 747 |
if (StateStack.top() != LexerState::TEMPLATE_STRING) |
| 748 |
return; |
| 749 |
// If back in TEMPLATE_STRING, fallthrough and continue parsing the |
| 750 |
} else if (BacktickToken->is(tok::unknown) && |
| 751 |
BacktickToken->TokenText == "`") { |
| 752 |
StateStack.push(LexerState::TEMPLATE_STRING); |
| 753 |
} else { |
| 754 |
return; // Not actually a template |
| 755 |
} |
| 756 |
|
| 757 |
// 'Manually' lex ahead in the current file buffer. |
| 758 |
const char *Offset = Lex->getBufferLocation(); |
| 759 |
const char *TmplBegin = Offset - BacktickToken->TokenText.size(); // at "`" |
| 760 |
for (; Offset != Lex->getBuffer().end(); ++Offset) { |
| 761 |
if (Offset[0] == '`') { |
| 762 |
StateStack.pop(); |
| 763 |
++Offset; |
| 764 |
break; |
| 765 |
} |
| 766 |
if (Offset[0] == '\\') { |
| 767 |
++Offset; // Skip the escaped character. |
| 768 |
} else if (Offset + 1 < Lex->getBuffer().end() && Offset[0] == '$' && |
| 769 |
Offset[1] == '{') { |
| 770 |
// '${' introduces an expression interpolation in the template string. |
| 771 |
StateStack.push(LexerState::NORMAL); |
| 772 |
Offset += 2; |
| 773 |
break; |
| 774 |
} |
| 775 |
} |
| 776 |
|
| 777 |
StringRef LiteralText(TmplBegin, Offset - TmplBegin); |
| 778 |
BacktickToken->setType(TT_TemplateString); |
| 779 |
BacktickToken->Tok.setKind(tok::string_literal); |
| 780 |
BacktickToken->TokenText = LiteralText; |
| 781 |
|
| 782 |
// Adjust width for potentially multiline string literals. |
| 783 |
size_t FirstBreak = LiteralText.find('\n'); |
| 784 |
StringRef FirstLineText = FirstBreak == StringRef::npos |
| 785 |
? LiteralText |
| 786 |
: LiteralText.substr(0, FirstBreak); |
| 787 |
BacktickToken->ColumnWidth = encoding::columnWidthWithTabs( |
| 788 |
FirstLineText, BacktickToken->OriginalColumn, Style.TabWidth, Encoding); |
| 789 |
size_t LastBreak = LiteralText.rfind('\n'); |
| 790 |
if (LastBreak != StringRef::npos) { |
| 791 |
BacktickToken->IsMultiline = true; |
| 792 |
unsigned StartColumn = 0; // The template tail spans the entire line. |
| 793 |
BacktickToken->LastLineColumnWidth = |
| 794 |
encoding::columnWidthWithTabs(LiteralText.substr(LastBreak + 1), |
| 795 |
StartColumn, Style.TabWidth, Encoding); |
| 796 |
} |
| 797 |
|
| 798 |
SourceLocation loc = Lex->getSourceLocation(Offset); |
| 799 |
resetLexer(SourceMgr.getFileOffset(loc)); |
| 800 |
} |
| 801 |
|
| 802 |
void FormatTokenLexer::tryParsePythonComment() { |
| 803 |
FormatToken *HashToken = Tokens.back(); |
| 804 |
if (!HashToken->isOneOf(tok::hash, tok::hashhash)) |
| 805 |
return; |
| 806 |
// Turn the remainder of this line into a comment. |
| 807 |
const char *CommentBegin = |
| 808 |
Lex->getBufferLocation() - HashToken->TokenText.size(); // at "#" |
| 809 |
size_t From = CommentBegin - Lex->getBuffer().begin(); |
| 810 |
size_t To = Lex->getBuffer().find_first_of('\n', From); |
| 811 |
if (To == StringRef::npos) |
| 812 |
To = Lex->getBuffer().size(); |
| 813 |
size_t Len = To - From; |
| 814 |
HashToken->setType(TT_LineComment); |
| 815 |
HashToken->Tok.setKind(tok::comment); |
| 816 |
HashToken->TokenText = Lex->getBuffer().substr(From, Len); |
| 817 |
SourceLocation Loc = To < Lex->getBuffer().size() |
| 818 |
? Lex->getSourceLocation(CommentBegin + Len) |
| 819 |
: SourceMgr.getLocForEndOfFile(ID); |
| 820 |
resetLexer(SourceMgr.getFileOffset(Loc)); |
| 821 |
} |
| 822 |
|
| 823 |
bool FormatTokenLexer::tryMerge_TMacro() { |
| 824 |
if (Tokens.size() < 4) |
| 825 |
return false; |
| 826 |
FormatToken *Last = Tokens.back(); |
| 827 |
if (!Last->is(tok::r_paren)) |
| 828 |
return false; |
| 829 |
|
| 830 |
FormatToken *String = Tokens[Tokens.size() - 2]; |
| 831 |
if (!String->is(tok::string_literal) || String->IsMultiline) |
| 832 |
return false; |
| 833 |
|
| 834 |
if (!Tokens[Tokens.size() - 3]->is(tok::l_paren)) |
| 835 |
return false; |
| 836 |
|
| 837 |
FormatToken *Macro = Tokens[Tokens.size() - 4]; |
| 838 |
if (Macro->TokenText != "_T") |
| 839 |
return false; |
| 840 |
|
| 841 |
const char *Start = Macro->TokenText.data(); |
| 842 |
const char *End = Last->TokenText.data() + Last->TokenText.size(); |
| 843 |
String->TokenText = StringRef(Start, End - Start); |
| 844 |
String->IsFirst = Macro->IsFirst; |
| 845 |
String->LastNewlineOffset = Macro->LastNewlineOffset; |
| 846 |
String->WhitespaceRange = Macro->WhitespaceRange; |
| 847 |
String->OriginalColumn = Macro->OriginalColumn; |
| 848 |
String->ColumnWidth = encoding::columnWidthWithTabs( |
| 849 |
String->TokenText, String->OriginalColumn, Style.TabWidth, Encoding); |
| 850 |
String->NewlinesBefore = Macro->NewlinesBefore; |
| 851 |
String->HasUnescapedNewline = Macro->HasUnescapedNewline; |
| 852 |
|
| 853 |
Tokens.pop_back(); |
| 854 |
Tokens.pop_back(); |
| 855 |
Tokens.pop_back(); |
| 856 |
Tokens.back() = String; |
| 857 |
if (FirstInLineIndex >= Tokens.size()) |
| 858 |
FirstInLineIndex = Tokens.size() - 1; |
| 859 |
return true; |
| 860 |
} |
| 861 |
|
| 862 |
bool FormatTokenLexer::tryMergeConflictMarkers() { |
| 863 |
if (Tokens.back()->NewlinesBefore == 0 && Tokens.back()->isNot(tok::eof)) |
| 864 |
return false; |
| 865 |
|
| 866 |
// Conflict lines look like: |
| 867 |
// <marker> <text from the vcs> |
| 868 |
// For example: |
| 869 |
// >>>>>>> /file/in/file/system at revision 1234 |
| 870 |
// |
| 871 |
// We merge all tokens in a line that starts with a conflict marker |
| 872 |
// into a single token with a special token type that the unwrapped line |
| 873 |
// parser will use to correctly rebuild the underlying code. |
| 874 |
|
| 875 |
FileID ID; |
| 876 |
// Get the position of the first token in the line. |
| 877 |
unsigned FirstInLineOffset; |
| 878 |
std::tie(ID, FirstInLineOffset) = SourceMgr.getDecomposedLoc( |
| 879 |
Tokens[FirstInLineIndex]->getStartOfNonWhitespace()); |
| 880 |
StringRef Buffer = SourceMgr.getBufferOrFake(ID).getBuffer(); |
| 881 |
// Calculate the offset of the start of the current line. |
| 882 |
auto LineOffset = Buffer.rfind('\n', FirstInLineOffset); |
| 883 |
if (LineOffset == StringRef::npos) |
| 884 |
LineOffset = 0; |
| 885 |
else |
| 886 |
++LineOffset; |
| 887 |
|
| 888 |
auto FirstSpace = Buffer.find_first_of(" \n", LineOffset); |
| 889 |
StringRef LineStart; |
| 890 |
if (FirstSpace == StringRef::npos) |
| 891 |
LineStart = Buffer.substr(LineOffset); |
| 892 |
else |
| 893 |
LineStart = Buffer.substr(LineOffset, FirstSpace - LineOffset); |
| 894 |
|
| 895 |
TokenType Type = TT_Unknown; |
| 896 |
if (LineStart == "<<<<<<<" || LineStart == ">>>>") { |
| 897 |
Type = TT_ConflictStart; |
| 898 |
} else if (LineStart == "|||||||" || LineStart == "=======" || |
| 899 |
LineStart == "====") { |
| 900 |
Type = TT_ConflictAlternative; |
| 901 |
} else if (LineStart == ">>>>>>>" || LineStart == "<<<<") { |
| 902 |
Type = TT_ConflictEnd; |
| 903 |
} |
| 904 |
|
| 905 |
if (Type != TT_Unknown) { |
| 906 |
FormatToken *Next = Tokens.back(); |
| 907 |
|
| 908 |
Tokens.resize(FirstInLineIndex + 1); |
| 909 |
// We do not need to build a complete token here, as we will skip it |
| 910 |
// during parsing anyway (as we must not touch whitespace around conflict |
| 911 |
// markers). |
| 912 |
Tokens.back()->setType(Type); |
| 913 |
Tokens.back()->Tok.setKind(tok::kw___unknown_anytype); |
| 914 |
|
| 915 |
Tokens.push_back(Next); |
| 916 |
return true; |
| 917 |
} |
| 918 |
|
| 919 |
return false; |
| 920 |
} |
| 921 |
|
| 922 |
FormatToken *FormatTokenLexer::getStashedToken() { |
| 923 |
// Create a synthesized second '>' or '<' token. |
| 924 |
Token Tok = FormatTok->Tok; |
| 925 |
StringRef TokenText = FormatTok->TokenText; |
| 926 |
|
| 927 |
unsigned OriginalColumn = FormatTok->OriginalColumn; |
| 928 |
FormatTok = new (Allocator.Allocate()) FormatToken; |
| 929 |
FormatTok->Tok = Tok; |
| 930 |
SourceLocation TokLocation = |
| 931 |
FormatTok->Tok.getLocation().getLocWithOffset(Tok.getLength() - 1); |
| 932 |
FormatTok->Tok.setLocation(TokLocation); |
| 933 |
FormatTok->WhitespaceRange = SourceRange(TokLocation, TokLocation); |
| 934 |
FormatTok->TokenText = TokenText; |
| 935 |
FormatTok->ColumnWidth = 1; |
| 936 |
FormatTok->OriginalColumn = OriginalColumn + 1; |
| 937 |
|
| 938 |
return FormatTok; |
| 939 |
} |
| 940 |
|
| 941 |
/// Truncate the current token to the new length and make the lexer continue |
| 942 |
/// from the end of the truncated token. Used for other languages that have |
| 943 |
/// different token boundaries, like JavaScript in which a comment ends at a |
| 944 |
/// line break regardless of whether the line break follows a backslash. Also |
| 945 |
/// used to set the lexer to the end of whitespace if the lexer regards |
| 946 |
/// whitespace and an unrecognized symbol as one token. |
| 947 |
void FormatTokenLexer::truncateToken(size_t NewLen) { |
| 948 |
assert(NewLen <= FormatTok->TokenText.size()); |
| 949 |
resetLexer(SourceMgr.getFileOffset(Lex->getSourceLocation( |
| 950 |
Lex->getBufferLocation() - FormatTok->TokenText.size() + NewLen))); |
| 951 |
FormatTok->TokenText = FormatTok->TokenText.substr(0, NewLen); |
| 952 |
FormatTok->ColumnWidth = encoding::columnWidthWithTabs( |
| 953 |
FormatTok->TokenText, FormatTok->OriginalColumn, Style.TabWidth, |
| 954 |
Encoding); |
| 955 |
FormatTok->Tok.setLength(NewLen); |
| 956 |
} |
| 957 |
|
| 958 |
/// Count the length of leading whitespace in a token. |
| 959 |
static size_t countLeadingWhitespace(StringRef Text) { |
| 960 |
// Basically counting the length matched by this regex. |
| 961 |
// "^([\n\r\f\v \t]|(\\\\|\\?\\?/)[\n\r])+" |
| 962 |
// Directly using the regex turned out to be slow. With the regex |
| 963 |
// version formatting all files in this directory took about 1.25 |
| 964 |
// seconds. This version took about 0.5 seconds. |
| 965 |
const unsigned char *const Begin = Text.bytes_begin(); |
| 966 |
const unsigned char *const End = Text.bytes_end(); |
| 967 |
const unsigned char *Cur = Begin; |
| 968 |
while (Cur < End) { |
| 969 |
if (isspace(Cur[0])) { |
| 970 |
++Cur; |
| 971 |
} else if (Cur[0] == '\\' && (Cur[1] == '\n' || Cur[1] == '\r')) { |
| 972 |
// A '\' followed by a newline always escapes the newline, regardless |
| 973 |
// of whether there is another '\' before it. |
| 974 |
// The source has a null byte at the end. So the end of the entire input |
| 975 |
// isn't reached yet. Also the lexer doesn't break apart an escaped |
| 976 |
// newline. |
| 977 |
assert(End - Cur >= 2); |
| 978 |
Cur += 2; |
| 979 |
} else if (Cur[0] == '?' && Cur[1] == '?' && Cur[2] == '/' && |
| 980 |
(Cur[3] == '\n' || Cur[3] == '\r')) { |
| 981 |
// Newlines can also be escaped by a '?' '?' '/' trigraph. By the way, the |
| 982 |
// characters are quoted individually in this comment because if we write |
| 983 |
// them together some compilers warn that we have a trigraph in the code. |
| 984 |
assert(End - Cur >= 4); |
| 985 |
Cur += 4; |
| 986 |
} else { |
| 987 |
break; |
| 988 |
} |
| 989 |
} |
| 990 |
return Cur - Begin; |
| 991 |
} |
| 992 |
|
| 993 |
FormatToken *FormatTokenLexer::getNextToken() { |
| 994 |
if (StateStack.top() == LexerState::TOKEN_STASHED) { |
| 995 |
StateStack.pop(); |
| 996 |
return getStashedToken(); |
| 997 |
} |
| 998 |
|
| 999 |
FormatTok = new (Allocator.Allocate()) FormatToken; |
| 1000 |
readRawToken(*FormatTok); |
| 1001 |
SourceLocation WhitespaceStart = |
| 1002 |
FormatTok->Tok.getLocation().getLocWithOffset(-TrailingWhitespace); |
| 1003 |
FormatTok->IsFirst = IsFirstToken; |
| 1004 |
IsFirstToken = false; |
| 1005 |
|
| 1006 |
// Consume and record whitespace until we find a significant token. |
| 1007 |
// Some tok::unknown tokens are not just whitespace, e.g. whitespace |
| 1008 |
// followed by a symbol such as backtick. Those symbols may be |
| 1009 |
// significant in other languages. |
| 1010 |
unsigned WhitespaceLength = TrailingWhitespace; |
| 1011 |
while (FormatTok->isNot(tok::eof)) { |
| 1012 |
auto LeadingWhitespace = countLeadingWhitespace(FormatTok->TokenText); |
| 1013 |
if (LeadingWhitespace == 0) |
| 1014 |
break; |
| 1015 |
if (LeadingWhitespace < FormatTok->TokenText.size()) |
| 1016 |
truncateToken(LeadingWhitespace); |
| 1017 |
StringRef Text = FormatTok->TokenText; |
| 1018 |
bool InEscape = false; |
| 1019 |
for (int i = 0, e = Text.size(); i != e; ++i) { |
| 1020 |
switch (Text[i]) { |
| 1021 |
case '\r': |
| 1022 |
// If this is a CRLF sequence, break here and the LF will be handled on |
| 1023 |
// the next loop iteration. Otherwise, this is a single Mac CR, treat it |
| 1024 |
// the same as a single LF. |
| 1025 |
if (i + 1 < e && Text[i + 1] == '\n') |
| 1026 |
break; |
| 1027 |
[[fallthrough]]; |
| 1028 |
case '\n': |
| 1029 |
++FormatTok->NewlinesBefore; |
| 1030 |
if (!InEscape) |
| 1031 |
FormatTok->HasUnescapedNewline = true; |
| 1032 |
else |
| 1033 |
InEscape = false; |
| 1034 |
FormatTok->LastNewlineOffset = WhitespaceLength + i + 1; |
| 1035 |
Column = 0; |
| 1036 |
break; |
| 1037 |
case '\f': |
| 1038 |
case '\v': |
| 1039 |
Column = 0; |
| 1040 |
break; |
| 1041 |
case ' ': |
| 1042 |
++Column; |
| 1043 |
break; |
| 1044 |
case '\t': |
| 1045 |
Column += |
| 1046 |
Style.TabWidth - (Style.TabWidth ? Column % Style.TabWidth : 0); |
| 1047 |
break; |
| 1048 |
case '\\': |
| 1049 |
case '?': |
| 1050 |
case '/': |
| 1051 |
// The text was entirely whitespace when this loop was entered. Thus |
| 1052 |
// this has to be an escape sequence. |
| 1053 |
assert(Text.substr(i, 2) == "\\\r" || Text.substr(i, 2) == "\\\n" || |
| 1054 |
Text.substr(i, 4) == "\?\?/\r" || |
| 1055 |
Text.substr(i, 4) == "\?\?/\n" || |
| 1056 |
(i >= 1 && (Text.substr(i - 1, 4) == "\?\?/\r" || |
| 1057 |
Text.substr(i - 1, 4) == "\?\?/\n")) || |
| 1058 |
(i >= 2 && (Text.substr(i - 2, 4) == "\?\?/\r" || |
| 1059 |
Text.substr(i - 2, 4) == "\?\?/\n"))); |
| 1060 |
InEscape = true; |
| 1061 |
break; |
| 1062 |
default: |
| 1063 |
// This shouldn't happen. |
| 1064 |
assert(false); |
| 1065 |
break; |
| 1066 |
} |
| 1067 |
} |
| 1068 |
WhitespaceLength += Text.size(); |
| 1069 |
readRawToken(*FormatTok); |
| 1070 |
} |
| 1071 |
|
| 1072 |
if (FormatTok->is(tok::unknown)) |
| 1073 |
FormatTok->setType(TT_ImplicitStringLiteral); |
| 1074 |
|
| 1075 |
// JavaScript and Java do not allow to escape the end of the line with a |
| 1076 |
// backslash. Backslashes are syntax errors in plain source, but can occur in |
| 1077 |
// comments. When a single line comment ends with a \, it'll cause the next |
| 1078 |
// line of code to be lexed as a comment, breaking formatting. The code below |
| 1079 |
// finds comments that contain a backslash followed by a line break, truncates |
| 1080 |
// the comment token at the backslash, and resets the lexer to restart behind |
| 1081 |
// the backslash. |
| 1082 |
if ((Style.isJavaScript() || Style.Language == FormatStyle::LK_Java) && |
| 1083 |
FormatTok->is(tok::comment) && FormatTok->TokenText.startswith("//")) { |
| 1084 |
size_t BackslashPos = FormatTok->TokenText.find('\\'); |
| 1085 |
while (BackslashPos != StringRef::npos) { |
| 1086 |
if (BackslashPos + 1 < FormatTok->TokenText.size() && |
| 1087 |
FormatTok->TokenText[BackslashPos + 1] == '\n') { |
| 1088 |
truncateToken(BackslashPos + 1); |
| 1089 |
break; |
| 1090 |
} |
| 1091 |
BackslashPos = FormatTok->TokenText.find('\\', BackslashPos + 1); |
| 1092 |
} |
| 1093 |
} |
| 1094 |
|
| 1095 |
if (Style.isVerilog()) { |
| 1096 |
static const llvm::Regex NumberBase("^s?[bdho]", llvm::Regex::IgnoreCase); |
| 1097 |
SmallVector<StringRef, 1> Matches; |
| 1098 |
// Verilog uses the backtick instead of the hash for preprocessor stuff. |
| 1099 |
// And it uses the hash for delays and parameter lists. In order to continue |
| 1100 |
// using `tok::hash` in other places, the backtick gets marked as the hash |
| 1101 |
// here. And in order to tell the backtick and hash apart for |
| 1102 |
// Verilog-specific stuff, the hash becomes an identifier. |
| 1103 |
if (FormatTok->is(tok::numeric_constant)) { |
| 1104 |
// In Verilog the quote is not part of a number. |
| 1105 |
auto Quote = FormatTok->TokenText.find('\''); |
| 1106 |
if (Quote != StringRef::npos) |
| 1107 |
truncateToken(Quote); |
| 1108 |
} else if (FormatTok->isOneOf(tok::hash, tok::hashhash)) { |
| 1109 |
FormatTok->Tok.setKind(tok::raw_identifier); |
| 1110 |
} else if (FormatTok->is(tok::raw_identifier)) { |
| 1111 |
if (FormatTok->TokenText == "`") { |
| 1112 |
FormatTok->Tok.setIdentifierInfo(nullptr); |
| 1113 |
FormatTok->Tok.setKind(tok::hash); |
| 1114 |
} else if (FormatTok->TokenText == "``") { |
| 1115 |
FormatTok->Tok.setIdentifierInfo(nullptr); |
| 1116 |
FormatTok->Tok.setKind(tok::hashhash); |
| 1117 |
} else if (Tokens.size() > 0 && |
| 1118 |
Tokens.back()->is(Keywords.kw_apostrophe) && |
| 1119 |
NumberBase.match(FormatTok->TokenText, &Matches)) { |
| 1120 |
// In Verilog in a based number literal like `'b10`, there may be |
| 1121 |
// whitespace between `'b` and `10`. Therefore we handle the base and |
| 1122 |
// the rest of the number literal as two tokens. But if there is no |
| 1123 |
// space in the input code, we need to manually separate the two parts. |
| 1124 |
truncateToken(Matches[0].size()); |
| 1125 |
FormatTok->setFinalizedType(TT_VerilogNumberBase); |
| 1126 |
} |
| 1127 |
} |
| 1128 |
} |
| 1129 |
|
| 1130 |
FormatTok->WhitespaceRange = SourceRange( |
| 1131 |
WhitespaceStart, WhitespaceStart.getLocWithOffset(WhitespaceLength)); |
| 1132 |
|
| 1133 |
FormatTok->OriginalColumn = Column; |
| 1134 |
|
| 1135 |
TrailingWhitespace = 0; |
| 1136 |
if (FormatTok->is(tok::comment)) { |
| 1137 |
// FIXME: Add the trimmed whitespace to Column. |
| 1138 |
StringRef UntrimmedText = FormatTok->TokenText; |
| 1139 |
FormatTok->TokenText = FormatTok->TokenText.rtrim(" \t\v\f"); |
| 1140 |
TrailingWhitespace = UntrimmedText.size() - FormatTok->TokenText.size(); |
| 1141 |
} else if (FormatTok->is(tok::raw_identifier)) { |
| 1142 |
IdentifierInfo &Info = IdentTable.get(FormatTok->TokenText); |
| 1143 |
FormatTok->Tok.setIdentifierInfo(&Info); |
| 1144 |
FormatTok->Tok.setKind(Info.getTokenID()); |
| 1145 |
if (Style.Language == FormatStyle::LK_Java && |
| 1146 |
FormatTok->isOneOf(tok::kw_struct, tok::kw_union, tok::kw_delete, |
| 1147 |
tok::kw_operator)) { |
| 1148 |
FormatTok->Tok.setKind(tok::identifier); |
| 1149 |
FormatTok->Tok.setIdentifierInfo(nullptr); |
| 1150 |
} else if (Style.isJavaScript() && |
| 1151 |
FormatTok->isOneOf(tok::kw_struct, tok::kw_union, |
| 1152 |
tok::kw_operator)) { |
| 1153 |
FormatTok->Tok.setKind(tok::identifier); |
| 1154 |
FormatTok->Tok.setIdentifierInfo(nullptr); |
| 1155 |
} |
| 1156 |
} else if (FormatTok->is(tok::greatergreater)) { |
| 1157 |
FormatTok->Tok.setKind(tok::greater); |
| 1158 |
FormatTok->TokenText = FormatTok->TokenText.substr(0, 1); |
| 1159 |
++Column; |
| 1160 |
StateStack.push(LexerState::TOKEN_STASHED); |
| 1161 |
} else if (FormatTok->is(tok::lessless)) { |
| 1162 |
FormatTok->Tok.setKind(tok::less); |
| 1163 |
FormatTok->TokenText = FormatTok->TokenText.substr(0, 1); |
| 1164 |
++Column; |
| 1165 |
StateStack.push(LexerState::TOKEN_STASHED); |
| 1166 |
} |
| 1167 |
|
| 1168 |
if (Style.isVerilog() && Tokens.size() > 0 && |
| 1169 |
Tokens.back()->is(TT_VerilogNumberBase) && |
| 1170 |
FormatTok->Tok.isOneOf(tok::identifier, tok::question)) { |
| 1171 |
// Mark the number following a base like `'h?a0` as a number. |
| 1172 |
FormatTok->Tok.setKind(tok::numeric_constant); |
| 1173 |
} |
| 1174 |
|
| 1175 |
// Now FormatTok is the next non-whitespace token. |
| 1176 |
|
| 1177 |
StringRef Text = FormatTok->TokenText; |
| 1178 |
size_t FirstNewlinePos = Text.find('\n'); |
| 1179 |
if (FirstNewlinePos == StringRef::npos) { |
| 1180 |
// FIXME: ColumnWidth actually depends on the start column, we need to |
| 1181 |
// take this into account when the token is moved. |
| 1182 |
FormatTok->ColumnWidth = |
| 1183 |
encoding::columnWidthWithTabs(Text, Column, Style.TabWidth, Encoding); |
| 1184 |
Column += FormatTok->ColumnWidth; |
| 1185 |
} else { |
| 1186 |
FormatTok->IsMultiline = true; |
| 1187 |
// FIXME: ColumnWidth actually depends on the start column, we need to |
| 1188 |
// take this into account when the token is moved. |
| 1189 |
FormatTok->ColumnWidth = encoding::columnWidthWithTabs( |
| 1190 |
Text.substr(0, FirstNewlinePos), Column, Style.TabWidth, Encoding); |
| 1191 |
|
| 1192 |
// The last line of the token always starts in column 0. |
| 1193 |
// Thus, the length can be precomputed even in the presence of tabs. |
| 1194 |
FormatTok->LastLineColumnWidth = encoding::columnWidthWithTabs( |
| 1195 |
Text.substr(Text.find_last_of('\n') + 1), 0, Style.TabWidth, Encoding); |
| 1196 |
Column = FormatTok->LastLineColumnWidth; |
| 1197 |
} |
| 1198 |
|
| 1199 |
if (Style.isCpp()) { |
| 1200 |
auto it = Macros.find(FormatTok->Tok.getIdentifierInfo()); |
| 1201 |
if (!(Tokens.size() > 0 && Tokens.back()->Tok.getIdentifierInfo() && |
| 1202 |
Tokens.back()->Tok.getIdentifierInfo()->getPPKeywordID() == |
| 1203 |
tok::pp_define) && |
| 1204 |
it != Macros.end()) { |
| 1205 |
FormatTok->setType(it->second); |
| 1206 |
if (it->second == TT_IfMacro) { |
| 1207 |
// The lexer token currently has type tok::kw_unknown. However, for this |
| 1208 |
// substitution to be treated correctly in the TokenAnnotator, faking |
| 1209 |
// the tok value seems to be needed. Not sure if there's a more elegant |
| 1210 |
// way. |
| 1211 |
FormatTok->Tok.setKind(tok::kw_if); |
| 1212 |
} |
| 1213 |
} else if (FormatTok->is(tok::identifier)) { |
| 1214 |
if (MacroBlockBeginRegex.match(Text)) |
| 1215 |
FormatTok->setType(TT_MacroBlockBegin); |
| 1216 |
else if (MacroBlockEndRegex.match(Text)) |
| 1217 |
FormatTok->setType(TT_MacroBlockEnd); |
| 1218 |
} |
| 1219 |
} |
| 1220 |
|
| 1221 |
return FormatTok; |
| 1222 |
} |
| 1223 |
|
| 1224 |
bool FormatTokenLexer::readRawTokenVerilogSpecific(Token &Tok) { |
| 1225 |
// In Verilog the quote is not a character literal. |
| 1226 |
// |
| 1227 |
// Make the backtick and double backtick identifiers to match against them |
| 1228 |
// more easily. |
| 1229 |
// |
| 1230 |
// In Verilog an escaped identifier starts with backslash and ends with |
| 1231 |
// whitespace. Unless that whitespace is an escaped newline. A backslash can |
| 1232 |
// also begin an escaped newline outside of an escaped identifier. We check |
| 1233 |
// for that outside of the Regex since we can't use negative lookhead |
| 1234 |
// assertions. Simply changing the '*' to '+' breaks stuff as the escaped |
| 1235 |
// identifier may have a length of 0 according to Section A.9.3. |
| 1236 |
// FIXME: If there is an escaped newline in the middle of an escaped |
| 1237 |
// identifier, allow for pasting the two lines together, But escaped |
| 1238 |
// identifiers usually occur only in generated code anyway. |
| 1239 |
static const llvm::Regex VerilogToken(R"re(^('|``?|\\(\\)re" |
| 1240 |
"(\r?\n|\r)|[^[:space:]])*)"); |
| 1241 |
|
| 1242 |
SmallVector<StringRef, 4> Matches; |
| 1243 |
const char *Start = Lex->getBufferLocation(); |
| 1244 |
if (!VerilogToken.match(StringRef(Start, Lex->getBuffer().end() - Start), |
| 1245 |
&Matches)) { |
| 1246 |
return false; |
| 1247 |
} |
| 1248 |
// There is a null byte at the end of the buffer, so we don't have to check |
| 1249 |
// Start[1] is within the buffer. |
| 1250 |
if (Start[0] == '\\' && (Start[1] == '\r' || Start[1] == '\n')) |
| 1251 |
return false; |
| 1252 |
size_t Len = Matches[0].size(); |
| 1253 |
|
| 1254 |
// The kind has to be an identifier so we can match it against those defined |
| 1255 |
// in Keywords. The kind has to be set before the length because the setLength |
| 1256 |
// function checks that the kind is not an annotation. |
| 1257 |
Tok.setKind(tok::raw_identifier); |
| 1258 |
Tok.setLength(Len); |
| 1259 |
Tok.setLocation(Lex->getSourceLocation(Start, Len)); |
| 1260 |
Tok.setRawIdentifierData(Start); |
| 1261 |
Lex->seek(Lex->getCurrentBufferOffset() + Len, /*IsAtStartofline=*/false); |
| 1262 |
return true; |
| 1263 |
} |
| 1264 |
|
| 1265 |
void FormatTokenLexer::readRawToken(FormatToken &Tok) { |
| 1266 |
// For Verilog, first see if there is a special token, and fall back to the |
| 1267 |
// normal lexer if there isn't one. |
| 1268 |
if (!Style.isVerilog() || !readRawTokenVerilogSpecific(Tok.Tok)) |
| 1269 |
Lex->LexFromRawLexer(Tok.Tok); |
| 1270 |
Tok.TokenText = StringRef(SourceMgr.getCharacterData(Tok.Tok.getLocation()), |
| 1271 |
Tok.Tok.getLength()); |
| 1272 |
// For formatting, treat unterminated string literals like normal string |
| 1273 |
// literals. |
| 1274 |
if (Tok.is(tok::unknown)) { |
| 1275 |
if (!Tok.TokenText.empty() && Tok.TokenText[0] == '"') { |
| 1276 |
Tok.Tok.setKind(tok::string_literal); |
| 1277 |
Tok.IsUnterminatedLiteral = true; |
| 1278 |
} else if (Style.isJavaScript() && Tok.TokenText == "''") { |
| 1279 |
Tok.Tok.setKind(tok::string_literal); |
| 1280 |
} |
| 1281 |
} |
| 1282 |
|
| 1283 |
if ((Style.isJavaScript() || Style.Language == FormatStyle::LK_Proto || |
| 1284 |
Style.Language == FormatStyle::LK_TextProto) && |
| 1285 |
Tok.is(tok::char_constant)) { |
| 1286 |
Tok.Tok.setKind(tok::string_literal); |
| 1287 |
} |
| 1288 |
|
| 1289 |
if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format on" || |
| 1290 |
Tok.TokenText == "/* clang-format on */")) { |
| 1291 |
FormattingDisabled = false; |
| 1292 |
} |
| 1293 |
|
| 1294 |
Tok.Finalized = FormattingDisabled; |
| 1295 |
|
| 1296 |
if (Tok.is(tok::comment) && (Tok.TokenText == "// clang-format off" || |
| 1297 |
Tok.TokenText == "/* clang-format off */")) { |
| 1298 |
FormattingDisabled = true; |
| 1299 |
} |
| 1300 |
} |
| 1301 |
|
| 1302 |
void FormatTokenLexer::resetLexer(unsigned Offset) { |
| 1303 |
StringRef Buffer = SourceMgr.getBufferData(ID); |
| 1304 |
LangOpts = getFormattingLangOpts(Style); |
| 1305 |
Lex.reset(new Lexer(SourceMgr.getLocForStartOfFile(ID), LangOpts, |
| 1306 |
Buffer.begin(), Buffer.begin() + Offset, Buffer.end())); |
| 1307 |
Lex->SetKeepWhitespaceMode(true); |
| 1308 |
TrailingWhitespace = 0; |
| 1309 |
} |
| 1310 |
|
| 1311 |
} // namespace format |
| 1312 |
} // namespace clang |