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 Style.BinPackArguments; 351 352 // Ternary operator 353 if (Changes[i].Tok->is(TT_ConditionalExpr)) 354 return true; 355 356 // Period Initializer .XXX = 1. 357 if (Changes[i].Tok->is(TT_DesignatedInitializerPeriod)) 358 return true; 359 360 // Continued ternary operator 361 if (Changes[i].Tok->Previous && 362 Changes[i].Tok->Previous->is(TT_ConditionalExpr)) 363 return true; 364 365 return false; 366 }; 367 368 if (ShouldShiftBeAdded()) 369 Changes[i].Spaces += Shift; 370 } 371 372 if (ContinuedStringLiteral) 373 Changes[i].Spaces += Shift; 374 375 Changes[i].StartOfTokenColumn += Shift; 376 if (i + 1 != Changes.size()) 377 Changes[i + 1].PreviousEndOfTokenColumn += Shift; 378 379 // If PointerAlignment is PAS_Right, keep *s or &s next to the token 380 if (Style.PointerAlignment == FormatStyle::PAS_Right && 381 Changes[i].Spaces != 0) { 382 for (int Previous = i - 1; 383 Previous >= 0 && 384 Changes[Previous].Tok->getType() == TT_PointerOrReference; 385 --Previous) { 386 Changes[Previous + 1].Spaces -= Shift; 387 Changes[Previous].Spaces += Shift; 388 } 389 } 390 } 391 } 392 393 // Walk through a subset of the changes, starting at StartAt, and find 394 // sequences of matching tokens to align. To do so, keep track of the lines and 395 // whether or not a matching token was found on a line. If a matching token is 396 // found, extend the current sequence. If the current line cannot be part of a 397 // sequence, e.g. because there is an empty line before it or it contains only 398 // non-matching tokens, finalize the previous sequence. 399 // The value returned is the token on which we stopped, either because we 400 // exhausted all items inside Changes, or because we hit a scope level higher 401 // than our initial scope. 402 // This function is recursive. Each invocation processes only the scope level 403 // equal to the initial level, which is the level of Changes[StartAt]. 404 // If we encounter a scope level greater than the initial level, then we call 405 // ourselves recursively, thereby avoiding the pollution of the current state 406 // with the alignment requirements of the nested sub-level. This recursive 407 // behavior is necessary for aligning function prototypes that have one or more 408 // arguments. 409 // If this function encounters a scope level less than the initial level, 410 // it returns the current position. 411 // There is a non-obvious subtlety in the recursive behavior: Even though we 412 // defer processing of nested levels to recursive invocations of this 413 // function, when it comes time to align a sequence of tokens, we run the 414 // alignment on the entire sequence, including the nested levels. 415 // When doing so, most of the nested tokens are skipped, because their 416 // alignment was already handled by the recursive invocations of this function. 417 // However, the special exception is that we do NOT skip function parameters 418 // that are split across multiple lines. See the test case in FormatTest.cpp 419 // that mentions "split function parameter alignment" for an example of this. 420 template <typename F> 421 static unsigned AlignTokens( 422 const FormatStyle &Style, F &&Matches, 423 SmallVector<WhitespaceManager::Change, 16> &Changes, unsigned StartAt, 424 const FormatStyle::AlignConsecutiveStyle &ACS = FormatStyle::ACS_None) { 425 unsigned MinColumn = 0; 426 unsigned MaxColumn = UINT_MAX; 427 428 // Line number of the start and the end of the current token sequence. 429 unsigned StartOfSequence = 0; 430 unsigned EndOfSequence = 0; 431 432 // Measure the scope level (i.e. depth of (), [], {}) of the first token, and 433 // abort when we hit any token in a higher scope than the starting one. 434 auto IndentAndNestingLevel = StartAt < Changes.size() 435 ? Changes[StartAt].indentAndNestingLevel() 436 : std::tuple<unsigned, unsigned, unsigned>(); 437 438 // Keep track of the number of commas before the matching tokens, we will only 439 // align a sequence of matching tokens if they are preceded by the same number 440 // of commas. 441 unsigned CommasBeforeLastMatch = 0; 442 unsigned CommasBeforeMatch = 0; 443 444 // Whether a matching token has been found on the current line. 445 bool FoundMatchOnLine = false; 446 447 // Whether the current line consists purely of comments. 448 bool LineIsComment = true; 449 450 // Aligns a sequence of matching tokens, on the MinColumn column. 451 // 452 // Sequences start from the first matching token to align, and end at the 453 // first token of the first line that doesn't need to be aligned. 454 // 455 // We need to adjust the StartOfTokenColumn of each Change that is on a line 456 // containing any matching token to be aligned and located after such token. 457 auto AlignCurrentSequence = [&] { 458 if (StartOfSequence > 0 && StartOfSequence < EndOfSequence) 459 AlignTokenSequence(Style, StartOfSequence, EndOfSequence, MinColumn, 460 Matches, Changes); 461 MinColumn = 0; 462 MaxColumn = UINT_MAX; 463 StartOfSequence = 0; 464 EndOfSequence = 0; 465 }; 466 467 unsigned i = StartAt; 468 for (unsigned e = Changes.size(); i != e; ++i) { 469 if (Changes[i].indentAndNestingLevel() < IndentAndNestingLevel) 470 break; 471 472 if (Changes[i].NewlinesBefore != 0) { 473 CommasBeforeMatch = 0; 474 EndOfSequence = i; 475 476 // Whether to break the alignment sequence because of an empty line. 477 bool EmptyLineBreak = 478 (Changes[i].NewlinesBefore > 1) && 479 (ACS != FormatStyle::ACS_AcrossEmptyLines) && 480 (ACS != FormatStyle::ACS_AcrossEmptyLinesAndComments); 481 482 // Whether to break the alignment sequence because of a line without a 483 // match. 484 bool NoMatchBreak = 485 !FoundMatchOnLine && 486 !(LineIsComment && 487 ((ACS == FormatStyle::ACS_AcrossComments) || 488 (ACS == FormatStyle::ACS_AcrossEmptyLinesAndComments))); 489 490 if (EmptyLineBreak || NoMatchBreak) 491 AlignCurrentSequence(); 492 493 // A new line starts, re-initialize line status tracking bools. 494 // Keep the match state if a string literal is continued on this line. 495 if (i == 0 || !Changes[i].Tok->is(tok::string_literal) || 496 !Changes[i - 1].Tok->is(tok::string_literal)) 497 FoundMatchOnLine = false; 498 LineIsComment = true; 499 } 500 501 if (!Changes[i].Tok->is(tok::comment)) { 502 LineIsComment = false; 503 } 504 505 if (Changes[i].Tok->is(tok::comma)) { 506 ++CommasBeforeMatch; 507 } else if (Changes[i].indentAndNestingLevel() > IndentAndNestingLevel) { 508 // Call AlignTokens recursively, skipping over this scope block. 509 unsigned StoppedAt = AlignTokens(Style, Matches, Changes, i, ACS); 510 i = StoppedAt - 1; 511 continue; 512 } 513 514 if (!Matches(Changes[i])) 515 continue; 516 517 // If there is more than one matching token per line, or if the number of 518 // preceding commas, do not match anymore, end the sequence. 519 if (FoundMatchOnLine || CommasBeforeMatch != CommasBeforeLastMatch) 520 AlignCurrentSequence(); 521 522 CommasBeforeLastMatch = CommasBeforeMatch; 523 FoundMatchOnLine = true; 524 525 if (StartOfSequence == 0) 526 StartOfSequence = i; 527 528 unsigned ChangeMinColumn = Changes[i].StartOfTokenColumn; 529 int LineLengthAfter = Changes[i].TokenLength; 530 for (unsigned j = i + 1; j != e && Changes[j].NewlinesBefore == 0; ++j) { 531 LineLengthAfter += Changes[j].Spaces; 532 // Changes are generally 1:1 with the tokens, but a change could also be 533 // inside of a token, in which case it's counted more than once: once for 534 // the whitespace surrounding the token (!IsInsideToken) and once for 535 // each whitespace change within it (IsInsideToken). 536 // Therefore, changes inside of a token should only count the space. 537 if (!Changes[j].IsInsideToken) 538 LineLengthAfter += Changes[j].TokenLength; 539 } 540 unsigned ChangeMaxColumn = Style.ColumnLimit - LineLengthAfter; 541 542 // If we are restricted by the maximum column width, end the sequence. 543 if (ChangeMinColumn > MaxColumn || ChangeMaxColumn < MinColumn || 544 CommasBeforeLastMatch != CommasBeforeMatch) { 545 AlignCurrentSequence(); 546 StartOfSequence = i; 547 } 548 549 MinColumn = std::max(MinColumn, ChangeMinColumn); 550 MaxColumn = std::min(MaxColumn, ChangeMaxColumn); 551 } 552 553 EndOfSequence = i; 554 AlignCurrentSequence(); 555 return i; 556 } 557 558 // Aligns a sequence of matching tokens, on the MinColumn column. 559 // 560 // Sequences start from the first matching token to align, and end at the 561 // first token of the first line that doesn't need to be aligned. 562 // 563 // We need to adjust the StartOfTokenColumn of each Change that is on a line 564 // containing any matching token to be aligned and located after such token. 565 static void AlignMacroSequence( 566 unsigned &StartOfSequence, unsigned &EndOfSequence, unsigned &MinColumn, 567 unsigned &MaxColumn, bool &FoundMatchOnLine, 568 std::function<bool(const WhitespaceManager::Change &C)> AlignMacrosMatches, 569 SmallVector<WhitespaceManager::Change, 16> &Changes) { 570 if (StartOfSequence > 0 && StartOfSequence < EndOfSequence) { 571 572 FoundMatchOnLine = false; 573 int Shift = 0; 574 575 for (unsigned I = StartOfSequence; I != EndOfSequence; ++I) { 576 if (Changes[I].NewlinesBefore > 0) { 577 Shift = 0; 578 FoundMatchOnLine = false; 579 } 580 581 // If this is the first matching token to be aligned, remember by how many 582 // spaces it has to be shifted, so the rest of the changes on the line are 583 // shifted by the same amount 584 if (!FoundMatchOnLine && AlignMacrosMatches(Changes[I])) { 585 FoundMatchOnLine = true; 586 Shift = MinColumn - Changes[I].StartOfTokenColumn; 587 Changes[I].Spaces += Shift; 588 } 589 590 assert(Shift >= 0); 591 Changes[I].StartOfTokenColumn += Shift; 592 if (I + 1 != Changes.size()) 593 Changes[I + 1].PreviousEndOfTokenColumn += Shift; 594 } 595 } 596 597 MinColumn = 0; 598 MaxColumn = UINT_MAX; 599 StartOfSequence = 0; 600 EndOfSequence = 0; 601 } 602 603 void WhitespaceManager::alignConsecutiveMacros() { 604 if (Style.AlignConsecutiveMacros == FormatStyle::ACS_None) 605 return; 606 607 auto AlignMacrosMatches = [](const Change &C) { 608 const FormatToken *Current = C.Tok; 609 unsigned SpacesRequiredBefore = 1; 610 611 if (Current->SpacesRequiredBefore == 0 || !Current->Previous) 612 return false; 613 614 Current = Current->Previous; 615 616 // If token is a ")", skip over the parameter list, to the 617 // token that precedes the "(" 618 if (Current->is(tok::r_paren) && Current->MatchingParen) { 619 Current = Current->MatchingParen->Previous; 620 SpacesRequiredBefore = 0; 621 } 622 623 if (!Current || !Current->is(tok::identifier)) 624 return false; 625 626 if (!Current->Previous || !Current->Previous->is(tok::pp_define)) 627 return false; 628 629 // For a macro function, 0 spaces are required between the 630 // identifier and the lparen that opens the parameter list. 631 // For a simple macro, 1 space is required between the 632 // identifier and the first token of the defined value. 633 return Current->Next->SpacesRequiredBefore == SpacesRequiredBefore; 634 }; 635 636 unsigned MinColumn = 0; 637 unsigned MaxColumn = UINT_MAX; 638 639 // Start and end of the token sequence we're processing. 640 unsigned StartOfSequence = 0; 641 unsigned EndOfSequence = 0; 642 643 // Whether a matching token has been found on the current line. 644 bool FoundMatchOnLine = false; 645 646 // Whether the current line consists only of comments 647 bool LineIsComment = true; 648 649 unsigned I = 0; 650 for (unsigned E = Changes.size(); I != E; ++I) { 651 if (Changes[I].NewlinesBefore != 0) { 652 EndOfSequence = I; 653 654 // Whether to break the alignment sequence because of an empty line. 655 bool EmptyLineBreak = 656 (Changes[I].NewlinesBefore > 1) && 657 (Style.AlignConsecutiveMacros != FormatStyle::ACS_AcrossEmptyLines) && 658 (Style.AlignConsecutiveMacros != 659 FormatStyle::ACS_AcrossEmptyLinesAndComments); 660 661 // Whether to break the alignment sequence because of a line without a 662 // match. 663 bool NoMatchBreak = 664 !FoundMatchOnLine && 665 !(LineIsComment && ((Style.AlignConsecutiveMacros == 666 FormatStyle::ACS_AcrossComments) || 667 (Style.AlignConsecutiveMacros == 668 FormatStyle::ACS_AcrossEmptyLinesAndComments))); 669 670 if (EmptyLineBreak || NoMatchBreak) 671 AlignMacroSequence(StartOfSequence, EndOfSequence, MinColumn, MaxColumn, 672 FoundMatchOnLine, AlignMacrosMatches, Changes); 673 674 // A new line starts, re-initialize line status tracking bools. 675 FoundMatchOnLine = false; 676 LineIsComment = true; 677 } 678 679 if (!Changes[I].Tok->is(tok::comment)) { 680 LineIsComment = false; 681 } 682 683 if (!AlignMacrosMatches(Changes[I])) 684 continue; 685 686 FoundMatchOnLine = true; 687 688 if (StartOfSequence == 0) 689 StartOfSequence = I; 690 691 unsigned ChangeMinColumn = Changes[I].StartOfTokenColumn; 692 int LineLengthAfter = -Changes[I].Spaces; 693 for (unsigned j = I; j != E && Changes[j].NewlinesBefore == 0; ++j) 694 LineLengthAfter += Changes[j].Spaces + Changes[j].TokenLength; 695 unsigned ChangeMaxColumn = Style.ColumnLimit - LineLengthAfter; 696 697 MinColumn = std::max(MinColumn, ChangeMinColumn); 698 MaxColumn = std::min(MaxColumn, ChangeMaxColumn); 699 } 700 701 EndOfSequence = I; 702 AlignMacroSequence(StartOfSequence, EndOfSequence, MinColumn, MaxColumn, 703 FoundMatchOnLine, AlignMacrosMatches, Changes); 704 } 705 706 void WhitespaceManager::alignConsecutiveAssignments() { 707 if (Style.AlignConsecutiveAssignments == FormatStyle::ACS_None) 708 return; 709 710 AlignTokens( 711 Style, 712 [&](const Change &C) { 713 // Do not align on equal signs that are first on a line. 714 if (C.NewlinesBefore > 0) 715 return false; 716 717 // Do not align on equal signs that are last on a line. 718 if (&C != &Changes.back() && (&C + 1)->NewlinesBefore > 0) 719 return false; 720 721 return C.Tok->is(tok::equal); 722 }, 723 Changes, /*StartAt=*/0, Style.AlignConsecutiveAssignments); 724 } 725 726 void WhitespaceManager::alignConsecutiveBitFields() { 727 if (Style.AlignConsecutiveBitFields == FormatStyle::ACS_None) 728 return; 729 730 AlignTokens( 731 Style, 732 [&](Change const &C) { 733 // Do not align on ':' that is first on a line. 734 if (C.NewlinesBefore > 0) 735 return false; 736 737 // Do not align on ':' that is last on a line. 738 if (&C != &Changes.back() && (&C + 1)->NewlinesBefore > 0) 739 return false; 740 741 return C.Tok->is(TT_BitFieldColon); 742 }, 743 Changes, /*StartAt=*/0, Style.AlignConsecutiveBitFields); 744 } 745 746 void WhitespaceManager::alignConsecutiveDeclarations() { 747 if (Style.AlignConsecutiveDeclarations == FormatStyle::ACS_None) 748 return; 749 750 AlignTokens( 751 Style, 752 [](Change const &C) { 753 // tok::kw_operator is necessary for aligning operator overload 754 // definitions. 755 if (C.Tok->isOneOf(TT_FunctionDeclarationName, tok::kw_operator)) 756 return true; 757 if (C.Tok->isNot(TT_StartOfName)) 758 return false; 759 if (C.Tok->Previous && 760 C.Tok->Previous->is(TT_StatementAttributeLikeMacro)) 761 return false; 762 // Check if there is a subsequent name that starts the same declaration. 763 for (FormatToken *Next = C.Tok->Next; Next; Next = Next->Next) { 764 if (Next->is(tok::comment)) 765 continue; 766 if (Next->is(TT_PointerOrReference)) 767 return false; 768 if (!Next->Tok.getIdentifierInfo()) 769 break; 770 if (Next->isOneOf(TT_StartOfName, TT_FunctionDeclarationName, 771 tok::kw_operator)) 772 return false; 773 } 774 return true; 775 }, 776 Changes, /*StartAt=*/0, Style.AlignConsecutiveDeclarations); 777 } 778 779 void WhitespaceManager::alignChainedConditionals() { 780 if (Style.BreakBeforeTernaryOperators) { 781 AlignTokens( 782 Style, 783 [](Change const &C) { 784 // Align question operators and last colon 785 return C.Tok->is(TT_ConditionalExpr) && 786 ((C.Tok->is(tok::question) && !C.NewlinesBefore) || 787 (C.Tok->is(tok::colon) && C.Tok->Next && 788 (C.Tok->Next->FakeLParens.size() == 0 || 789 C.Tok->Next->FakeLParens.back() != prec::Conditional))); 790 }, 791 Changes, /*StartAt=*/0); 792 } else { 793 static auto AlignWrappedOperand = [](Change const &C) { 794 FormatToken *Previous = C.Tok->getPreviousNonComment(); 795 return C.NewlinesBefore && Previous && Previous->is(TT_ConditionalExpr) && 796 (Previous->is(tok::colon) && 797 (C.Tok->FakeLParens.size() == 0 || 798 C.Tok->FakeLParens.back() != prec::Conditional)); 799 }; 800 // Ensure we keep alignment of wrapped operands with non-wrapped operands 801 // Since we actually align the operators, the wrapped operands need the 802 // extra offset to be properly aligned. 803 for (Change &C : Changes) { 804 if (AlignWrappedOperand(C)) 805 C.StartOfTokenColumn -= 2; 806 } 807 AlignTokens( 808 Style, 809 [this](Change const &C) { 810 // Align question operators if next operand is not wrapped, as 811 // well as wrapped operands after question operator or last 812 // colon in conditional sequence 813 return (C.Tok->is(TT_ConditionalExpr) && C.Tok->is(tok::question) && 814 &C != &Changes.back() && (&C + 1)->NewlinesBefore == 0 && 815 !(&C + 1)->IsTrailingComment) || 816 AlignWrappedOperand(C); 817 }, 818 Changes, /*StartAt=*/0); 819 } 820 } 821 822 void WhitespaceManager::alignTrailingComments() { 823 unsigned MinColumn = 0; 824 unsigned MaxColumn = UINT_MAX; 825 unsigned StartOfSequence = 0; 826 bool BreakBeforeNext = false; 827 unsigned Newlines = 0; 828 for (unsigned i = 0, e = Changes.size(); i != e; ++i) { 829 if (Changes[i].StartOfBlockComment) 830 continue; 831 Newlines += Changes[i].NewlinesBefore; 832 if (!Changes[i].IsTrailingComment) 833 continue; 834 835 unsigned ChangeMinColumn = Changes[i].StartOfTokenColumn; 836 unsigned ChangeMaxColumn; 837 838 if (Style.ColumnLimit == 0) 839 ChangeMaxColumn = UINT_MAX; 840 else if (Style.ColumnLimit >= Changes[i].TokenLength) 841 ChangeMaxColumn = Style.ColumnLimit - Changes[i].TokenLength; 842 else 843 ChangeMaxColumn = ChangeMinColumn; 844 845 // If we don't create a replacement for this change, we have to consider 846 // it to be immovable. 847 if (!Changes[i].CreateReplacement) 848 ChangeMaxColumn = ChangeMinColumn; 849 850 if (i + 1 != e && Changes[i + 1].ContinuesPPDirective) 851 ChangeMaxColumn -= 2; 852 // If this comment follows an } in column 0, it probably documents the 853 // closing of a namespace and we don't want to align it. 854 bool FollowsRBraceInColumn0 = i > 0 && Changes[i].NewlinesBefore == 0 && 855 Changes[i - 1].Tok->is(tok::r_brace) && 856 Changes[i - 1].StartOfTokenColumn == 0; 857 bool WasAlignedWithStartOfNextLine = false; 858 if (Changes[i].NewlinesBefore == 1) { // A comment on its own line. 859 unsigned CommentColumn = SourceMgr.getSpellingColumnNumber( 860 Changes[i].OriginalWhitespaceRange.getEnd()); 861 for (unsigned j = i + 1; j != e; ++j) { 862 if (Changes[j].Tok->is(tok::comment)) 863 continue; 864 865 unsigned NextColumn = SourceMgr.getSpellingColumnNumber( 866 Changes[j].OriginalWhitespaceRange.getEnd()); 867 // The start of the next token was previously aligned with the 868 // start of this comment. 869 WasAlignedWithStartOfNextLine = 870 CommentColumn == NextColumn || 871 CommentColumn == NextColumn + Style.IndentWidth; 872 break; 873 } 874 } 875 if (!Style.AlignTrailingComments || FollowsRBraceInColumn0) { 876 alignTrailingComments(StartOfSequence, i, MinColumn); 877 MinColumn = ChangeMinColumn; 878 MaxColumn = ChangeMinColumn; 879 StartOfSequence = i; 880 } else if (BreakBeforeNext || Newlines > 1 || 881 (ChangeMinColumn > MaxColumn || ChangeMaxColumn < MinColumn) || 882 // Break the comment sequence if the previous line did not end 883 // in a trailing comment. 884 (Changes[i].NewlinesBefore == 1 && i > 0 && 885 !Changes[i - 1].IsTrailingComment) || 886 WasAlignedWithStartOfNextLine) { 887 alignTrailingComments(StartOfSequence, i, MinColumn); 888 MinColumn = ChangeMinColumn; 889 MaxColumn = ChangeMaxColumn; 890 StartOfSequence = i; 891 } else { 892 MinColumn = std::max(MinColumn, ChangeMinColumn); 893 MaxColumn = std::min(MaxColumn, ChangeMaxColumn); 894 } 895 BreakBeforeNext = (i == 0) || (Changes[i].NewlinesBefore > 1) || 896 // Never start a sequence with a comment at the beginning 897 // of the line. 898 (Changes[i].NewlinesBefore == 1 && StartOfSequence == i); 899 Newlines = 0; 900 } 901 alignTrailingComments(StartOfSequence, Changes.size(), MinColumn); 902 } 903 904 void WhitespaceManager::alignTrailingComments(unsigned Start, unsigned End, 905 unsigned Column) { 906 for (unsigned i = Start; i != End; ++i) { 907 int Shift = 0; 908 if (Changes[i].IsTrailingComment) { 909 Shift = Column - Changes[i].StartOfTokenColumn; 910 } 911 if (Changes[i].StartOfBlockComment) { 912 Shift = Changes[i].IndentationOffset + 913 Changes[i].StartOfBlockComment->StartOfTokenColumn - 914 Changes[i].StartOfTokenColumn; 915 } 916 if (Shift < 0) continue; 917 Changes[i].Spaces += Shift; 918 if (i + 1 != Changes.size()) 919 Changes[i + 1].PreviousEndOfTokenColumn += Shift; 920 Changes[i].StartOfTokenColumn += Shift; 921 } 922 } 923 924 void WhitespaceManager::alignEscapedNewlines() { 925 if (Style.AlignEscapedNewlines == FormatStyle::ENAS_DontAlign) 926 return; 927 928 bool AlignLeft = Style.AlignEscapedNewlines == FormatStyle::ENAS_Left; 929 unsigned MaxEndOfLine = AlignLeft ? 0 : Style.ColumnLimit; 930 unsigned StartOfMacro = 0; 931 for (unsigned i = 1, e = Changes.size(); i < e; ++i) { 932 Change &C = Changes[i]; 933 if (C.NewlinesBefore > 0) { 934 if (C.ContinuesPPDirective) { 935 MaxEndOfLine = std::max(C.PreviousEndOfTokenColumn + 2, MaxEndOfLine); 936 } else { 937 alignEscapedNewlines(StartOfMacro + 1, i, MaxEndOfLine); 938 MaxEndOfLine = AlignLeft ? 0 : Style.ColumnLimit; 939 StartOfMacro = i; 940 } 941 } 942 } 943 alignEscapedNewlines(StartOfMacro + 1, Changes.size(), MaxEndOfLine); 944 } 945 946 void WhitespaceManager::alignEscapedNewlines(unsigned Start, unsigned End, 947 unsigned Column) { 948 for (unsigned i = Start; i < End; ++i) { 949 Change &C = Changes[i]; 950 if (C.NewlinesBefore > 0) { 951 assert(C.ContinuesPPDirective); 952 if (C.PreviousEndOfTokenColumn + 1 > Column) 953 C.EscapedNewlineColumn = 0; 954 else 955 C.EscapedNewlineColumn = Column; 956 } 957 } 958 } 959 960 void WhitespaceManager::alignArrayInitializers() { 961 if (Style.AlignArrayOfStructures == FormatStyle::AIAS_None) 962 return; 963 964 for (unsigned ChangeIndex = 1U, ChangeEnd = Changes.size(); 965 ChangeIndex < ChangeEnd; ++ChangeIndex) { 966 auto &C = Changes[ChangeIndex]; 967 if (C.Tok->IsArrayInitializer) { 968 bool FoundComplete = false; 969 for (unsigned InsideIndex = ChangeIndex + 1; InsideIndex < ChangeEnd; 970 ++InsideIndex) { 971 if (Changes[InsideIndex].Tok == C.Tok->MatchingParen) { 972 alignArrayInitializers(ChangeIndex, InsideIndex + 1); 973 ChangeIndex = InsideIndex + 1; 974 FoundComplete = true; 975 break; 976 } 977 } 978 if (!FoundComplete) 979 ChangeIndex = ChangeEnd; 980 } 981 } 982 } 983 984 void WhitespaceManager::alignArrayInitializers(unsigned Start, unsigned End) { 985 986 if (Style.AlignArrayOfStructures == FormatStyle::AIAS_Right) 987 alignArrayInitializersRightJustified(getCells(Start, End)); 988 else if (Style.AlignArrayOfStructures == FormatStyle::AIAS_Left) 989 alignArrayInitializersLeftJustified(getCells(Start, End)); 990 } 991 992 void WhitespaceManager::alignArrayInitializersRightJustified( 993 CellDescriptions &&CellDescs) { 994 auto &Cells = CellDescs.Cells; 995 996 // Now go through and fixup the spaces. 997 auto *CellIter = Cells.begin(); 998 for (auto i = 0U; i < CellDescs.CellCount; i++, ++CellIter) { 999 unsigned NetWidth = 0U; 1000 if (isSplitCell(*CellIter)) 1001 NetWidth = getNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces); 1002 auto CellWidth = getMaximumCellWidth(CellIter, NetWidth); 1003 1004 if (Changes[CellIter->Index].Tok->is(tok::r_brace)) { 1005 // So in here we want to see if there is a brace that falls 1006 // on a line that was split. If so on that line we make sure that 1007 // the spaces in front of the brace are enough. 1008 Changes[CellIter->Index].NewlinesBefore = 0; 1009 Changes[CellIter->Index].Spaces = 0; 1010 for (const auto *Next = CellIter->NextColumnElement; Next != nullptr; 1011 Next = Next->NextColumnElement) { 1012 Changes[Next->Index].Spaces = 0; 1013 Changes[Next->Index].NewlinesBefore = 0; 1014 } 1015 // Unless the array is empty, we need the position of all the 1016 // immediately adjacent cells 1017 if (CellIter != Cells.begin()) { 1018 auto ThisNetWidth = 1019 getNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces); 1020 auto MaxNetWidth = 1021 getMaximumNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces, 1022 CellDescs.CellCount); 1023 if (ThisNetWidth < MaxNetWidth) 1024 Changes[CellIter->Index].Spaces = (MaxNetWidth - ThisNetWidth); 1025 auto RowCount = 1U; 1026 auto Offset = std::distance(Cells.begin(), CellIter); 1027 for (const auto *Next = CellIter->NextColumnElement; Next != nullptr; 1028 Next = Next->NextColumnElement) { 1029 auto *Start = (Cells.begin() + RowCount * CellDescs.CellCount); 1030 auto *End = Start + Offset; 1031 ThisNetWidth = getNetWidth(Start, End, CellDescs.InitialSpaces); 1032 if (ThisNetWidth < MaxNetWidth) 1033 Changes[Next->Index].Spaces = (MaxNetWidth - ThisNetWidth); 1034 ++RowCount; 1035 } 1036 } 1037 } else { 1038 auto ThisWidth = 1039 calculateCellWidth(CellIter->Index, CellIter->EndIndex, true) + 1040 NetWidth; 1041 if (Changes[CellIter->Index].NewlinesBefore == 0) { 1042 Changes[CellIter->Index].Spaces = (CellWidth - (ThisWidth + NetWidth)); 1043 Changes[CellIter->Index].Spaces += (i > 0) ? 1 : 0; 1044 } 1045 alignToStartOfCell(CellIter->Index, CellIter->EndIndex); 1046 for (const auto *Next = CellIter->NextColumnElement; Next != nullptr; 1047 Next = Next->NextColumnElement) { 1048 ThisWidth = 1049 calculateCellWidth(Next->Index, Next->EndIndex, true) + NetWidth; 1050 if (Changes[Next->Index].NewlinesBefore == 0) { 1051 Changes[Next->Index].Spaces = (CellWidth - ThisWidth); 1052 Changes[Next->Index].Spaces += (i > 0) ? 1 : 0; 1053 } 1054 alignToStartOfCell(Next->Index, Next->EndIndex); 1055 } 1056 } 1057 } 1058 } 1059 1060 void WhitespaceManager::alignArrayInitializersLeftJustified( 1061 CellDescriptions &&CellDescs) { 1062 auto &Cells = CellDescs.Cells; 1063 1064 // Now go through and fixup the spaces. 1065 auto *CellIter = Cells.begin(); 1066 // The first cell needs to be against the left brace. 1067 if (Changes[CellIter->Index].NewlinesBefore == 0) 1068 Changes[CellIter->Index].Spaces = 0; 1069 else 1070 Changes[CellIter->Index].Spaces = CellDescs.InitialSpaces; 1071 ++CellIter; 1072 for (auto i = 1U; i < CellDescs.CellCount; i++, ++CellIter) { 1073 auto MaxNetWidth = getMaximumNetWidth( 1074 Cells.begin(), CellIter, CellDescs.InitialSpaces, CellDescs.CellCount); 1075 auto ThisNetWidth = 1076 getNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces); 1077 if (Changes[CellIter->Index].NewlinesBefore == 0) { 1078 Changes[CellIter->Index].Spaces = 1079 MaxNetWidth - ThisNetWidth + 1080 (Changes[CellIter->Index].Tok->isNot(tok::r_brace) ? 1 : 0); 1081 } 1082 auto RowCount = 1U; 1083 auto Offset = std::distance(Cells.begin(), CellIter); 1084 for (const auto *Next = CellIter->NextColumnElement; Next != nullptr; 1085 Next = Next->NextColumnElement) { 1086 auto *Start = (Cells.begin() + RowCount * CellDescs.CellCount); 1087 auto *End = Start + Offset; 1088 auto ThisNetWidth = getNetWidth(Start, End, CellDescs.InitialSpaces); 1089 if (Changes[Next->Index].NewlinesBefore == 0) { 1090 Changes[Next->Index].Spaces = 1091 MaxNetWidth - ThisNetWidth + 1092 (Changes[Next->Index].Tok->isNot(tok::r_brace) ? 1 : 0); 1093 } 1094 ++RowCount; 1095 } 1096 } 1097 } 1098 1099 bool WhitespaceManager::isSplitCell(const CellDescription &Cell) { 1100 if (Cell.HasSplit) 1101 return true; 1102 for (const auto *Next = Cell.NextColumnElement; Next != nullptr; 1103 Next = Next->NextColumnElement) { 1104 if (Next->HasSplit) 1105 return true; 1106 } 1107 return false; 1108 } 1109 1110 WhitespaceManager::CellDescriptions WhitespaceManager::getCells(unsigned Start, 1111 unsigned End) { 1112 1113 unsigned Depth = 0; 1114 unsigned Cell = 0; 1115 unsigned CellCount = 0; 1116 unsigned InitialSpaces = 0; 1117 unsigned InitialTokenLength = 0; 1118 unsigned EndSpaces = 0; 1119 SmallVector<CellDescription> Cells; 1120 const FormatToken *MatchingParen = nullptr; 1121 for (unsigned i = Start; i < End; ++i) { 1122 auto &C = Changes[i]; 1123 if (C.Tok->is(tok::l_brace)) 1124 ++Depth; 1125 else if (C.Tok->is(tok::r_brace)) 1126 --Depth; 1127 if (Depth == 2) { 1128 if (C.Tok->is(tok::l_brace)) { 1129 Cell = 0; 1130 MatchingParen = C.Tok->MatchingParen; 1131 if (InitialSpaces == 0) { 1132 InitialSpaces = C.Spaces + C.TokenLength; 1133 InitialTokenLength = C.TokenLength; 1134 auto j = i - 1; 1135 for (; Changes[j].NewlinesBefore == 0 && j > Start; --j) { 1136 InitialSpaces += Changes[j].Spaces + Changes[j].TokenLength; 1137 InitialTokenLength += Changes[j].TokenLength; 1138 } 1139 if (C.NewlinesBefore == 0) { 1140 InitialSpaces += Changes[j].Spaces + Changes[j].TokenLength; 1141 InitialTokenLength += Changes[j].TokenLength; 1142 } 1143 } 1144 } else if (C.Tok->is(tok::comma)) { 1145 if (!Cells.empty()) 1146 Cells.back().EndIndex = i; 1147 if (C.Tok->getNextNonComment()->isNot(tok::r_brace)) // dangling comma 1148 ++Cell; 1149 } 1150 } else if (Depth == 1) { 1151 if (C.Tok == MatchingParen) { 1152 if (!Cells.empty()) 1153 Cells.back().EndIndex = i; 1154 Cells.push_back(CellDescription{i, ++Cell, i + 1, false, nullptr}); 1155 CellCount = C.Tok->Previous->isNot(tok::comma) ? Cell + 1 : Cell; 1156 // Go to the next non-comment and ensure there is a break in front 1157 const auto *NextNonComment = C.Tok->getNextNonComment(); 1158 while (NextNonComment->is(tok::comma)) 1159 NextNonComment = NextNonComment->getNextNonComment(); 1160 auto j = i; 1161 while (Changes[j].Tok != NextNonComment && j < End) 1162 j++; 1163 if (j < End && Changes[j].NewlinesBefore == 0 && 1164 Changes[j].Tok->isNot(tok::r_brace)) { 1165 Changes[j].NewlinesBefore = 1; 1166 // Account for the added token lengths 1167 Changes[j].Spaces = InitialSpaces - InitialTokenLength; 1168 } 1169 } else if (C.Tok->is(tok::comment)) { 1170 // Trailing comments stay at a space past the last token 1171 C.Spaces = Changes[i - 1].Tok->is(tok::comma) ? 1 : 2; 1172 } else if (C.Tok->is(tok::l_brace)) { 1173 // We need to make sure that the ending braces is aligned to the 1174 // start of our initializer 1175 auto j = i - 1; 1176 for (; j > 0 && !Changes[j].Tok->ArrayInitializerLineStart; --j) 1177 ; // Nothing the loop does the work 1178 EndSpaces = Changes[j].Spaces; 1179 } 1180 } else if (Depth == 0 && C.Tok->is(tok::r_brace)) { 1181 C.NewlinesBefore = 1; 1182 C.Spaces = EndSpaces; 1183 } 1184 if (C.Tok->StartsColumn) { 1185 // This gets us past tokens that have been split over multiple 1186 // lines 1187 bool HasSplit = false; 1188 if (Changes[i].NewlinesBefore > 0) { 1189 // So if we split a line previously and the tail line + this token is 1190 // less then the column limit we remove the split here and just put 1191 // the column start at a space past the comma 1192 // 1193 // FIXME This if branch covers the cases where the column is not 1194 // the first column. This leads to weird pathologies like the formatting 1195 // auto foo = Items{ 1196 // Section{ 1197 // 0, bar(), 1198 // } 1199 // }; 1200 // Well if it doesn't lead to that it's indicative that the line 1201 // breaking should be revisited. Unfortunately alot of other options 1202 // interact with this 1203 auto j = i - 1; 1204 if ((j - 1) > Start && Changes[j].Tok->is(tok::comma) && 1205 Changes[j - 1].NewlinesBefore > 0) { 1206 --j; 1207 auto LineLimit = Changes[j].Spaces + Changes[j].TokenLength; 1208 if (LineLimit < Style.ColumnLimit) { 1209 Changes[i].NewlinesBefore = 0; 1210 Changes[i].Spaces = 1; 1211 } 1212 } 1213 } 1214 while (Changes[i].NewlinesBefore > 0 && Changes[i].Tok == C.Tok) { 1215 Changes[i].Spaces = InitialSpaces; 1216 ++i; 1217 HasSplit = true; 1218 } 1219 if (Changes[i].Tok != C.Tok) 1220 --i; 1221 Cells.push_back(CellDescription{i, Cell, i, HasSplit, nullptr}); 1222 } 1223 } 1224 1225 return linkCells({Cells, CellCount, InitialSpaces}); 1226 } 1227 1228 unsigned WhitespaceManager::calculateCellWidth(unsigned Start, unsigned End, 1229 bool WithSpaces) const { 1230 unsigned CellWidth = 0; 1231 for (auto i = Start; i < End; i++) { 1232 if (Changes[i].NewlinesBefore > 0) 1233 CellWidth = 0; 1234 CellWidth += Changes[i].TokenLength; 1235 CellWidth += (WithSpaces ? Changes[i].Spaces : 0); 1236 } 1237 return CellWidth; 1238 } 1239 1240 void WhitespaceManager::alignToStartOfCell(unsigned Start, unsigned End) { 1241 if ((End - Start) <= 1) 1242 return; 1243 // If the line is broken anywhere in there make sure everything 1244 // is aligned to the parent 1245 for (auto i = Start + 1; i < End; i++) { 1246 if (Changes[i].NewlinesBefore > 0) 1247 Changes[i].Spaces = Changes[Start].Spaces; 1248 } 1249 } 1250 1251 WhitespaceManager::CellDescriptions 1252 WhitespaceManager::linkCells(CellDescriptions &&CellDesc) { 1253 auto &Cells = CellDesc.Cells; 1254 for (auto *CellIter = Cells.begin(); CellIter != Cells.end(); ++CellIter) { 1255 if (CellIter->NextColumnElement == nullptr && 1256 ((CellIter + 1) != Cells.end())) { 1257 for (auto *NextIter = CellIter + 1; NextIter != Cells.end(); ++NextIter) { 1258 if (NextIter->Cell == CellIter->Cell) { 1259 CellIter->NextColumnElement = &(*NextIter); 1260 break; 1261 } 1262 } 1263 } 1264 } 1265 return std::move(CellDesc); 1266 } 1267 1268 void WhitespaceManager::generateChanges() { 1269 for (unsigned i = 0, e = Changes.size(); i != e; ++i) { 1270 const Change &C = Changes[i]; 1271 if (i > 0 && Changes[i - 1].OriginalWhitespaceRange.getBegin() == 1272 C.OriginalWhitespaceRange.getBegin()) { 1273 // Do not generate two replacements for the same location. 1274 continue; 1275 } 1276 if (C.CreateReplacement) { 1277 std::string ReplacementText = C.PreviousLinePostfix; 1278 if (C.ContinuesPPDirective) 1279 appendEscapedNewlineText(ReplacementText, C.NewlinesBefore, 1280 C.PreviousEndOfTokenColumn, 1281 C.EscapedNewlineColumn); 1282 else 1283 appendNewlineText(ReplacementText, C.NewlinesBefore); 1284 appendIndentText( 1285 ReplacementText, C.Tok->IndentLevel, std::max(0, C.Spaces), 1286 C.StartOfTokenColumn - std::max(0, C.Spaces), C.IsAligned); 1287 ReplacementText.append(C.CurrentLinePrefix); 1288 storeReplacement(C.OriginalWhitespaceRange, ReplacementText); 1289 } 1290 } 1291 } 1292 1293 void WhitespaceManager::storeReplacement(SourceRange Range, StringRef Text) { 1294 unsigned WhitespaceLength = SourceMgr.getFileOffset(Range.getEnd()) - 1295 SourceMgr.getFileOffset(Range.getBegin()); 1296 // Don't create a replacement, if it does not change anything. 1297 if (StringRef(SourceMgr.getCharacterData(Range.getBegin()), 1298 WhitespaceLength) == Text) 1299 return; 1300 auto Err = Replaces.add(tooling::Replacement( 1301 SourceMgr, CharSourceRange::getCharRange(Range), Text)); 1302 // FIXME: better error handling. For now, just print an error message in the 1303 // release version. 1304 if (Err) { 1305 llvm::errs() << llvm::toString(std::move(Err)) << "\n"; 1306 assert(false); 1307 } 1308 } 1309 1310 void WhitespaceManager::appendNewlineText(std::string &Text, 1311 unsigned Newlines) { 1312 for (unsigned i = 0; i < Newlines; ++i) 1313 Text.append(UseCRLF ? "\r\n" : "\n"); 1314 } 1315 1316 void WhitespaceManager::appendEscapedNewlineText( 1317 std::string &Text, unsigned Newlines, unsigned PreviousEndOfTokenColumn, 1318 unsigned EscapedNewlineColumn) { 1319 if (Newlines > 0) { 1320 unsigned Spaces = 1321 std::max<int>(1, EscapedNewlineColumn - PreviousEndOfTokenColumn - 1); 1322 for (unsigned i = 0; i < Newlines; ++i) { 1323 Text.append(Spaces, ' '); 1324 Text.append(UseCRLF ? "\\\r\n" : "\\\n"); 1325 Spaces = std::max<int>(0, EscapedNewlineColumn - 1); 1326 } 1327 } 1328 } 1329 1330 void WhitespaceManager::appendIndentText(std::string &Text, 1331 unsigned IndentLevel, unsigned Spaces, 1332 unsigned WhitespaceStartColumn, 1333 bool IsAligned) { 1334 switch (Style.UseTab) { 1335 case FormatStyle::UT_Never: 1336 Text.append(Spaces, ' '); 1337 break; 1338 case FormatStyle::UT_Always: { 1339 if (Style.TabWidth) { 1340 unsigned FirstTabWidth = 1341 Style.TabWidth - WhitespaceStartColumn % Style.TabWidth; 1342 1343 // Insert only spaces when we want to end up before the next tab. 1344 if (Spaces < FirstTabWidth || Spaces == 1) { 1345 Text.append(Spaces, ' '); 1346 break; 1347 } 1348 // Align to the next tab. 1349 Spaces -= FirstTabWidth; 1350 Text.append("\t"); 1351 1352 Text.append(Spaces / Style.TabWidth, '\t'); 1353 Text.append(Spaces % Style.TabWidth, ' '); 1354 } else if (Spaces == 1) { 1355 Text.append(Spaces, ' '); 1356 } 1357 break; 1358 } 1359 case FormatStyle::UT_ForIndentation: 1360 if (WhitespaceStartColumn == 0) { 1361 unsigned Indentation = IndentLevel * Style.IndentWidth; 1362 Spaces = appendTabIndent(Text, Spaces, Indentation); 1363 } 1364 Text.append(Spaces, ' '); 1365 break; 1366 case FormatStyle::UT_ForContinuationAndIndentation: 1367 if (WhitespaceStartColumn == 0) 1368 Spaces = appendTabIndent(Text, Spaces, Spaces); 1369 Text.append(Spaces, ' '); 1370 break; 1371 case FormatStyle::UT_AlignWithSpaces: 1372 if (WhitespaceStartColumn == 0) { 1373 unsigned Indentation = 1374 IsAligned ? IndentLevel * Style.IndentWidth : Spaces; 1375 Spaces = appendTabIndent(Text, Spaces, Indentation); 1376 } 1377 Text.append(Spaces, ' '); 1378 break; 1379 } 1380 } 1381 1382 unsigned WhitespaceManager::appendTabIndent(std::string &Text, unsigned Spaces, 1383 unsigned Indentation) { 1384 // This happens, e.g. when a line in a block comment is indented less than the 1385 // first one. 1386 if (Indentation > Spaces) 1387 Indentation = Spaces; 1388 if (Style.TabWidth) { 1389 unsigned Tabs = Indentation / Style.TabWidth; 1390 Text.append(Tabs, '\t'); 1391 Spaces -= Tabs * Style.TabWidth; 1392 } 1393 return Spaces; 1394 } 1395 1396 } // namespace format 1397 } // namespace clang 1398