1 //===--- WhitespaceManager.cpp - Format C++ code --------------------------===// 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 WhitespaceManager class. 11 /// 12 //===----------------------------------------------------------------------===// 13 14 #include "WhitespaceManager.h" 15 #include "llvm/ADT/STLExtras.h" 16 #include "llvm/ADT/SmallVector.h" 17 #include <algorithm> 18 19 namespace clang { 20 namespace format { 21 22 bool WhitespaceManager::Change::IsBeforeInFile::operator()( 23 const Change &C1, const Change &C2) const { 24 return SourceMgr.isBeforeInTranslationUnit( 25 C1.OriginalWhitespaceRange.getBegin(), 26 C2.OriginalWhitespaceRange.getBegin()); 27 } 28 29 WhitespaceManager::Change::Change(const FormatToken &Tok, 30 bool CreateReplacement, 31 SourceRange OriginalWhitespaceRange, 32 int Spaces, unsigned StartOfTokenColumn, 33 unsigned NewlinesBefore, 34 StringRef PreviousLinePostfix, 35 StringRef CurrentLinePrefix, bool IsAligned, 36 bool ContinuesPPDirective, bool IsInsideToken) 37 : Tok(&Tok), CreateReplacement(CreateReplacement), 38 OriginalWhitespaceRange(OriginalWhitespaceRange), 39 StartOfTokenColumn(StartOfTokenColumn), NewlinesBefore(NewlinesBefore), 40 PreviousLinePostfix(PreviousLinePostfix), 41 CurrentLinePrefix(CurrentLinePrefix), IsAligned(IsAligned), 42 ContinuesPPDirective(ContinuesPPDirective), Spaces(Spaces), 43 IsInsideToken(IsInsideToken), IsTrailingComment(false), TokenLength(0), 44 PreviousEndOfTokenColumn(0), EscapedNewlineColumn(0), 45 StartOfBlockComment(nullptr), IndentationOffset(0), ConditionalsLevel(0) { 46 } 47 48 void WhitespaceManager::replaceWhitespace(FormatToken &Tok, unsigned Newlines, 49 unsigned Spaces, 50 unsigned StartOfTokenColumn, 51 bool IsAligned, bool InPPDirective) { 52 if (Tok.Finalized) 53 return; 54 Tok.setDecision((Newlines > 0) ? FD_Break : FD_Continue); 55 Changes.push_back(Change(Tok, /*CreateReplacement=*/true, Tok.WhitespaceRange, 56 Spaces, StartOfTokenColumn, Newlines, "", "", 57 IsAligned, InPPDirective && !Tok.IsFirst, 58 /*IsInsideToken=*/false)); 59 } 60 61 void WhitespaceManager::addUntouchableToken(const FormatToken &Tok, 62 bool InPPDirective) { 63 if (Tok.Finalized) 64 return; 65 Changes.push_back(Change(Tok, /*CreateReplacement=*/false, 66 Tok.WhitespaceRange, /*Spaces=*/0, 67 Tok.OriginalColumn, Tok.NewlinesBefore, "", "", 68 /*IsAligned=*/false, InPPDirective && !Tok.IsFirst, 69 /*IsInsideToken=*/false)); 70 } 71 72 llvm::Error 73 WhitespaceManager::addReplacement(const tooling::Replacement &Replacement) { 74 return Replaces.add(Replacement); 75 } 76 77 void WhitespaceManager::replaceWhitespaceInToken( 78 const FormatToken &Tok, unsigned Offset, unsigned ReplaceChars, 79 StringRef PreviousPostfix, StringRef CurrentPrefix, bool InPPDirective, 80 unsigned Newlines, int Spaces) { 81 if (Tok.Finalized) 82 return; 83 SourceLocation Start = Tok.getStartOfNonWhitespace().getLocWithOffset(Offset); 84 Changes.push_back( 85 Change(Tok, /*CreateReplacement=*/true, 86 SourceRange(Start, Start.getLocWithOffset(ReplaceChars)), Spaces, 87 std::max(0, Spaces), Newlines, PreviousPostfix, CurrentPrefix, 88 /*IsAligned=*/true, InPPDirective && !Tok.IsFirst, 89 /*IsInsideToken=*/true)); 90 } 91 92 const tooling::Replacements &WhitespaceManager::generateReplacements() { 93 if (Changes.empty()) 94 return Replaces; 95 96 llvm::sort(Changes, Change::IsBeforeInFile(SourceMgr)); 97 calculateLineBreakInformation(); 98 alignConsecutiveMacros(); 99 alignConsecutiveDeclarations(); 100 alignConsecutiveBitFields(); 101 alignConsecutiveAssignments(); 102 alignChainedConditionals(); 103 alignTrailingComments(); 104 alignEscapedNewlines(); 105 alignArrayInitializers(); 106 generateChanges(); 107 108 return Replaces; 109 } 110 111 void WhitespaceManager::calculateLineBreakInformation() { 112 Changes[0].PreviousEndOfTokenColumn = 0; 113 Change *LastOutsideTokenChange = &Changes[0]; 114 for (unsigned i = 1, e = Changes.size(); i != e; ++i) { 115 SourceLocation OriginalWhitespaceStart = 116 Changes[i].OriginalWhitespaceRange.getBegin(); 117 SourceLocation PreviousOriginalWhitespaceEnd = 118 Changes[i - 1].OriginalWhitespaceRange.getEnd(); 119 unsigned OriginalWhitespaceStartOffset = 120 SourceMgr.getFileOffset(OriginalWhitespaceStart); 121 unsigned PreviousOriginalWhitespaceEndOffset = 122 SourceMgr.getFileOffset(PreviousOriginalWhitespaceEnd); 123 assert(PreviousOriginalWhitespaceEndOffset <= 124 OriginalWhitespaceStartOffset); 125 const char *const PreviousOriginalWhitespaceEndData = 126 SourceMgr.getCharacterData(PreviousOriginalWhitespaceEnd); 127 StringRef Text(PreviousOriginalWhitespaceEndData, 128 SourceMgr.getCharacterData(OriginalWhitespaceStart) - 129 PreviousOriginalWhitespaceEndData); 130 // Usually consecutive changes would occur in consecutive tokens. This is 131 // not the case however when analyzing some preprocessor runs of the 132 // annotated lines. For example, in this code: 133 // 134 // #if A // line 1 135 // int i = 1; 136 // #else B // line 2 137 // int i = 2; 138 // #endif // line 3 139 // 140 // one of the runs will produce the sequence of lines marked with line 1, 2 141 // and 3. So the two consecutive whitespace changes just before '// line 2' 142 // and before '#endif // line 3' span multiple lines and tokens: 143 // 144 // #else B{change X}[// line 2 145 // int i = 2; 146 // ]{change Y}#endif // line 3 147 // 148 // For this reason, if the text between consecutive changes spans multiple 149 // newlines, the token length must be adjusted to the end of the original 150 // line of the token. 151 auto NewlinePos = Text.find_first_of('\n'); 152 if (NewlinePos == StringRef::npos) { 153 Changes[i - 1].TokenLength = OriginalWhitespaceStartOffset - 154 PreviousOriginalWhitespaceEndOffset + 155 Changes[i].PreviousLinePostfix.size() + 156 Changes[i - 1].CurrentLinePrefix.size(); 157 } else { 158 Changes[i - 1].TokenLength = 159 NewlinePos + Changes[i - 1].CurrentLinePrefix.size(); 160 } 161 162 // If there are multiple changes in this token, sum up all the changes until 163 // the end of the line. 164 if (Changes[i - 1].IsInsideToken && Changes[i - 1].NewlinesBefore == 0) 165 LastOutsideTokenChange->TokenLength += 166 Changes[i - 1].TokenLength + Changes[i - 1].Spaces; 167 else 168 LastOutsideTokenChange = &Changes[i - 1]; 169 170 Changes[i].PreviousEndOfTokenColumn = 171 Changes[i - 1].StartOfTokenColumn + Changes[i - 1].TokenLength; 172 173 Changes[i - 1].IsTrailingComment = 174 (Changes[i].NewlinesBefore > 0 || Changes[i].Tok->is(tok::eof) || 175 (Changes[i].IsInsideToken && Changes[i].Tok->is(tok::comment))) && 176 Changes[i - 1].Tok->is(tok::comment) && 177 // FIXME: This is a dirty hack. The problem is that 178 // BreakableLineCommentSection does comment reflow changes and here is 179 // the aligning of trailing comments. Consider the case where we reflow 180 // the second line up in this example: 181 // 182 // // line 1 183 // // line 2 184 // 185 // That amounts to 2 changes by BreakableLineCommentSection: 186 // - the first, delimited by (), for the whitespace between the tokens, 187 // - and second, delimited by [], for the whitespace at the beginning 188 // of the second token: 189 // 190 // // line 1( 191 // )[// ]line 2 192 // 193 // So in the end we have two changes like this: 194 // 195 // // line1()[ ]line 2 196 // 197 // Note that the OriginalWhitespaceStart of the second change is the 198 // same as the PreviousOriginalWhitespaceEnd of the first change. 199 // In this case, the below check ensures that the second change doesn't 200 // get treated as a trailing comment change here, since this might 201 // trigger additional whitespace to be wrongly inserted before "line 2" 202 // by the comment aligner here. 203 // 204 // For a proper solution we need a mechanism to say to WhitespaceManager 205 // that a particular change breaks the current sequence of trailing 206 // comments. 207 OriginalWhitespaceStart != PreviousOriginalWhitespaceEnd; 208 } 209 // FIXME: The last token is currently not always an eof token; in those 210 // cases, setting TokenLength of the last token to 0 is wrong. 211 Changes.back().TokenLength = 0; 212 Changes.back().IsTrailingComment = Changes.back().Tok->is(tok::comment); 213 214 const WhitespaceManager::Change *LastBlockComment = nullptr; 215 for (auto &Change : Changes) { 216 // Reset the IsTrailingComment flag for changes inside of trailing comments 217 // so they don't get realigned later. Comment line breaks however still need 218 // to be aligned. 219 if (Change.IsInsideToken && Change.NewlinesBefore == 0) 220 Change.IsTrailingComment = false; 221 Change.StartOfBlockComment = nullptr; 222 Change.IndentationOffset = 0; 223 if (Change.Tok->is(tok::comment)) { 224 if (Change.Tok->is(TT_LineComment) || !Change.IsInsideToken) 225 LastBlockComment = &Change; 226 else { 227 if ((Change.StartOfBlockComment = LastBlockComment)) 228 Change.IndentationOffset = 229 Change.StartOfTokenColumn - 230 Change.StartOfBlockComment->StartOfTokenColumn; 231 } 232 } else { 233 LastBlockComment = nullptr; 234 } 235 } 236 237 // Compute conditional nesting level 238 // Level is increased for each conditional, unless this conditional continues 239 // a chain of conditional, i.e. starts immediately after the colon of another 240 // conditional. 241 SmallVector<bool, 16> ScopeStack; 242 int ConditionalsLevel = 0; 243 for (auto &Change : Changes) { 244 for (unsigned i = 0, e = Change.Tok->FakeLParens.size(); i != e; ++i) { 245 bool isNestedConditional = 246 Change.Tok->FakeLParens[e - 1 - i] == prec::Conditional && 247 !(i == 0 && Change.Tok->Previous && 248 Change.Tok->Previous->is(TT_ConditionalExpr) && 249 Change.Tok->Previous->is(tok::colon)); 250 if (isNestedConditional) 251 ++ConditionalsLevel; 252 ScopeStack.push_back(isNestedConditional); 253 } 254 255 Change.ConditionalsLevel = ConditionalsLevel; 256 257 for (unsigned i = Change.Tok->FakeRParens; i > 0 && ScopeStack.size(); 258 --i) { 259 if (ScopeStack.pop_back_val()) 260 --ConditionalsLevel; 261 } 262 } 263 } 264 265 // Align a single sequence of tokens, see AlignTokens below. 266 template <typename F> 267 static void 268 AlignTokenSequence(const FormatStyle &Style, unsigned Start, unsigned End, 269 unsigned Column, F &&Matches, 270 SmallVector<WhitespaceManager::Change, 16> &Changes) { 271 bool FoundMatchOnLine = false; 272 int Shift = 0; 273 274 // ScopeStack keeps track of the current scope depth. It contains indices of 275 // the first token on each scope. 276 // We only run the "Matches" function on tokens from the outer-most scope. 277 // However, we do need to pay special attention to one class of tokens 278 // that are not in the outer-most scope, and that is function parameters 279 // which are split across multiple lines, as illustrated by this example: 280 // double a(int x); 281 // int b(int y, 282 // double z); 283 // In the above example, we need to take special care to ensure that 284 // 'double z' is indented along with it's owning function 'b'. 285 // The same holds for calling a function: 286 // double a = foo(x); 287 // int b = bar(foo(y), 288 // foor(z)); 289 // Similar for broken string literals: 290 // double x = 3.14; 291 // auto s = "Hello" 292 // "World"; 293 // Special handling is required for 'nested' ternary operators. 294 SmallVector<unsigned, 16> ScopeStack; 295 296 for (unsigned i = Start; i != End; ++i) { 297 if (ScopeStack.size() != 0 && 298 Changes[i].indentAndNestingLevel() < 299 Changes[ScopeStack.back()].indentAndNestingLevel()) 300 ScopeStack.pop_back(); 301 302 // Compare current token to previous non-comment token to ensure whether 303 // it is in a deeper scope or not. 304 unsigned PreviousNonComment = i - 1; 305 while (PreviousNonComment > Start && 306 Changes[PreviousNonComment].Tok->is(tok::comment)) 307 PreviousNonComment--; 308 if (i != Start && Changes[i].indentAndNestingLevel() > 309 Changes[PreviousNonComment].indentAndNestingLevel()) 310 ScopeStack.push_back(i); 311 312 bool InsideNestedScope = ScopeStack.size() != 0; 313 bool ContinuedStringLiteral = i > Start && 314 Changes[i].Tok->is(tok::string_literal) && 315 Changes[i - 1].Tok->is(tok::string_literal); 316 bool SkipMatchCheck = InsideNestedScope || ContinuedStringLiteral; 317 318 if (Changes[i].NewlinesBefore > 0 && !SkipMatchCheck) { 319 Shift = 0; 320 FoundMatchOnLine = false; 321 } 322 323 // If this is the first matching token to be aligned, remember by how many 324 // spaces it has to be shifted, so the rest of the changes on the line are 325 // shifted by the same amount 326 if (!FoundMatchOnLine && !SkipMatchCheck && Matches(Changes[i])) { 327 FoundMatchOnLine = true; 328 Shift = Column - Changes[i].StartOfTokenColumn; 329 Changes[i].Spaces += Shift; 330 } 331 332 // This is for function parameters that are split across multiple lines, 333 // as mentioned in the ScopeStack comment. 334 if (InsideNestedScope && Changes[i].NewlinesBefore > 0) { 335 unsigned ScopeStart = ScopeStack.back(); 336 auto ShouldShiftBeAdded = [&] { 337 // Function declaration 338 if (Changes[ScopeStart - 1].Tok->is(TT_FunctionDeclarationName)) 339 return true; 340 341 // Continued function declaration 342 if (ScopeStart > Start + 1 && 343 Changes[ScopeStart - 2].Tok->is(TT_FunctionDeclarationName)) 344 return true; 345 346 // Continued function call 347 if (ScopeStart > Start + 1 && 348 Changes[ScopeStart - 2].Tok->is(tok::identifier) && 349 Changes[ScopeStart - 1].Tok->is(tok::l_paren)) 350 return true; 351 352 // Ternary operator 353 if (Changes[i].Tok->is(TT_ConditionalExpr)) 354 return true; 355 356 // Continued ternary operator 357 if (Changes[i].Tok->Previous && 358 Changes[i].Tok->Previous->is(TT_ConditionalExpr)) 359 return true; 360 361 return false; 362 }; 363 364 if (ShouldShiftBeAdded()) 365 Changes[i].Spaces += Shift; 366 } 367 368 if (ContinuedStringLiteral) 369 Changes[i].Spaces += Shift; 370 371 assert(Shift >= 0); 372 373 Changes[i].StartOfTokenColumn += Shift; 374 if (i + 1 != Changes.size()) 375 Changes[i + 1].PreviousEndOfTokenColumn += Shift; 376 377 // If PointerAlignment is PAS_Right, keep *s or &s next to the token 378 if (Style.PointerAlignment == FormatStyle::PAS_Right && 379 Changes[i].Spaces != 0) { 380 for (int Previous = i - 1; 381 Previous >= 0 && 382 Changes[Previous].Tok->getType() == TT_PointerOrReference; 383 --Previous) { 384 Changes[Previous + 1].Spaces -= Shift; 385 Changes[Previous].Spaces += Shift; 386 } 387 } 388 } 389 } 390 391 // Walk through a subset of the changes, starting at StartAt, and find 392 // sequences of matching tokens to align. To do so, keep track of the lines and 393 // whether or not a matching token was found on a line. If a matching token is 394 // found, extend the current sequence. If the current line cannot be part of a 395 // sequence, e.g. because there is an empty line before it or it contains only 396 // non-matching tokens, finalize the previous sequence. 397 // The value returned is the token on which we stopped, either because we 398 // exhausted all items inside Changes, or because we hit a scope level higher 399 // than our initial scope. 400 // This function is recursive. Each invocation processes only the scope level 401 // equal to the initial level, which is the level of Changes[StartAt]. 402 // If we encounter a scope level greater than the initial level, then we call 403 // ourselves recursively, thereby avoiding the pollution of the current state 404 // with the alignment requirements of the nested sub-level. This recursive 405 // behavior is necessary for aligning function prototypes that have one or more 406 // arguments. 407 // If this function encounters a scope level less than the initial level, 408 // it returns the current position. 409 // There is a non-obvious subtlety in the recursive behavior: Even though we 410 // defer processing of nested levels to recursive invocations of this 411 // function, when it comes time to align a sequence of tokens, we run the 412 // alignment on the entire sequence, including the nested levels. 413 // When doing so, most of the nested tokens are skipped, because their 414 // alignment was already handled by the recursive invocations of this function. 415 // However, the special exception is that we do NOT skip function parameters 416 // that are split across multiple lines. See the test case in FormatTest.cpp 417 // that mentions "split function parameter alignment" for an example of this. 418 template <typename F> 419 static unsigned AlignTokens( 420 const FormatStyle &Style, F &&Matches, 421 SmallVector<WhitespaceManager::Change, 16> &Changes, unsigned StartAt, 422 const FormatStyle::AlignConsecutiveStyle &ACS = FormatStyle::ACS_None) { 423 unsigned MinColumn = 0; 424 unsigned MaxColumn = UINT_MAX; 425 426 // Line number of the start and the end of the current token sequence. 427 unsigned StartOfSequence = 0; 428 unsigned EndOfSequence = 0; 429 430 // Measure the scope level (i.e. depth of (), [], {}) of the first token, and 431 // abort when we hit any token in a higher scope than the starting one. 432 auto IndentAndNestingLevel = StartAt < Changes.size() 433 ? Changes[StartAt].indentAndNestingLevel() 434 : std::tuple<unsigned, unsigned, unsigned>(); 435 436 // Keep track of the number of commas before the matching tokens, we will only 437 // align a sequence of matching tokens if they are preceded by the same number 438 // of commas. 439 unsigned CommasBeforeLastMatch = 0; 440 unsigned CommasBeforeMatch = 0; 441 442 // Whether a matching token has been found on the current line. 443 bool FoundMatchOnLine = false; 444 445 // Whether the current line consists purely of comments. 446 bool LineIsComment = true; 447 448 // Aligns a sequence of matching tokens, on the MinColumn column. 449 // 450 // Sequences start from the first matching token to align, and end at the 451 // first token of the first line that doesn't need to be aligned. 452 // 453 // We need to adjust the StartOfTokenColumn of each Change that is on a line 454 // containing any matching token to be aligned and located after such token. 455 auto AlignCurrentSequence = [&] { 456 if (StartOfSequence > 0 && StartOfSequence < EndOfSequence) 457 AlignTokenSequence(Style, StartOfSequence, EndOfSequence, MinColumn, 458 Matches, Changes); 459 MinColumn = 0; 460 MaxColumn = UINT_MAX; 461 StartOfSequence = 0; 462 EndOfSequence = 0; 463 }; 464 465 unsigned i = StartAt; 466 for (unsigned e = Changes.size(); i != e; ++i) { 467 if (Changes[i].indentAndNestingLevel() < IndentAndNestingLevel) 468 break; 469 470 if (Changes[i].NewlinesBefore != 0) { 471 CommasBeforeMatch = 0; 472 EndOfSequence = i; 473 474 // Whether to break the alignment sequence because of an empty line. 475 bool EmptyLineBreak = 476 (Changes[i].NewlinesBefore > 1) && 477 (ACS != FormatStyle::ACS_AcrossEmptyLines) && 478 (ACS != FormatStyle::ACS_AcrossEmptyLinesAndComments); 479 480 // Whether to break the alignment sequence because of a line without a 481 // match. 482 bool NoMatchBreak = 483 !FoundMatchOnLine && 484 !(LineIsComment && 485 ((ACS == FormatStyle::ACS_AcrossComments) || 486 (ACS == FormatStyle::ACS_AcrossEmptyLinesAndComments))); 487 488 if (EmptyLineBreak || NoMatchBreak) 489 AlignCurrentSequence(); 490 491 // A new line starts, re-initialize line status tracking bools. 492 // Keep the match state if a string literal is continued on this line. 493 if (i == 0 || !Changes[i].Tok->is(tok::string_literal) || 494 !Changes[i - 1].Tok->is(tok::string_literal)) 495 FoundMatchOnLine = false; 496 LineIsComment = true; 497 } 498 499 if (!Changes[i].Tok->is(tok::comment)) { 500 LineIsComment = false; 501 } 502 503 if (Changes[i].Tok->is(tok::comma)) { 504 ++CommasBeforeMatch; 505 } else if (Changes[i].indentAndNestingLevel() > IndentAndNestingLevel) { 506 // Call AlignTokens recursively, skipping over this scope block. 507 unsigned StoppedAt = AlignTokens(Style, Matches, Changes, i, ACS); 508 i = StoppedAt - 1; 509 continue; 510 } 511 512 if (!Matches(Changes[i])) 513 continue; 514 515 // If there is more than one matching token per line, or if the number of 516 // preceding commas, do not match anymore, end the sequence. 517 if (FoundMatchOnLine || CommasBeforeMatch != CommasBeforeLastMatch) 518 AlignCurrentSequence(); 519 520 CommasBeforeLastMatch = CommasBeforeMatch; 521 FoundMatchOnLine = true; 522 523 if (StartOfSequence == 0) 524 StartOfSequence = i; 525 526 unsigned ChangeMinColumn = Changes[i].StartOfTokenColumn; 527 int LineLengthAfter = Changes[i].TokenLength; 528 for (unsigned j = i + 1; j != e && Changes[j].NewlinesBefore == 0; ++j) { 529 LineLengthAfter += Changes[j].Spaces; 530 // Changes are generally 1:1 with the tokens, but a change could also be 531 // inside of a token, in which case it's counted more than once: once for 532 // the whitespace surrounding the token (!IsInsideToken) and once for 533 // each whitespace change within it (IsInsideToken). 534 // Therefore, changes inside of a token should only count the space. 535 if (!Changes[j].IsInsideToken) 536 LineLengthAfter += Changes[j].TokenLength; 537 } 538 unsigned ChangeMaxColumn = Style.ColumnLimit - LineLengthAfter; 539 540 // If we are restricted by the maximum column width, end the sequence. 541 if (ChangeMinColumn > MaxColumn || ChangeMaxColumn < MinColumn || 542 CommasBeforeLastMatch != CommasBeforeMatch) { 543 AlignCurrentSequence(); 544 StartOfSequence = i; 545 } 546 547 MinColumn = std::max(MinColumn, ChangeMinColumn); 548 MaxColumn = std::min(MaxColumn, ChangeMaxColumn); 549 } 550 551 EndOfSequence = i; 552 AlignCurrentSequence(); 553 return i; 554 } 555 556 // Aligns a sequence of matching tokens, on the MinColumn column. 557 // 558 // Sequences start from the first matching token to align, and end at the 559 // first token of the first line that doesn't need to be aligned. 560 // 561 // We need to adjust the StartOfTokenColumn of each Change that is on a line 562 // containing any matching token to be aligned and located after such token. 563 static void AlignMacroSequence( 564 unsigned &StartOfSequence, unsigned &EndOfSequence, unsigned &MinColumn, 565 unsigned &MaxColumn, bool &FoundMatchOnLine, 566 std::function<bool(const WhitespaceManager::Change &C)> AlignMacrosMatches, 567 SmallVector<WhitespaceManager::Change, 16> &Changes) { 568 if (StartOfSequence > 0 && StartOfSequence < EndOfSequence) { 569 570 FoundMatchOnLine = false; 571 int Shift = 0; 572 573 for (unsigned I = StartOfSequence; I != EndOfSequence; ++I) { 574 if (Changes[I].NewlinesBefore > 0) { 575 Shift = 0; 576 FoundMatchOnLine = false; 577 } 578 579 // If this is the first matching token to be aligned, remember by how many 580 // spaces it has to be shifted, so the rest of the changes on the line are 581 // shifted by the same amount 582 if (!FoundMatchOnLine && AlignMacrosMatches(Changes[I])) { 583 FoundMatchOnLine = true; 584 Shift = MinColumn - Changes[I].StartOfTokenColumn; 585 Changes[I].Spaces += Shift; 586 } 587 588 assert(Shift >= 0); 589 Changes[I].StartOfTokenColumn += Shift; 590 if (I + 1 != Changes.size()) 591 Changes[I + 1].PreviousEndOfTokenColumn += Shift; 592 } 593 } 594 595 MinColumn = 0; 596 MaxColumn = UINT_MAX; 597 StartOfSequence = 0; 598 EndOfSequence = 0; 599 } 600 601 void WhitespaceManager::alignConsecutiveMacros() { 602 if (Style.AlignConsecutiveMacros == FormatStyle::ACS_None) 603 return; 604 605 auto AlignMacrosMatches = [](const Change &C) { 606 const FormatToken *Current = C.Tok; 607 unsigned SpacesRequiredBefore = 1; 608 609 if (Current->SpacesRequiredBefore == 0 || !Current->Previous) 610 return false; 611 612 Current = Current->Previous; 613 614 // If token is a ")", skip over the parameter list, to the 615 // token that precedes the "(" 616 if (Current->is(tok::r_paren) && Current->MatchingParen) { 617 Current = Current->MatchingParen->Previous; 618 SpacesRequiredBefore = 0; 619 } 620 621 if (!Current || !Current->is(tok::identifier)) 622 return false; 623 624 if (!Current->Previous || !Current->Previous->is(tok::pp_define)) 625 return false; 626 627 // For a macro function, 0 spaces are required between the 628 // identifier and the lparen that opens the parameter list. 629 // For a simple macro, 1 space is required between the 630 // identifier and the first token of the defined value. 631 return Current->Next->SpacesRequiredBefore == SpacesRequiredBefore; 632 }; 633 634 unsigned MinColumn = 0; 635 unsigned MaxColumn = UINT_MAX; 636 637 // Start and end of the token sequence we're processing. 638 unsigned StartOfSequence = 0; 639 unsigned EndOfSequence = 0; 640 641 // Whether a matching token has been found on the current line. 642 bool FoundMatchOnLine = false; 643 644 // Whether the current line consists only of comments 645 bool LineIsComment = true; 646 647 unsigned I = 0; 648 for (unsigned E = Changes.size(); I != E; ++I) { 649 if (Changes[I].NewlinesBefore != 0) { 650 EndOfSequence = I; 651 652 // Whether to break the alignment sequence because of an empty line. 653 bool EmptyLineBreak = 654 (Changes[I].NewlinesBefore > 1) && 655 (Style.AlignConsecutiveMacros != FormatStyle::ACS_AcrossEmptyLines) && 656 (Style.AlignConsecutiveMacros != 657 FormatStyle::ACS_AcrossEmptyLinesAndComments); 658 659 // Whether to break the alignment sequence because of a line without a 660 // match. 661 bool NoMatchBreak = 662 !FoundMatchOnLine && 663 !(LineIsComment && ((Style.AlignConsecutiveMacros == 664 FormatStyle::ACS_AcrossComments) || 665 (Style.AlignConsecutiveMacros == 666 FormatStyle::ACS_AcrossEmptyLinesAndComments))); 667 668 if (EmptyLineBreak || NoMatchBreak) 669 AlignMacroSequence(StartOfSequence, EndOfSequence, MinColumn, MaxColumn, 670 FoundMatchOnLine, AlignMacrosMatches, Changes); 671 672 // A new line starts, re-initialize line status tracking bools. 673 FoundMatchOnLine = false; 674 LineIsComment = true; 675 } 676 677 if (!Changes[I].Tok->is(tok::comment)) { 678 LineIsComment = false; 679 } 680 681 if (!AlignMacrosMatches(Changes[I])) 682 continue; 683 684 FoundMatchOnLine = true; 685 686 if (StartOfSequence == 0) 687 StartOfSequence = I; 688 689 unsigned ChangeMinColumn = Changes[I].StartOfTokenColumn; 690 int LineLengthAfter = -Changes[I].Spaces; 691 for (unsigned j = I; j != E && Changes[j].NewlinesBefore == 0; ++j) 692 LineLengthAfter += Changes[j].Spaces + Changes[j].TokenLength; 693 unsigned ChangeMaxColumn = Style.ColumnLimit - LineLengthAfter; 694 695 MinColumn = std::max(MinColumn, ChangeMinColumn); 696 MaxColumn = std::min(MaxColumn, ChangeMaxColumn); 697 } 698 699 EndOfSequence = I; 700 AlignMacroSequence(StartOfSequence, EndOfSequence, MinColumn, MaxColumn, 701 FoundMatchOnLine, AlignMacrosMatches, Changes); 702 } 703 704 void WhitespaceManager::alignConsecutiveAssignments() { 705 if (Style.AlignConsecutiveAssignments == FormatStyle::ACS_None) 706 return; 707 708 AlignTokens( 709 Style, 710 [&](const Change &C) { 711 // Do not align on equal signs that are first on a line. 712 if (C.NewlinesBefore > 0) 713 return false; 714 715 // Do not align on equal signs that are last on a line. 716 if (&C != &Changes.back() && (&C + 1)->NewlinesBefore > 0) 717 return false; 718 719 return C.Tok->is(tok::equal); 720 }, 721 Changes, /*StartAt=*/0, Style.AlignConsecutiveAssignments); 722 } 723 724 void WhitespaceManager::alignConsecutiveBitFields() { 725 if (Style.AlignConsecutiveBitFields == FormatStyle::ACS_None) 726 return; 727 728 AlignTokens( 729 Style, 730 [&](Change const &C) { 731 // Do not align on ':' that is first on a line. 732 if (C.NewlinesBefore > 0) 733 return false; 734 735 // Do not align on ':' that is last on a line. 736 if (&C != &Changes.back() && (&C + 1)->NewlinesBefore > 0) 737 return false; 738 739 return C.Tok->is(TT_BitFieldColon); 740 }, 741 Changes, /*StartAt=*/0, Style.AlignConsecutiveBitFields); 742 } 743 744 void WhitespaceManager::alignConsecutiveDeclarations() { 745 if (Style.AlignConsecutiveDeclarations == FormatStyle::ACS_None) 746 return; 747 748 AlignTokens( 749 Style, 750 [](Change const &C) { 751 // tok::kw_operator is necessary for aligning operator overload 752 // definitions. 753 if (C.Tok->isOneOf(TT_FunctionDeclarationName, tok::kw_operator)) 754 return true; 755 if (C.Tok->isNot(TT_StartOfName)) 756 return false; 757 if (C.Tok->Previous && 758 C.Tok->Previous->is(TT_StatementAttributeLikeMacro)) 759 return false; 760 // Check if there is a subsequent name that starts the same declaration. 761 for (FormatToken *Next = C.Tok->Next; Next; Next = Next->Next) { 762 if (Next->is(tok::comment)) 763 continue; 764 if (Next->is(TT_PointerOrReference)) 765 return false; 766 if (!Next->Tok.getIdentifierInfo()) 767 break; 768 if (Next->isOneOf(TT_StartOfName, TT_FunctionDeclarationName, 769 tok::kw_operator)) 770 return false; 771 } 772 return true; 773 }, 774 Changes, /*StartAt=*/0, Style.AlignConsecutiveDeclarations); 775 } 776 777 void WhitespaceManager::alignChainedConditionals() { 778 if (Style.BreakBeforeTernaryOperators) { 779 AlignTokens( 780 Style, 781 [](Change const &C) { 782 // Align question operators and last colon 783 return C.Tok->is(TT_ConditionalExpr) && 784 ((C.Tok->is(tok::question) && !C.NewlinesBefore) || 785 (C.Tok->is(tok::colon) && C.Tok->Next && 786 (C.Tok->Next->FakeLParens.size() == 0 || 787 C.Tok->Next->FakeLParens.back() != prec::Conditional))); 788 }, 789 Changes, /*StartAt=*/0); 790 } else { 791 static auto AlignWrappedOperand = [](Change const &C) { 792 FormatToken *Previous = C.Tok->getPreviousNonComment(); 793 return C.NewlinesBefore && Previous && Previous->is(TT_ConditionalExpr) && 794 (Previous->is(tok::colon) && 795 (C.Tok->FakeLParens.size() == 0 || 796 C.Tok->FakeLParens.back() != prec::Conditional)); 797 }; 798 // Ensure we keep alignment of wrapped operands with non-wrapped operands 799 // Since we actually align the operators, the wrapped operands need the 800 // extra offset to be properly aligned. 801 for (Change &C : Changes) { 802 if (AlignWrappedOperand(C)) 803 C.StartOfTokenColumn -= 2; 804 } 805 AlignTokens( 806 Style, 807 [this](Change const &C) { 808 // Align question operators if next operand is not wrapped, as 809 // well as wrapped operands after question operator or last 810 // colon in conditional sequence 811 return (C.Tok->is(TT_ConditionalExpr) && C.Tok->is(tok::question) && 812 &C != &Changes.back() && (&C + 1)->NewlinesBefore == 0 && 813 !(&C + 1)->IsTrailingComment) || 814 AlignWrappedOperand(C); 815 }, 816 Changes, /*StartAt=*/0); 817 } 818 } 819 820 void WhitespaceManager::alignTrailingComments() { 821 unsigned MinColumn = 0; 822 unsigned MaxColumn = UINT_MAX; 823 unsigned StartOfSequence = 0; 824 bool BreakBeforeNext = false; 825 unsigned Newlines = 0; 826 for (unsigned i = 0, e = Changes.size(); i != e; ++i) { 827 if (Changes[i].StartOfBlockComment) 828 continue; 829 Newlines += Changes[i].NewlinesBefore; 830 if (!Changes[i].IsTrailingComment) 831 continue; 832 833 unsigned ChangeMinColumn = Changes[i].StartOfTokenColumn; 834 unsigned ChangeMaxColumn; 835 836 if (Style.ColumnLimit == 0) 837 ChangeMaxColumn = UINT_MAX; 838 else if (Style.ColumnLimit >= Changes[i].TokenLength) 839 ChangeMaxColumn = Style.ColumnLimit - Changes[i].TokenLength; 840 else 841 ChangeMaxColumn = ChangeMinColumn; 842 843 // If we don't create a replacement for this change, we have to consider 844 // it to be immovable. 845 if (!Changes[i].CreateReplacement) 846 ChangeMaxColumn = ChangeMinColumn; 847 848 if (i + 1 != e && Changes[i + 1].ContinuesPPDirective) 849 ChangeMaxColumn -= 2; 850 // If this comment follows an } in column 0, it probably documents the 851 // closing of a namespace and we don't want to align it. 852 bool FollowsRBraceInColumn0 = i > 0 && Changes[i].NewlinesBefore == 0 && 853 Changes[i - 1].Tok->is(tok::r_brace) && 854 Changes[i - 1].StartOfTokenColumn == 0; 855 bool WasAlignedWithStartOfNextLine = false; 856 if (Changes[i].NewlinesBefore == 1) { // A comment on its own line. 857 unsigned CommentColumn = SourceMgr.getSpellingColumnNumber( 858 Changes[i].OriginalWhitespaceRange.getEnd()); 859 for (unsigned j = i + 1; j != e; ++j) { 860 if (Changes[j].Tok->is(tok::comment)) 861 continue; 862 863 unsigned NextColumn = SourceMgr.getSpellingColumnNumber( 864 Changes[j].OriginalWhitespaceRange.getEnd()); 865 // The start of the next token was previously aligned with the 866 // start of this comment. 867 WasAlignedWithStartOfNextLine = 868 CommentColumn == NextColumn || 869 CommentColumn == NextColumn + Style.IndentWidth; 870 break; 871 } 872 } 873 if (!Style.AlignTrailingComments || FollowsRBraceInColumn0) { 874 alignTrailingComments(StartOfSequence, i, MinColumn); 875 MinColumn = ChangeMinColumn; 876 MaxColumn = ChangeMinColumn; 877 StartOfSequence = i; 878 } else if (BreakBeforeNext || Newlines > 1 || 879 (ChangeMinColumn > MaxColumn || ChangeMaxColumn < MinColumn) || 880 // Break the comment sequence if the previous line did not end 881 // in a trailing comment. 882 (Changes[i].NewlinesBefore == 1 && i > 0 && 883 !Changes[i - 1].IsTrailingComment) || 884 WasAlignedWithStartOfNextLine) { 885 alignTrailingComments(StartOfSequence, i, MinColumn); 886 MinColumn = ChangeMinColumn; 887 MaxColumn = ChangeMaxColumn; 888 StartOfSequence = i; 889 } else { 890 MinColumn = std::max(MinColumn, ChangeMinColumn); 891 MaxColumn = std::min(MaxColumn, ChangeMaxColumn); 892 } 893 BreakBeforeNext = (i == 0) || (Changes[i].NewlinesBefore > 1) || 894 // Never start a sequence with a comment at the beginning 895 // of the line. 896 (Changes[i].NewlinesBefore == 1 && StartOfSequence == i); 897 Newlines = 0; 898 } 899 alignTrailingComments(StartOfSequence, Changes.size(), MinColumn); 900 } 901 902 void WhitespaceManager::alignTrailingComments(unsigned Start, unsigned End, 903 unsigned Column) { 904 for (unsigned i = Start; i != End; ++i) { 905 int Shift = 0; 906 if (Changes[i].IsTrailingComment) { 907 Shift = Column - Changes[i].StartOfTokenColumn; 908 } 909 if (Changes[i].StartOfBlockComment) { 910 Shift = Changes[i].IndentationOffset + 911 Changes[i].StartOfBlockComment->StartOfTokenColumn - 912 Changes[i].StartOfTokenColumn; 913 } 914 assert(Shift >= 0); 915 Changes[i].Spaces += Shift; 916 if (i + 1 != Changes.size()) 917 Changes[i + 1].PreviousEndOfTokenColumn += Shift; 918 Changes[i].StartOfTokenColumn += Shift; 919 } 920 } 921 922 void WhitespaceManager::alignEscapedNewlines() { 923 if (Style.AlignEscapedNewlines == FormatStyle::ENAS_DontAlign) 924 return; 925 926 bool AlignLeft = Style.AlignEscapedNewlines == FormatStyle::ENAS_Left; 927 unsigned MaxEndOfLine = AlignLeft ? 0 : Style.ColumnLimit; 928 unsigned StartOfMacro = 0; 929 for (unsigned i = 1, e = Changes.size(); i < e; ++i) { 930 Change &C = Changes[i]; 931 if (C.NewlinesBefore > 0) { 932 if (C.ContinuesPPDirective) { 933 MaxEndOfLine = std::max(C.PreviousEndOfTokenColumn + 2, MaxEndOfLine); 934 } else { 935 alignEscapedNewlines(StartOfMacro + 1, i, MaxEndOfLine); 936 MaxEndOfLine = AlignLeft ? 0 : Style.ColumnLimit; 937 StartOfMacro = i; 938 } 939 } 940 } 941 alignEscapedNewlines(StartOfMacro + 1, Changes.size(), MaxEndOfLine); 942 } 943 944 void WhitespaceManager::alignEscapedNewlines(unsigned Start, unsigned End, 945 unsigned Column) { 946 for (unsigned i = Start; i < End; ++i) { 947 Change &C = Changes[i]; 948 if (C.NewlinesBefore > 0) { 949 assert(C.ContinuesPPDirective); 950 if (C.PreviousEndOfTokenColumn + 1 > Column) 951 C.EscapedNewlineColumn = 0; 952 else 953 C.EscapedNewlineColumn = Column; 954 } 955 } 956 } 957 958 void WhitespaceManager::alignArrayInitializers() { 959 if (Style.AlignArrayOfStructures == FormatStyle::AIAS_None) 960 return; 961 962 for (unsigned ChangeIndex = 1U, ChangeEnd = Changes.size(); 963 ChangeIndex < ChangeEnd; ++ChangeIndex) { 964 auto &C = Changes[ChangeIndex]; 965 if (C.Tok->IsArrayInitializer) { 966 bool FoundComplete = false; 967 for (unsigned InsideIndex = ChangeIndex + 1; InsideIndex < ChangeEnd; 968 ++InsideIndex) { 969 if (Changes[InsideIndex].Tok == C.Tok->MatchingParen) { 970 alignArrayInitializers(ChangeIndex, InsideIndex + 1); 971 ChangeIndex = InsideIndex + 1; 972 FoundComplete = true; 973 break; 974 } 975 } 976 if (!FoundComplete) 977 ChangeIndex = ChangeEnd; 978 } 979 } 980 } 981 982 void WhitespaceManager::alignArrayInitializers(unsigned Start, unsigned End) { 983 984 if (Style.AlignArrayOfStructures == FormatStyle::AIAS_Right) 985 alignArrayInitializersRightJustified(getCells(Start, End)); 986 else if (Style.AlignArrayOfStructures == FormatStyle::AIAS_Left) 987 alignArrayInitializersLeftJustified(getCells(Start, End)); 988 } 989 990 void WhitespaceManager::alignArrayInitializersRightJustified( 991 CellDescriptions &&CellDescs) { 992 auto &Cells = CellDescs.Cells; 993 994 // Now go through and fixup the spaces. 995 auto *CellIter = Cells.begin(); 996 for (auto i = 0U; i < CellDescs.CellCount; i++, ++CellIter) { 997 unsigned NetWidth = 0U; 998 if (isSplitCell(*CellIter)) 999 NetWidth = getNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces); 1000 auto CellWidth = getMaximumCellWidth(CellIter, NetWidth); 1001 1002 if (Changes[CellIter->Index].Tok->is(tok::r_brace)) { 1003 // So in here we want to see if there is a brace that falls 1004 // on a line that was split. If so on that line we make sure that 1005 // the spaces in front of the brace are enough. 1006 Changes[CellIter->Index].NewlinesBefore = 0; 1007 Changes[CellIter->Index].Spaces = 0; 1008 for (const auto *Next = CellIter->NextColumnElement; Next != nullptr; 1009 Next = Next->NextColumnElement) { 1010 Changes[Next->Index].Spaces = 0; 1011 Changes[Next->Index].NewlinesBefore = 0; 1012 } 1013 // Unless the array is empty, we need the position of all the 1014 // immediately adjacent cells 1015 if (CellIter != Cells.begin()) { 1016 auto ThisNetWidth = 1017 getNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces); 1018 auto MaxNetWidth = 1019 getMaximumNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces, 1020 CellDescs.CellCount); 1021 if (ThisNetWidth < MaxNetWidth) 1022 Changes[CellIter->Index].Spaces = (MaxNetWidth - ThisNetWidth); 1023 auto RowCount = 1U; 1024 auto Offset = std::distance(Cells.begin(), CellIter); 1025 for (const auto *Next = CellIter->NextColumnElement; Next != nullptr; 1026 Next = Next->NextColumnElement) { 1027 auto *Start = (Cells.begin() + RowCount * CellDescs.CellCount); 1028 auto *End = Start + Offset; 1029 ThisNetWidth = getNetWidth(Start, End, CellDescs.InitialSpaces); 1030 if (ThisNetWidth < MaxNetWidth) 1031 Changes[Next->Index].Spaces = (MaxNetWidth - ThisNetWidth); 1032 ++RowCount; 1033 } 1034 } 1035 } else { 1036 auto ThisWidth = 1037 calculateCellWidth(CellIter->Index, CellIter->EndIndex, true) + 1038 NetWidth; 1039 if (Changes[CellIter->Index].NewlinesBefore == 0) { 1040 Changes[CellIter->Index].Spaces = (CellWidth - (ThisWidth + NetWidth)); 1041 Changes[CellIter->Index].Spaces += (i > 0) ? 1 : 0; 1042 } 1043 alignToStartOfCell(CellIter->Index, CellIter->EndIndex); 1044 for (const auto *Next = CellIter->NextColumnElement; Next != nullptr; 1045 Next = Next->NextColumnElement) { 1046 ThisWidth = 1047 calculateCellWidth(Next->Index, Next->EndIndex, true) + NetWidth; 1048 if (Changes[Next->Index].NewlinesBefore == 0) { 1049 Changes[Next->Index].Spaces = (CellWidth - ThisWidth); 1050 Changes[Next->Index].Spaces += (i > 0) ? 1 : 0; 1051 } 1052 alignToStartOfCell(Next->Index, Next->EndIndex); 1053 } 1054 } 1055 } 1056 } 1057 1058 void WhitespaceManager::alignArrayInitializersLeftJustified( 1059 CellDescriptions &&CellDescs) { 1060 auto &Cells = CellDescs.Cells; 1061 1062 // Now go through and fixup the spaces. 1063 auto *CellIter = Cells.begin(); 1064 // The first cell needs to be against the left brace. 1065 if (Changes[CellIter->Index].NewlinesBefore == 0) 1066 Changes[CellIter->Index].Spaces = 0; 1067 else 1068 Changes[CellIter->Index].Spaces = CellDescs.InitialSpaces; 1069 ++CellIter; 1070 for (auto i = 1U; i < CellDescs.CellCount; i++, ++CellIter) { 1071 auto MaxNetWidth = getMaximumNetWidth( 1072 Cells.begin(), CellIter, CellDescs.InitialSpaces, CellDescs.CellCount); 1073 auto ThisNetWidth = 1074 getNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces); 1075 if (Changes[CellIter->Index].NewlinesBefore == 0) { 1076 Changes[CellIter->Index].Spaces = 1077 MaxNetWidth - ThisNetWidth + 1078 (Changes[CellIter->Index].Tok->isNot(tok::r_brace) ? 1 : 0); 1079 } 1080 auto RowCount = 1U; 1081 auto Offset = std::distance(Cells.begin(), CellIter); 1082 for (const auto *Next = CellIter->NextColumnElement; Next != nullptr; 1083 Next = Next->NextColumnElement) { 1084 auto *Start = (Cells.begin() + RowCount * CellDescs.CellCount); 1085 auto *End = Start + Offset; 1086 auto ThisNetWidth = getNetWidth(Start, End, CellDescs.InitialSpaces); 1087 if (Changes[Next->Index].NewlinesBefore == 0) { 1088 Changes[Next->Index].Spaces = 1089 MaxNetWidth - ThisNetWidth + 1090 (Changes[Next->Index].Tok->isNot(tok::r_brace) ? 1 : 0); 1091 } 1092 ++RowCount; 1093 } 1094 } 1095 } 1096 1097 bool WhitespaceManager::isSplitCell(const CellDescription &Cell) { 1098 if (Cell.HasSplit) 1099 return true; 1100 for (const auto *Next = Cell.NextColumnElement; Next != nullptr; 1101 Next = Next->NextColumnElement) { 1102 if (Next->HasSplit) 1103 return true; 1104 } 1105 return false; 1106 } 1107 1108 WhitespaceManager::CellDescriptions WhitespaceManager::getCells(unsigned Start, 1109 unsigned End) { 1110 1111 unsigned Depth = 0; 1112 unsigned Cell = 0; 1113 unsigned CellCount = 0; 1114 unsigned InitialSpaces = 0; 1115 unsigned InitialTokenLength = 0; 1116 unsigned EndSpaces = 0; 1117 SmallVector<CellDescription> Cells; 1118 const FormatToken *MatchingParen = nullptr; 1119 for (unsigned i = Start; i < End; ++i) { 1120 auto &C = Changes[i]; 1121 if (C.Tok->is(tok::l_brace)) 1122 ++Depth; 1123 else if (C.Tok->is(tok::r_brace)) 1124 --Depth; 1125 if (Depth == 2) { 1126 if (C.Tok->is(tok::l_brace)) { 1127 Cell = 0; 1128 MatchingParen = C.Tok->MatchingParen; 1129 if (InitialSpaces == 0) { 1130 InitialSpaces = C.Spaces + C.TokenLength; 1131 InitialTokenLength = C.TokenLength; 1132 auto j = i - 1; 1133 for (; Changes[j].NewlinesBefore == 0 && j > Start; --j) { 1134 InitialSpaces += Changes[j].Spaces + Changes[j].TokenLength; 1135 InitialTokenLength += Changes[j].TokenLength; 1136 } 1137 if (C.NewlinesBefore == 0) { 1138 InitialSpaces += Changes[j].Spaces + Changes[j].TokenLength; 1139 InitialTokenLength += Changes[j].TokenLength; 1140 } 1141 } 1142 } else if (C.Tok->is(tok::comma)) { 1143 if (!Cells.empty()) 1144 Cells.back().EndIndex = i; 1145 Cell++; 1146 } 1147 } else if (Depth == 1) { 1148 if (C.Tok == MatchingParen) { 1149 if (!Cells.empty()) 1150 Cells.back().EndIndex = i; 1151 Cells.push_back(CellDescription{i, ++Cell, i + 1, false, nullptr}); 1152 CellCount = Cell + 1; 1153 // Go to the next non-comment and ensure there is a break in front 1154 const auto *NextNonComment = C.Tok->getNextNonComment(); 1155 while (NextNonComment->is(tok::comma)) 1156 NextNonComment = NextNonComment->getNextNonComment(); 1157 auto j = i; 1158 while (Changes[j].Tok != NextNonComment && j < End) 1159 j++; 1160 if (j < End && Changes[j].NewlinesBefore == 0 && 1161 Changes[j].Tok->isNot(tok::r_brace)) { 1162 Changes[j].NewlinesBefore = 1; 1163 // Account for the added token lengths 1164 Changes[j].Spaces = InitialSpaces - InitialTokenLength; 1165 } 1166 } else if (C.Tok->is(tok::comment)) { 1167 // Trailing comments stay at a space past the last token 1168 C.Spaces = Changes[i - 1].Tok->is(tok::comma) ? 1 : 2; 1169 } else if (C.Tok->is(tok::l_brace)) { 1170 // We need to make sure that the ending braces is aligned to the 1171 // start of our initializer 1172 auto j = i - 1; 1173 for (; j > 0 && !Changes[j].Tok->ArrayInitializerLineStart; --j) 1174 ; // Nothing the loop does the work 1175 EndSpaces = Changes[j].Spaces; 1176 } 1177 } else if (Depth == 0 && C.Tok->is(tok::r_brace)) { 1178 C.NewlinesBefore = 1; 1179 C.Spaces = EndSpaces; 1180 } 1181 if (C.Tok->StartsColumn) { 1182 // This gets us past tokens that have been split over multiple 1183 // lines 1184 bool HasSplit = false; 1185 if (Changes[i].NewlinesBefore > 0) { 1186 // So if we split a line previously and the tail line + this token is 1187 // less then the column limit we remove the split here and just put 1188 // the column start at a space past the comma 1189 auto j = i - 1; 1190 if ((j - 1) > Start && Changes[j].Tok->is(tok::comma) && 1191 Changes[j - 1].NewlinesBefore > 0) { 1192 --j; 1193 auto LineLimit = Changes[j].Spaces + Changes[j].TokenLength; 1194 if (LineLimit < Style.ColumnLimit) { 1195 Changes[i].NewlinesBefore = 0; 1196 Changes[i].Spaces = 1; 1197 } 1198 } 1199 } 1200 while (Changes[i].NewlinesBefore > 0 && Changes[i].Tok == C.Tok) { 1201 Changes[i].Spaces = InitialSpaces; 1202 ++i; 1203 HasSplit = true; 1204 } 1205 if (Changes[i].Tok != C.Tok) 1206 --i; 1207 Cells.push_back(CellDescription{i, Cell, i, HasSplit, nullptr}); 1208 } 1209 } 1210 1211 return linkCells({Cells, CellCount, InitialSpaces}); 1212 } 1213 1214 unsigned WhitespaceManager::calculateCellWidth(unsigned Start, unsigned End, 1215 bool WithSpaces) const { 1216 unsigned CellWidth = 0; 1217 for (auto i = Start; i < End; i++) { 1218 if (Changes[i].NewlinesBefore > 0) 1219 CellWidth = 0; 1220 CellWidth += Changes[i].TokenLength; 1221 CellWidth += (WithSpaces ? Changes[i].Spaces : 0); 1222 } 1223 return CellWidth; 1224 } 1225 1226 void WhitespaceManager::alignToStartOfCell(unsigned Start, unsigned End) { 1227 if ((End - Start) <= 1) 1228 return; 1229 // If the line is broken anywhere in there make sure everything 1230 // is aligned to the parent 1231 for (auto i = Start + 1; i < End; i++) { 1232 if (Changes[i].NewlinesBefore > 0) 1233 Changes[i].Spaces = Changes[Start].Spaces; 1234 } 1235 } 1236 1237 WhitespaceManager::CellDescriptions 1238 WhitespaceManager::linkCells(CellDescriptions &&CellDesc) { 1239 auto &Cells = CellDesc.Cells; 1240 for (auto *CellIter = Cells.begin(); CellIter != Cells.end(); ++CellIter) { 1241 if (CellIter->NextColumnElement == nullptr && 1242 ((CellIter + 1) != Cells.end())) { 1243 for (auto *NextIter = CellIter + 1; NextIter != Cells.end(); ++NextIter) { 1244 if (NextIter->Cell == CellIter->Cell) { 1245 CellIter->NextColumnElement = &(*NextIter); 1246 break; 1247 } 1248 } 1249 } 1250 } 1251 return std::move(CellDesc); 1252 } 1253 1254 void WhitespaceManager::generateChanges() { 1255 for (unsigned i = 0, e = Changes.size(); i != e; ++i) { 1256 const Change &C = Changes[i]; 1257 if (i > 0) { 1258 assert(Changes[i - 1].OriginalWhitespaceRange.getBegin() != 1259 C.OriginalWhitespaceRange.getBegin() && 1260 "Generating two replacements for the same location"); 1261 } 1262 if (C.CreateReplacement) { 1263 std::string ReplacementText = C.PreviousLinePostfix; 1264 if (C.ContinuesPPDirective) 1265 appendEscapedNewlineText(ReplacementText, C.NewlinesBefore, 1266 C.PreviousEndOfTokenColumn, 1267 C.EscapedNewlineColumn); 1268 else 1269 appendNewlineText(ReplacementText, C.NewlinesBefore); 1270 appendIndentText( 1271 ReplacementText, C.Tok->IndentLevel, std::max(0, C.Spaces), 1272 C.StartOfTokenColumn - std::max(0, C.Spaces), C.IsAligned); 1273 ReplacementText.append(C.CurrentLinePrefix); 1274 storeReplacement(C.OriginalWhitespaceRange, ReplacementText); 1275 } 1276 } 1277 } 1278 1279 void WhitespaceManager::storeReplacement(SourceRange Range, StringRef Text) { 1280 unsigned WhitespaceLength = SourceMgr.getFileOffset(Range.getEnd()) - 1281 SourceMgr.getFileOffset(Range.getBegin()); 1282 // Don't create a replacement, if it does not change anything. 1283 if (StringRef(SourceMgr.getCharacterData(Range.getBegin()), 1284 WhitespaceLength) == Text) 1285 return; 1286 auto Err = Replaces.add(tooling::Replacement( 1287 SourceMgr, CharSourceRange::getCharRange(Range), Text)); 1288 // FIXME: better error handling. For now, just print an error message in the 1289 // release version. 1290 if (Err) { 1291 llvm::errs() << llvm::toString(std::move(Err)) << "\n"; 1292 assert(false); 1293 } 1294 } 1295 1296 void WhitespaceManager::appendNewlineText(std::string &Text, 1297 unsigned Newlines) { 1298 for (unsigned i = 0; i < Newlines; ++i) 1299 Text.append(UseCRLF ? "\r\n" : "\n"); 1300 } 1301 1302 void WhitespaceManager::appendEscapedNewlineText( 1303 std::string &Text, unsigned Newlines, unsigned PreviousEndOfTokenColumn, 1304 unsigned EscapedNewlineColumn) { 1305 if (Newlines > 0) { 1306 unsigned Spaces = 1307 std::max<int>(1, EscapedNewlineColumn - PreviousEndOfTokenColumn - 1); 1308 for (unsigned i = 0; i < Newlines; ++i) { 1309 Text.append(Spaces, ' '); 1310 Text.append(UseCRLF ? "\\\r\n" : "\\\n"); 1311 Spaces = std::max<int>(0, EscapedNewlineColumn - 1); 1312 } 1313 } 1314 } 1315 1316 void WhitespaceManager::appendIndentText(std::string &Text, 1317 unsigned IndentLevel, unsigned Spaces, 1318 unsigned WhitespaceStartColumn, 1319 bool IsAligned) { 1320 switch (Style.UseTab) { 1321 case FormatStyle::UT_Never: 1322 Text.append(Spaces, ' '); 1323 break; 1324 case FormatStyle::UT_Always: { 1325 if (Style.TabWidth) { 1326 unsigned FirstTabWidth = 1327 Style.TabWidth - WhitespaceStartColumn % Style.TabWidth; 1328 1329 // Insert only spaces when we want to end up before the next tab. 1330 if (Spaces < FirstTabWidth || Spaces == 1) { 1331 Text.append(Spaces, ' '); 1332 break; 1333 } 1334 // Align to the next tab. 1335 Spaces -= FirstTabWidth; 1336 Text.append("\t"); 1337 1338 Text.append(Spaces / Style.TabWidth, '\t'); 1339 Text.append(Spaces % Style.TabWidth, ' '); 1340 } else if (Spaces == 1) { 1341 Text.append(Spaces, ' '); 1342 } 1343 break; 1344 } 1345 case FormatStyle::UT_ForIndentation: 1346 if (WhitespaceStartColumn == 0) { 1347 unsigned Indentation = IndentLevel * Style.IndentWidth; 1348 Spaces = appendTabIndent(Text, Spaces, Indentation); 1349 } 1350 Text.append(Spaces, ' '); 1351 break; 1352 case FormatStyle::UT_ForContinuationAndIndentation: 1353 if (WhitespaceStartColumn == 0) 1354 Spaces = appendTabIndent(Text, Spaces, Spaces); 1355 Text.append(Spaces, ' '); 1356 break; 1357 case FormatStyle::UT_AlignWithSpaces: 1358 if (WhitespaceStartColumn == 0) { 1359 unsigned Indentation = 1360 IsAligned ? IndentLevel * Style.IndentWidth : Spaces; 1361 Spaces = appendTabIndent(Text, Spaces, Indentation); 1362 } 1363 Text.append(Spaces, ' '); 1364 break; 1365 } 1366 } 1367 1368 unsigned WhitespaceManager::appendTabIndent(std::string &Text, unsigned Spaces, 1369 unsigned Indentation) { 1370 // This happens, e.g. when a line in a block comment is indented less than the 1371 // first one. 1372 if (Indentation > Spaces) 1373 Indentation = Spaces; 1374 if (Style.TabWidth) { 1375 unsigned Tabs = Indentation / Style.TabWidth; 1376 Text.append(Tabs, '\t'); 1377 Spaces -= Tabs * Style.TabWidth; 1378 } 1379 return Spaces; 1380 } 1381 1382 } // namespace format 1383 } // namespace clang 1384