clang  19.0.0git
WhitespaceManager.cpp
Go to the documentation of this file.
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 
23  const Change &C1, const Change &C2) const {
24  return SourceMgr.isBeforeInTranslationUnit(
29  SourceMgr.isBeforeInTranslationUnit(
32 }
33 
35  bool CreateReplacement,
37  int Spaces, unsigned StartOfTokenColumn,
38  unsigned NewlinesBefore,
39  StringRef PreviousLinePostfix,
40  StringRef CurrentLinePrefix, bool IsAligned,
51 }
52 
54  unsigned Spaces,
55  unsigned StartOfTokenColumn,
56  bool IsAligned, bool InPPDirective) {
57  if (Tok.Finalized || (Tok.MacroCtx && Tok.MacroCtx->Role == MR_ExpandedArg))
58  return;
59  Tok.setDecision((Newlines > 0) ? FD_Break : FD_Continue);
60  Changes.push_back(Change(Tok, /*CreateReplacement=*/true, Tok.WhitespaceRange,
61  Spaces, StartOfTokenColumn, Newlines, "", "",
62  IsAligned, InPPDirective && !Tok.IsFirst,
63  /*IsInsideToken=*/false));
64 }
65 
67  bool InPPDirective) {
68  if (Tok.Finalized || (Tok.MacroCtx && Tok.MacroCtx->Role == MR_ExpandedArg))
69  return;
70  Changes.push_back(Change(Tok, /*CreateReplacement=*/false,
71  Tok.WhitespaceRange, /*Spaces=*/0,
72  Tok.OriginalColumn, Tok.NewlinesBefore, "", "",
73  /*IsAligned=*/false, InPPDirective && !Tok.IsFirst,
74  /*IsInsideToken=*/false));
75 }
76 
77 llvm::Error
79  return Replaces.add(Replacement);
80 }
81 
82 bool WhitespaceManager::inputUsesCRLF(StringRef Text, bool DefaultToCRLF) {
83  size_t LF = Text.count('\n');
84  size_t CR = Text.count('\r') * 2;
85  return LF == CR ? DefaultToCRLF : CR > LF;
86 }
87 
89  const FormatToken &Tok, unsigned Offset, unsigned ReplaceChars,
90  StringRef PreviousPostfix, StringRef CurrentPrefix, bool InPPDirective,
91  unsigned Newlines, int Spaces) {
92  if (Tok.Finalized || (Tok.MacroCtx && Tok.MacroCtx->Role == MR_ExpandedArg))
93  return;
95  Changes.push_back(
96  Change(Tok, /*CreateReplacement=*/true,
97  SourceRange(Start, Start.getLocWithOffset(ReplaceChars)), Spaces,
98  std::max(0, Spaces), Newlines, PreviousPostfix, CurrentPrefix,
99  /*IsAligned=*/true, InPPDirective && !Tok.IsFirst,
100  /*IsInsideToken=*/true));
101 }
102 
104  if (Changes.empty())
105  return Replaces;
106 
107  llvm::sort(Changes, Change::IsBeforeInFile(SourceMgr));
108  calculateLineBreakInformation();
109  alignConsecutiveMacros();
110  alignConsecutiveShortCaseStatements(/*IsExpr=*/true);
111  alignConsecutiveShortCaseStatements(/*IsExpr=*/false);
112  alignConsecutiveDeclarations();
113  alignConsecutiveBitFields();
114  alignConsecutiveAssignments();
115  if (Style.isTableGen()) {
116  alignConsecutiveTableGenBreakingDAGArgColons();
117  alignConsecutiveTableGenCondOperatorColons();
118  alignConsecutiveTableGenDefinitions();
119  }
120  alignChainedConditionals();
121  alignTrailingComments();
122  alignEscapedNewlines();
123  alignArrayInitializers();
124  generateChanges();
125 
126  return Replaces;
127 }
128 
129 void WhitespaceManager::calculateLineBreakInformation() {
130  Changes[0].PreviousEndOfTokenColumn = 0;
131  Change *LastOutsideTokenChange = &Changes[0];
132  for (unsigned I = 1, e = Changes.size(); I != e; ++I) {
133  auto &C = Changes[I];
134  auto &P = Changes[I - 1];
135  auto &PrevTokLength = P.TokenLength;
136  SourceLocation OriginalWhitespaceStart =
137  C.OriginalWhitespaceRange.getBegin();
138  SourceLocation PreviousOriginalWhitespaceEnd =
139  P.OriginalWhitespaceRange.getEnd();
140  unsigned OriginalWhitespaceStartOffset =
141  SourceMgr.getFileOffset(OriginalWhitespaceStart);
142  unsigned PreviousOriginalWhitespaceEndOffset =
143  SourceMgr.getFileOffset(PreviousOriginalWhitespaceEnd);
144  assert(PreviousOriginalWhitespaceEndOffset <=
145  OriginalWhitespaceStartOffset);
146  const char *const PreviousOriginalWhitespaceEndData =
147  SourceMgr.getCharacterData(PreviousOriginalWhitespaceEnd);
148  StringRef Text(PreviousOriginalWhitespaceEndData,
149  SourceMgr.getCharacterData(OriginalWhitespaceStart) -
150  PreviousOriginalWhitespaceEndData);
151  // Usually consecutive changes would occur in consecutive tokens. This is
152  // not the case however when analyzing some preprocessor runs of the
153  // annotated lines. For example, in this code:
154  //
155  // #if A // line 1
156  // int i = 1;
157  // #else B // line 2
158  // int i = 2;
159  // #endif // line 3
160  //
161  // one of the runs will produce the sequence of lines marked with line 1, 2
162  // and 3. So the two consecutive whitespace changes just before '// line 2'
163  // and before '#endif // line 3' span multiple lines and tokens:
164  //
165  // #else B{change X}[// line 2
166  // int i = 2;
167  // ]{change Y}#endif // line 3
168  //
169  // For this reason, if the text between consecutive changes spans multiple
170  // newlines, the token length must be adjusted to the end of the original
171  // line of the token.
172  auto NewlinePos = Text.find_first_of('\n');
173  if (NewlinePos == StringRef::npos) {
174  PrevTokLength = OriginalWhitespaceStartOffset -
175  PreviousOriginalWhitespaceEndOffset +
176  C.PreviousLinePostfix.size() + P.CurrentLinePrefix.size();
177  if (!P.IsInsideToken)
178  PrevTokLength = std::min(PrevTokLength, P.Tok->ColumnWidth);
179  } else {
180  PrevTokLength = NewlinePos + P.CurrentLinePrefix.size();
181  }
182 
183  // If there are multiple changes in this token, sum up all the changes until
184  // the end of the line.
185  if (P.IsInsideToken && P.NewlinesBefore == 0)
186  LastOutsideTokenChange->TokenLength += PrevTokLength + P.Spaces;
187  else
188  LastOutsideTokenChange = &P;
189 
190  C.PreviousEndOfTokenColumn = P.StartOfTokenColumn + PrevTokLength;
191 
192  P.IsTrailingComment =
193  (C.NewlinesBefore > 0 || C.Tok->is(tok::eof) ||
194  (C.IsInsideToken && C.Tok->is(tok::comment))) &&
195  P.Tok->is(tok::comment) &&
196  // FIXME: This is a dirty hack. The problem is that
197  // BreakableLineCommentSection does comment reflow changes and here is
198  // the aligning of trailing comments. Consider the case where we reflow
199  // the second line up in this example:
200  //
201  // // line 1
202  // // line 2
203  //
204  // That amounts to 2 changes by BreakableLineCommentSection:
205  // - the first, delimited by (), for the whitespace between the tokens,
206  // - and second, delimited by [], for the whitespace at the beginning
207  // of the second token:
208  //
209  // // line 1(
210  // )[// ]line 2
211  //
212  // So in the end we have two changes like this:
213  //
214  // // line1()[ ]line 2
215  //
216  // Note that the OriginalWhitespaceStart of the second change is the
217  // same as the PreviousOriginalWhitespaceEnd of the first change.
218  // In this case, the below check ensures that the second change doesn't
219  // get treated as a trailing comment change here, since this might
220  // trigger additional whitespace to be wrongly inserted before "line 2"
221  // by the comment aligner here.
222  //
223  // For a proper solution we need a mechanism to say to WhitespaceManager
224  // that a particular change breaks the current sequence of trailing
225  // comments.
226  OriginalWhitespaceStart != PreviousOriginalWhitespaceEnd;
227  }
228  // FIXME: The last token is currently not always an eof token; in those
229  // cases, setting TokenLength of the last token to 0 is wrong.
230  Changes.back().TokenLength = 0;
231  Changes.back().IsTrailingComment = Changes.back().Tok->is(tok::comment);
232 
233  const WhitespaceManager::Change *LastBlockComment = nullptr;
234  for (auto &Change : Changes) {
235  // Reset the IsTrailingComment flag for changes inside of trailing comments
236  // so they don't get realigned later. Comment line breaks however still need
237  // to be aligned.
238  if (Change.IsInsideToken && Change.NewlinesBefore == 0)
239  Change.IsTrailingComment = false;
240  Change.StartOfBlockComment = nullptr;
241  Change.IndentationOffset = 0;
242  if (Change.Tok->is(tok::comment)) {
243  if (Change.Tok->is(TT_LineComment) || !Change.IsInsideToken) {
244  LastBlockComment = &Change;
245  } else if ((Change.StartOfBlockComment = LastBlockComment)) {
246  Change.IndentationOffset =
247  Change.StartOfTokenColumn -
248  Change.StartOfBlockComment->StartOfTokenColumn;
249  }
250  } else {
251  LastBlockComment = nullptr;
252  }
253  }
254 
255  // Compute conditional nesting level
256  // Level is increased for each conditional, unless this conditional continues
257  // a chain of conditional, i.e. starts immediately after the colon of another
258  // conditional.
259  SmallVector<bool, 16> ScopeStack;
260  int ConditionalsLevel = 0;
261  for (auto &Change : Changes) {
262  for (unsigned i = 0, e = Change.Tok->FakeLParens.size(); i != e; ++i) {
263  bool isNestedConditional =
264  Change.Tok->FakeLParens[e - 1 - i] == prec::Conditional &&
265  !(i == 0 && Change.Tok->Previous &&
266  Change.Tok->Previous->is(TT_ConditionalExpr) &&
267  Change.Tok->Previous->is(tok::colon));
268  if (isNestedConditional)
269  ++ConditionalsLevel;
270  ScopeStack.push_back(isNestedConditional);
271  }
272 
273  Change.ConditionalsLevel = ConditionalsLevel;
274 
275  for (unsigned i = Change.Tok->FakeRParens; i > 0 && ScopeStack.size(); --i)
276  if (ScopeStack.pop_back_val())
277  --ConditionalsLevel;
278  }
279 }
280 
281 // Align a single sequence of tokens, see AlignTokens below.
282 // Column - The token for which Matches returns true is moved to this column.
283 // RightJustify - Whether it is the token's right end or left end that gets
284 // moved to that column.
285 template <typename F>
286 static void
287 AlignTokenSequence(const FormatStyle &Style, unsigned Start, unsigned End,
288  unsigned Column, bool RightJustify, F &&Matches,
290  bool FoundMatchOnLine = false;
291  int Shift = 0;
292 
293  // ScopeStack keeps track of the current scope depth. It contains indices of
294  // the first token on each scope.
295  // We only run the "Matches" function on tokens from the outer-most scope.
296  // However, we do need to pay special attention to one class of tokens
297  // that are not in the outer-most scope, and that is function parameters
298  // which are split across multiple lines, as illustrated by this example:
299  // double a(int x);
300  // int b(int y,
301  // double z);
302  // In the above example, we need to take special care to ensure that
303  // 'double z' is indented along with it's owning function 'b'.
304  // The same holds for calling a function:
305  // double a = foo(x);
306  // int b = bar(foo(y),
307  // foor(z));
308  // Similar for broken string literals:
309  // double x = 3.14;
310  // auto s = "Hello"
311  // "World";
312  // Special handling is required for 'nested' ternary operators.
313  SmallVector<unsigned, 16> ScopeStack;
314 
315  for (unsigned i = Start; i != End; ++i) {
316  auto &CurrentChange = Changes[i];
317  if (ScopeStack.size() != 0 &&
318  CurrentChange.indentAndNestingLevel() <
319  Changes[ScopeStack.back()].indentAndNestingLevel()) {
320  ScopeStack.pop_back();
321  }
322 
323  // Compare current token to previous non-comment token to ensure whether
324  // it is in a deeper scope or not.
325  unsigned PreviousNonComment = i - 1;
326  while (PreviousNonComment > Start &&
327  Changes[PreviousNonComment].Tok->is(tok::comment)) {
328  --PreviousNonComment;
329  }
330  if (i != Start && CurrentChange.indentAndNestingLevel() >
331  Changes[PreviousNonComment].indentAndNestingLevel()) {
332  ScopeStack.push_back(i);
333  }
334 
335  bool InsideNestedScope = ScopeStack.size() != 0;
336  bool ContinuedStringLiteral = i > Start &&
337  CurrentChange.Tok->is(tok::string_literal) &&
338  Changes[i - 1].Tok->is(tok::string_literal);
339  bool SkipMatchCheck = InsideNestedScope || ContinuedStringLiteral;
340 
341  if (CurrentChange.NewlinesBefore > 0 && !SkipMatchCheck) {
342  Shift = 0;
343  FoundMatchOnLine = false;
344  }
345 
346  // If this is the first matching token to be aligned, remember by how many
347  // spaces it has to be shifted, so the rest of the changes on the line are
348  // shifted by the same amount
349  if (!FoundMatchOnLine && !SkipMatchCheck && Matches(CurrentChange)) {
350  FoundMatchOnLine = true;
351  Shift = Column - (RightJustify ? CurrentChange.TokenLength : 0) -
352  CurrentChange.StartOfTokenColumn;
353  CurrentChange.Spaces += Shift;
354  // FIXME: This is a workaround that should be removed when we fix
355  // http://llvm.org/PR53699. An assertion later below verifies this.
356  if (CurrentChange.NewlinesBefore == 0) {
357  CurrentChange.Spaces =
358  std::max(CurrentChange.Spaces,
359  static_cast<int>(CurrentChange.Tok->SpacesRequiredBefore));
360  }
361  }
362 
363  if (Shift == 0)
364  continue;
365 
366  // This is for function parameters that are split across multiple lines,
367  // as mentioned in the ScopeStack comment.
368  if (InsideNestedScope && CurrentChange.NewlinesBefore > 0) {
369  unsigned ScopeStart = ScopeStack.back();
370  auto ShouldShiftBeAdded = [&] {
371  // Function declaration
372  if (Changes[ScopeStart - 1].Tok->is(TT_FunctionDeclarationName))
373  return true;
374 
375  // Lambda.
376  if (Changes[ScopeStart - 1].Tok->is(TT_LambdaLBrace))
377  return false;
378 
379  // Continued function declaration
380  if (ScopeStart > Start + 1 &&
381  Changes[ScopeStart - 2].Tok->is(TT_FunctionDeclarationName)) {
382  return true;
383  }
384 
385  // Continued (template) function call.
386  if (ScopeStart > Start + 1 &&
387  Changes[ScopeStart - 2].Tok->isOneOf(tok::identifier,
388  TT_TemplateCloser) &&
389  Changes[ScopeStart - 1].Tok->is(tok::l_paren) &&
390  Changes[ScopeStart].Tok->isNot(TT_LambdaLSquare)) {
391  if (CurrentChange.Tok->MatchingParen &&
392  CurrentChange.Tok->MatchingParen->is(TT_LambdaLBrace)) {
393  return false;
394  }
395  if (Changes[ScopeStart].NewlinesBefore > 0)
396  return false;
397  if (CurrentChange.Tok->is(tok::l_brace) &&
398  CurrentChange.Tok->is(BK_BracedInit)) {
399  return true;
400  }
401  return Style.BinPackArguments;
402  }
403 
404  // Ternary operator
405  if (CurrentChange.Tok->is(TT_ConditionalExpr))
406  return true;
407 
408  // Period Initializer .XXX = 1.
409  if (CurrentChange.Tok->is(TT_DesignatedInitializerPeriod))
410  return true;
411 
412  // Continued ternary operator
413  if (CurrentChange.Tok->Previous &&
414  CurrentChange.Tok->Previous->is(TT_ConditionalExpr)) {
415  return true;
416  }
417 
418  // Continued direct-list-initialization using braced list.
419  if (ScopeStart > Start + 1 &&
420  Changes[ScopeStart - 2].Tok->is(tok::identifier) &&
421  Changes[ScopeStart - 1].Tok->is(tok::l_brace) &&
422  CurrentChange.Tok->is(tok::l_brace) &&
423  CurrentChange.Tok->is(BK_BracedInit)) {
424  return true;
425  }
426 
427  // Continued braced list.
428  if (ScopeStart > Start + 1 &&
429  Changes[ScopeStart - 2].Tok->isNot(tok::identifier) &&
430  Changes[ScopeStart - 1].Tok->is(tok::l_brace) &&
431  CurrentChange.Tok->isNot(tok::r_brace)) {
432  for (unsigned OuterScopeStart : llvm::reverse(ScopeStack)) {
433  // Lambda.
434  if (OuterScopeStart > Start &&
435  Changes[OuterScopeStart - 1].Tok->is(TT_LambdaLBrace)) {
436  return false;
437  }
438  }
439  if (Changes[ScopeStart].NewlinesBefore > 0)
440  return false;
441  return true;
442  }
443 
444  // Continued template parameter.
445  if (Changes[ScopeStart - 1].Tok->is(TT_TemplateOpener))
446  return true;
447 
448  return false;
449  };
450 
451  if (ShouldShiftBeAdded())
452  CurrentChange.Spaces += Shift;
453  }
454 
455  if (ContinuedStringLiteral)
456  CurrentChange.Spaces += Shift;
457 
458  // We should not remove required spaces unless we break the line before.
459  assert(Shift > 0 || Changes[i].NewlinesBefore > 0 ||
460  CurrentChange.Spaces >=
461  static_cast<int>(Changes[i].Tok->SpacesRequiredBefore) ||
462  CurrentChange.Tok->is(tok::eof));
463 
464  CurrentChange.StartOfTokenColumn += Shift;
465  if (i + 1 != Changes.size())
466  Changes[i + 1].PreviousEndOfTokenColumn += Shift;
467 
468  // If PointerAlignment is PAS_Right, keep *s or &s next to the token,
469  // except if the token is equal, then a space is needed.
470  if ((Style.PointerAlignment == FormatStyle::PAS_Right ||
471  Style.ReferenceAlignment == FormatStyle::RAS_Right) &&
472  CurrentChange.Spaces != 0 && CurrentChange.Tok->isNot(tok::equal)) {
473  const bool ReferenceNotRightAligned =
474  Style.ReferenceAlignment != FormatStyle::RAS_Right &&
475  Style.ReferenceAlignment != FormatStyle::RAS_Pointer;
476  for (int Previous = i - 1;
477  Previous >= 0 && Changes[Previous].Tok->is(TT_PointerOrReference);
478  --Previous) {
479  assert(Changes[Previous].Tok->isPointerOrReference());
480  if (Changes[Previous].Tok->isNot(tok::star)) {
481  if (ReferenceNotRightAligned)
482  continue;
483  } else if (Style.PointerAlignment != FormatStyle::PAS_Right) {
484  continue;
485  }
486  Changes[Previous + 1].Spaces -= Shift;
487  Changes[Previous].Spaces += Shift;
488  Changes[Previous].StartOfTokenColumn += Shift;
489  }
490  }
491  }
492 }
493 
494 // Walk through a subset of the changes, starting at StartAt, and find
495 // sequences of matching tokens to align. To do so, keep track of the lines and
496 // whether or not a matching token was found on a line. If a matching token is
497 // found, extend the current sequence. If the current line cannot be part of a
498 // sequence, e.g. because there is an empty line before it or it contains only
499 // non-matching tokens, finalize the previous sequence.
500 // The value returned is the token on which we stopped, either because we
501 // exhausted all items inside Changes, or because we hit a scope level higher
502 // than our initial scope.
503 // This function is recursive. Each invocation processes only the scope level
504 // equal to the initial level, which is the level of Changes[StartAt].
505 // If we encounter a scope level greater than the initial level, then we call
506 // ourselves recursively, thereby avoiding the pollution of the current state
507 // with the alignment requirements of the nested sub-level. This recursive
508 // behavior is necessary for aligning function prototypes that have one or more
509 // arguments.
510 // If this function encounters a scope level less than the initial level,
511 // it returns the current position.
512 // There is a non-obvious subtlety in the recursive behavior: Even though we
513 // defer processing of nested levels to recursive invocations of this
514 // function, when it comes time to align a sequence of tokens, we run the
515 // alignment on the entire sequence, including the nested levels.
516 // When doing so, most of the nested tokens are skipped, because their
517 // alignment was already handled by the recursive invocations of this function.
518 // However, the special exception is that we do NOT skip function parameters
519 // that are split across multiple lines. See the test case in FormatTest.cpp
520 // that mentions "split function parameter alignment" for an example of this.
521 // When the parameter RightJustify is true, the operator will be
522 // right-justified. It is used to align compound assignments like `+=` and `=`.
523 // When RightJustify and ACS.PadOperators are true, operators in each block to
524 // be aligned will be padded on the left to the same length before aligning.
525 template <typename F>
526 static unsigned AlignTokens(const FormatStyle &Style, F &&Matches,
528  unsigned StartAt,
529  const FormatStyle::AlignConsecutiveStyle &ACS = {},
530  bool RightJustify = false) {
531  // We arrange each line in 3 parts. The operator to be aligned (the anchor),
532  // and text to its left and right. In the aligned text the width of each part
533  // will be the maximum of that over the block that has been aligned. Maximum
534  // widths of each part so far. When RightJustify is true and ACS.PadOperators
535  // is false, the part from start of line to the right end of the anchor.
536  // Otherwise, only the part to the left of the anchor. Including the space
537  // that exists on its left from the start. Not including the padding added on
538  // the left to right-justify the anchor.
539  unsigned WidthLeft = 0;
540  // The operator to be aligned when RightJustify is true and ACS.PadOperators
541  // is false. 0 otherwise.
542  unsigned WidthAnchor = 0;
543  // Width to the right of the anchor. Plus width of the anchor when
544  // RightJustify is false.
545  unsigned WidthRight = 0;
546 
547  // Line number of the start and the end of the current token sequence.
548  unsigned StartOfSequence = 0;
549  unsigned EndOfSequence = 0;
550 
551  // Measure the scope level (i.e. depth of (), [], {}) of the first token, and
552  // abort when we hit any token in a higher scope than the starting one.
553  auto IndentAndNestingLevel = StartAt < Changes.size()
554  ? Changes[StartAt].indentAndNestingLevel()
555  : std::tuple<unsigned, unsigned, unsigned>();
556 
557  // Keep track of the number of commas before the matching tokens, we will only
558  // align a sequence of matching tokens if they are preceded by the same number
559  // of commas.
560  unsigned CommasBeforeLastMatch = 0;
561  unsigned CommasBeforeMatch = 0;
562 
563  // Whether a matching token has been found on the current line.
564  bool FoundMatchOnLine = false;
565 
566  // Whether the current line consists purely of comments.
567  bool LineIsComment = true;
568 
569  // Aligns a sequence of matching tokens, on the MinColumn column.
570  //
571  // Sequences start from the first matching token to align, and end at the
572  // first token of the first line that doesn't need to be aligned.
573  //
574  // We need to adjust the StartOfTokenColumn of each Change that is on a line
575  // containing any matching token to be aligned and located after such token.
576  auto AlignCurrentSequence = [&] {
577  if (StartOfSequence > 0 && StartOfSequence < EndOfSequence) {
578  AlignTokenSequence(Style, StartOfSequence, EndOfSequence,
579  WidthLeft + WidthAnchor, RightJustify, Matches,
580  Changes);
581  }
582  WidthLeft = 0;
583  WidthAnchor = 0;
584  WidthRight = 0;
585  StartOfSequence = 0;
586  EndOfSequence = 0;
587  };
588 
589  unsigned i = StartAt;
590  for (unsigned e = Changes.size(); i != e; ++i) {
591  auto &CurrentChange = Changes[i];
592  if (CurrentChange.indentAndNestingLevel() < IndentAndNestingLevel)
593  break;
594 
595  if (CurrentChange.NewlinesBefore != 0) {
596  CommasBeforeMatch = 0;
597  EndOfSequence = i;
598 
599  // Whether to break the alignment sequence because of an empty line.
600  bool EmptyLineBreak =
601  (CurrentChange.NewlinesBefore > 1) && !ACS.AcrossEmptyLines;
602 
603  // Whether to break the alignment sequence because of a line without a
604  // match.
605  bool NoMatchBreak =
606  !FoundMatchOnLine && !(LineIsComment && ACS.AcrossComments);
607 
608  if (EmptyLineBreak || NoMatchBreak)
609  AlignCurrentSequence();
610 
611  // A new line starts, re-initialize line status tracking bools.
612  // Keep the match state if a string literal is continued on this line.
613  if (i == 0 || CurrentChange.Tok->isNot(tok::string_literal) ||
614  Changes[i - 1].Tok->isNot(tok::string_literal)) {
615  FoundMatchOnLine = false;
616  }
617  LineIsComment = true;
618  }
619 
620  if (CurrentChange.Tok->isNot(tok::comment))
621  LineIsComment = false;
622 
623  if (CurrentChange.Tok->is(tok::comma)) {
624  ++CommasBeforeMatch;
625  } else if (CurrentChange.indentAndNestingLevel() > IndentAndNestingLevel) {
626  // Call AlignTokens recursively, skipping over this scope block.
627  unsigned StoppedAt =
628  AlignTokens(Style, Matches, Changes, i, ACS, RightJustify);
629  i = StoppedAt - 1;
630  continue;
631  }
632 
633  if (!Matches(CurrentChange))
634  continue;
635 
636  // If there is more than one matching token per line, or if the number of
637  // preceding commas, do not match anymore, end the sequence.
638  if (FoundMatchOnLine || CommasBeforeMatch != CommasBeforeLastMatch)
639  AlignCurrentSequence();
640 
641  CommasBeforeLastMatch = CommasBeforeMatch;
642  FoundMatchOnLine = true;
643 
644  if (StartOfSequence == 0)
645  StartOfSequence = i;
646 
647  unsigned ChangeWidthLeft = CurrentChange.StartOfTokenColumn;
648  unsigned ChangeWidthAnchor = 0;
649  unsigned ChangeWidthRight = 0;
650  if (RightJustify)
651  if (ACS.PadOperators)
652  ChangeWidthAnchor = CurrentChange.TokenLength;
653  else
654  ChangeWidthLeft += CurrentChange.TokenLength;
655  else
656  ChangeWidthRight = CurrentChange.TokenLength;
657  for (unsigned j = i + 1; j != e && Changes[j].NewlinesBefore == 0; ++j) {
658  ChangeWidthRight += Changes[j].Spaces;
659  // Changes are generally 1:1 with the tokens, but a change could also be
660  // inside of a token, in which case it's counted more than once: once for
661  // the whitespace surrounding the token (!IsInsideToken) and once for
662  // each whitespace change within it (IsInsideToken).
663  // Therefore, changes inside of a token should only count the space.
664  if (!Changes[j].IsInsideToken)
665  ChangeWidthRight += Changes[j].TokenLength;
666  }
667 
668  // If we are restricted by the maximum column width, end the sequence.
669  unsigned NewLeft = std::max(ChangeWidthLeft, WidthLeft);
670  unsigned NewAnchor = std::max(ChangeWidthAnchor, WidthAnchor);
671  unsigned NewRight = std::max(ChangeWidthRight, WidthRight);
672  // `ColumnLimit == 0` means there is no column limit.
673  if (Style.ColumnLimit != 0 &&
674  Style.ColumnLimit < NewLeft + NewAnchor + NewRight) {
675  AlignCurrentSequence();
676  StartOfSequence = i;
677  WidthLeft = ChangeWidthLeft;
678  WidthAnchor = ChangeWidthAnchor;
679  WidthRight = ChangeWidthRight;
680  } else {
681  WidthLeft = NewLeft;
682  WidthAnchor = NewAnchor;
683  WidthRight = NewRight;
684  }
685  }
686 
687  EndOfSequence = i;
688  AlignCurrentSequence();
689  return i;
690 }
691 
692 // Aligns a sequence of matching tokens, on the MinColumn column.
693 //
694 // Sequences start from the first matching token to align, and end at the
695 // first token of the first line that doesn't need to be aligned.
696 //
697 // We need to adjust the StartOfTokenColumn of each Change that is on a line
698 // containing any matching token to be aligned and located after such token.
700  unsigned &StartOfSequence, unsigned &EndOfSequence, unsigned &MinColumn,
701  std::function<bool(const WhitespaceManager::Change &C)> Matches,
703  if (StartOfSequence > 0 && StartOfSequence < EndOfSequence) {
704  bool FoundMatchOnLine = false;
705  int Shift = 0;
706 
707  for (unsigned I = StartOfSequence; I != EndOfSequence; ++I) {
708  if (Changes[I].NewlinesBefore > 0) {
709  Shift = 0;
710  FoundMatchOnLine = false;
711  }
712 
713  // If this is the first matching token to be aligned, remember by how many
714  // spaces it has to be shifted, so the rest of the changes on the line are
715  // shifted by the same amount.
716  if (!FoundMatchOnLine && Matches(Changes[I])) {
717  FoundMatchOnLine = true;
718  Shift = MinColumn - Changes[I].StartOfTokenColumn;
719  Changes[I].Spaces += Shift;
720  }
721 
722  assert(Shift >= 0);
723  Changes[I].StartOfTokenColumn += Shift;
724  if (I + 1 != Changes.size())
725  Changes[I + 1].PreviousEndOfTokenColumn += Shift;
726  }
727  }
728 
729  MinColumn = 0;
730  StartOfSequence = 0;
731  EndOfSequence = 0;
732 }
733 
734 void WhitespaceManager::alignConsecutiveMacros() {
735  if (!Style.AlignConsecutiveMacros.Enabled)
736  return;
737 
738  auto AlignMacrosMatches = [](const Change &C) {
739  const FormatToken *Current = C.Tok;
740  unsigned SpacesRequiredBefore = 1;
741 
742  if (Current->SpacesRequiredBefore == 0 || !Current->Previous)
743  return false;
744 
745  Current = Current->Previous;
746 
747  // If token is a ")", skip over the parameter list, to the
748  // token that precedes the "("
749  if (Current->is(tok::r_paren) && Current->MatchingParen) {
750  Current = Current->MatchingParen->Previous;
751  SpacesRequiredBefore = 0;
752  }
753 
754  if (!Current || Current->isNot(tok::identifier))
755  return false;
756 
757  if (!Current->Previous || Current->Previous->isNot(tok::pp_define))
758  return false;
759 
760  // For a macro function, 0 spaces are required between the
761  // identifier and the lparen that opens the parameter list.
762  // For a simple macro, 1 space is required between the
763  // identifier and the first token of the defined value.
764  return Current->Next->SpacesRequiredBefore == SpacesRequiredBefore;
765  };
766 
767  unsigned MinColumn = 0;
768 
769  // Start and end of the token sequence we're processing.
770  unsigned StartOfSequence = 0;
771  unsigned EndOfSequence = 0;
772 
773  // Whether a matching token has been found on the current line.
774  bool FoundMatchOnLine = false;
775 
776  // Whether the current line consists only of comments
777  bool LineIsComment = true;
778 
779  unsigned I = 0;
780  for (unsigned E = Changes.size(); I != E; ++I) {
781  if (Changes[I].NewlinesBefore != 0) {
782  EndOfSequence = I;
783 
784  // Whether to break the alignment sequence because of an empty line.
785  bool EmptyLineBreak = (Changes[I].NewlinesBefore > 1) &&
786  !Style.AlignConsecutiveMacros.AcrossEmptyLines;
787 
788  // Whether to break the alignment sequence because of a line without a
789  // match.
790  bool NoMatchBreak =
791  !FoundMatchOnLine &&
792  !(LineIsComment && Style.AlignConsecutiveMacros.AcrossComments);
793 
794  if (EmptyLineBreak || NoMatchBreak) {
795  AlignMatchingTokenSequence(StartOfSequence, EndOfSequence, MinColumn,
796  AlignMacrosMatches, Changes);
797  }
798 
799  // A new line starts, re-initialize line status tracking bools.
800  FoundMatchOnLine = false;
801  LineIsComment = true;
802  }
803 
804  if (Changes[I].Tok->isNot(tok::comment))
805  LineIsComment = false;
806 
807  if (!AlignMacrosMatches(Changes[I]))
808  continue;
809 
810  FoundMatchOnLine = true;
811 
812  if (StartOfSequence == 0)
813  StartOfSequence = I;
814 
815  unsigned ChangeMinColumn = Changes[I].StartOfTokenColumn;
816  MinColumn = std::max(MinColumn, ChangeMinColumn);
817  }
818 
819  EndOfSequence = I;
820  AlignMatchingTokenSequence(StartOfSequence, EndOfSequence, MinColumn,
821  AlignMacrosMatches, Changes);
822 }
823 
824 void WhitespaceManager::alignConsecutiveAssignments() {
825  if (!Style.AlignConsecutiveAssignments.Enabled)
826  return;
827 
828  AlignTokens(
829  Style,
830  [&](const Change &C) {
831  // Do not align on equal signs that are first on a line.
832  if (C.NewlinesBefore > 0)
833  return false;
834 
835  // Do not align on equal signs that are last on a line.
836  if (&C != &Changes.back() && (&C + 1)->NewlinesBefore > 0)
837  return false;
838 
839  // Do not align operator= overloads.
840  FormatToken *Previous = C.Tok->getPreviousNonComment();
841  if (Previous && Previous->is(tok::kw_operator))
842  return false;
843 
844  return Style.AlignConsecutiveAssignments.AlignCompound
845  ? C.Tok->getPrecedence() == prec::Assignment
846  : (C.Tok->is(tok::equal) ||
847  // In Verilog the '<=' is not a compound assignment, thus
848  // it is aligned even when the AlignCompound option is not
849  // set.
850  (Style.isVerilog() && C.Tok->is(tok::lessequal) &&
851  C.Tok->getPrecedence() == prec::Assignment));
852  },
853  Changes, /*StartAt=*/0, Style.AlignConsecutiveAssignments,
854  /*RightJustify=*/true);
855 }
856 
857 void WhitespaceManager::alignConsecutiveBitFields() {
858  alignConsecutiveColons(Style.AlignConsecutiveBitFields, TT_BitFieldColon);
859 }
860 
861 void WhitespaceManager::alignConsecutiveColons(
862  const FormatStyle::AlignConsecutiveStyle &AlignStyle, TokenType Type) {
863  if (!AlignStyle.Enabled)
864  return;
865 
866  AlignTokens(
867  Style,
868  [&](Change const &C) {
869  // Do not align on ':' that is first on a line.
870  if (C.NewlinesBefore > 0)
871  return false;
872 
873  // Do not align on ':' that is last on a line.
874  if (&C != &Changes.back() && (&C + 1)->NewlinesBefore > 0)
875  return false;
876 
877  return C.Tok->is(Type);
878  },
879  Changes, /*StartAt=*/0, AlignStyle);
880 }
881 
882 void WhitespaceManager::alignConsecutiveShortCaseStatements(bool IsExpr) {
883  if (!Style.AlignConsecutiveShortCaseStatements.Enabled ||
884  !(IsExpr ? Style.AllowShortCaseExpressionOnASingleLine
885  : Style.AllowShortCaseLabelsOnASingleLine)) {
886  return;
887  }
888 
889  const auto Type = IsExpr ? TT_CaseLabelArrow : TT_CaseLabelColon;
890  const auto &Option = Style.AlignConsecutiveShortCaseStatements;
891  const bool AlignArrowOrColon =
892  IsExpr ? Option.AlignCaseArrows : Option.AlignCaseColons;
893 
894  auto Matches = [&](const Change &C) {
895  if (AlignArrowOrColon)
896  return C.Tok->is(Type);
897 
898  // Ignore 'IsInsideToken' to allow matching trailing comments which
899  // need to be reflowed as that causes the token to appear in two
900  // different changes, which will cause incorrect alignment as we'll
901  // reflow early due to detecting multiple aligning tokens per line.
902  return !C.IsInsideToken && C.Tok->Previous && C.Tok->Previous->is(Type);
903  };
904 
905  unsigned MinColumn = 0;
906 
907  // Empty case statements don't break the alignment, but don't necessarily
908  // match our predicate, so we need to track their column so they can push out
909  // our alignment.
910  unsigned MinEmptyCaseColumn = 0;
911 
912  // Start and end of the token sequence we're processing.
913  unsigned StartOfSequence = 0;
914  unsigned EndOfSequence = 0;
915 
916  // Whether a matching token has been found on the current line.
917  bool FoundMatchOnLine = false;
918 
919  bool LineIsComment = true;
920  bool LineIsEmptyCase = false;
921 
922  unsigned I = 0;
923  for (unsigned E = Changes.size(); I != E; ++I) {
924  if (Changes[I].NewlinesBefore != 0) {
925  // Whether to break the alignment sequence because of an empty line.
926  bool EmptyLineBreak =
927  (Changes[I].NewlinesBefore > 1) &&
928  !Style.AlignConsecutiveShortCaseStatements.AcrossEmptyLines;
929 
930  // Whether to break the alignment sequence because of a line without a
931  // match.
932  bool NoMatchBreak =
933  !FoundMatchOnLine &&
934  !(LineIsComment &&
935  Style.AlignConsecutiveShortCaseStatements.AcrossComments) &&
936  !LineIsEmptyCase;
937 
938  if (EmptyLineBreak || NoMatchBreak) {
939  AlignMatchingTokenSequence(StartOfSequence, EndOfSequence, MinColumn,
940  Matches, Changes);
941  MinEmptyCaseColumn = 0;
942  }
943 
944  // A new line starts, re-initialize line status tracking bools.
945  FoundMatchOnLine = false;
946  LineIsComment = true;
947  LineIsEmptyCase = false;
948  }
949 
950  if (Changes[I].Tok->isNot(tok::comment))
951  LineIsComment = false;
952 
953  if (Changes[I].Tok->is(Type)) {
954  LineIsEmptyCase =
955  !Changes[I].Tok->Next || Changes[I].Tok->Next->isTrailingComment();
956 
957  if (LineIsEmptyCase) {
958  if (Style.AlignConsecutiveShortCaseStatements.AlignCaseColons) {
959  MinEmptyCaseColumn =
960  std::max(MinEmptyCaseColumn, Changes[I].StartOfTokenColumn);
961  } else {
962  MinEmptyCaseColumn =
963  std::max(MinEmptyCaseColumn, Changes[I].StartOfTokenColumn + 2);
964  }
965  }
966  }
967 
968  if (!Matches(Changes[I]))
969  continue;
970 
971  if (LineIsEmptyCase)
972  continue;
973 
974  FoundMatchOnLine = true;
975 
976  if (StartOfSequence == 0)
977  StartOfSequence = I;
978 
979  EndOfSequence = I + 1;
980 
981  MinColumn = std::max(MinColumn, Changes[I].StartOfTokenColumn);
982 
983  // Allow empty case statements to push out our alignment.
984  MinColumn = std::max(MinColumn, MinEmptyCaseColumn);
985  }
986 
987  AlignMatchingTokenSequence(StartOfSequence, EndOfSequence, MinColumn, Matches,
988  Changes);
989 }
990 
991 void WhitespaceManager::alignConsecutiveTableGenBreakingDAGArgColons() {
992  alignConsecutiveColons(Style.AlignConsecutiveTableGenBreakingDAGArgColons,
993  TT_TableGenDAGArgListColonToAlign);
994 }
995 
996 void WhitespaceManager::alignConsecutiveTableGenCondOperatorColons() {
997  alignConsecutiveColons(Style.AlignConsecutiveTableGenCondOperatorColons,
998  TT_TableGenCondOperatorColon);
999 }
1000 
1001 void WhitespaceManager::alignConsecutiveTableGenDefinitions() {
1002  alignConsecutiveColons(Style.AlignConsecutiveTableGenDefinitionColons,
1003  TT_InheritanceColon);
1004 }
1005 
1006 void WhitespaceManager::alignConsecutiveDeclarations() {
1007  if (!Style.AlignConsecutiveDeclarations.Enabled)
1008  return;
1009 
1010  AlignTokens(
1011  Style,
1012  [&](Change const &C) {
1013  if (Style.AlignConsecutiveDeclarations.AlignFunctionPointers) {
1014  for (const auto *Prev = C.Tok->Previous; Prev; Prev = Prev->Previous)
1015  if (Prev->is(tok::equal))
1016  return false;
1017  if (C.Tok->is(TT_FunctionTypeLParen))
1018  return true;
1019  }
1020  if (C.Tok->is(TT_FunctionDeclarationName))
1021  return true;
1022  if (C.Tok->isNot(TT_StartOfName))
1023  return false;
1024  if (C.Tok->Previous &&
1025  C.Tok->Previous->is(TT_StatementAttributeLikeMacro))
1026  return false;
1027  // Check if there is a subsequent name that starts the same declaration.
1028  for (FormatToken *Next = C.Tok->Next; Next; Next = Next->Next) {
1029  if (Next->is(tok::comment))
1030  continue;
1031  if (Next->is(TT_PointerOrReference))
1032  return false;
1033  if (!Next->Tok.getIdentifierInfo())
1034  break;
1035  if (Next->isOneOf(TT_StartOfName, TT_FunctionDeclarationName,
1036  tok::kw_operator)) {
1037  return false;
1038  }
1039  }
1040  return true;
1041  },
1042  Changes, /*StartAt=*/0, Style.AlignConsecutiveDeclarations);
1043 }
1044 
1045 void WhitespaceManager::alignChainedConditionals() {
1046  if (Style.BreakBeforeTernaryOperators) {
1047  AlignTokens(
1048  Style,
1049  [](Change const &C) {
1050  // Align question operators and last colon
1051  return C.Tok->is(TT_ConditionalExpr) &&
1052  ((C.Tok->is(tok::question) && !C.NewlinesBefore) ||
1053  (C.Tok->is(tok::colon) && C.Tok->Next &&
1054  (C.Tok->Next->FakeLParens.size() == 0 ||
1055  C.Tok->Next->FakeLParens.back() != prec::Conditional)));
1056  },
1057  Changes, /*StartAt=*/0);
1058  } else {
1059  static auto AlignWrappedOperand = [](Change const &C) {
1060  FormatToken *Previous = C.Tok->getPreviousNonComment();
1061  return C.NewlinesBefore && Previous && Previous->is(TT_ConditionalExpr) &&
1062  (Previous->is(tok::colon) &&
1063  (C.Tok->FakeLParens.size() == 0 ||
1064  C.Tok->FakeLParens.back() != prec::Conditional));
1065  };
1066  // Ensure we keep alignment of wrapped operands with non-wrapped operands
1067  // Since we actually align the operators, the wrapped operands need the
1068  // extra offset to be properly aligned.
1069  for (Change &C : Changes)
1070  if (AlignWrappedOperand(C))
1071  C.StartOfTokenColumn -= 2;
1072  AlignTokens(
1073  Style,
1074  [this](Change const &C) {
1075  // Align question operators if next operand is not wrapped, as
1076  // well as wrapped operands after question operator or last
1077  // colon in conditional sequence
1078  return (C.Tok->is(TT_ConditionalExpr) && C.Tok->is(tok::question) &&
1079  &C != &Changes.back() && (&C + 1)->NewlinesBefore == 0 &&
1080  !(&C + 1)->IsTrailingComment) ||
1081  AlignWrappedOperand(C);
1082  },
1083  Changes, /*StartAt=*/0);
1084  }
1085 }
1086 
1087 void WhitespaceManager::alignTrailingComments() {
1088  if (Style.AlignTrailingComments.Kind == FormatStyle::TCAS_Never)
1089  return;
1090 
1091  const int Size = Changes.size();
1092  int MinColumn = 0;
1093  int StartOfSequence = 0;
1094  bool BreakBeforeNext = false;
1095  int NewLineThreshold = 1;
1096  if (Style.AlignTrailingComments.Kind == FormatStyle::TCAS_Always)
1097  NewLineThreshold = Style.AlignTrailingComments.OverEmptyLines + 1;
1098 
1099  for (int I = 0, MaxColumn = INT_MAX, Newlines = 0; I < Size; ++I) {
1100  auto &C = Changes[I];
1101  if (C.StartOfBlockComment)
1102  continue;
1103  Newlines += C.NewlinesBefore;
1104  if (!C.IsTrailingComment)
1105  continue;
1106 
1107  if (Style.AlignTrailingComments.Kind == FormatStyle::TCAS_Leave) {
1108  const int OriginalSpaces =
1109  C.OriginalWhitespaceRange.getEnd().getRawEncoding() -
1110  C.OriginalWhitespaceRange.getBegin().getRawEncoding() -
1111  C.Tok->LastNewlineOffset;
1112  assert(OriginalSpaces >= 0);
1113  const auto RestoredLineLength =
1114  C.StartOfTokenColumn + C.TokenLength + OriginalSpaces;
1115  // If leaving comments makes the line exceed the column limit, give up to
1116  // leave the comments.
1117  if (RestoredLineLength >= Style.ColumnLimit && Style.ColumnLimit > 0)
1118  break;
1119  C.Spaces = OriginalSpaces;
1120  continue;
1121  }
1122 
1123  const int ChangeMinColumn = C.StartOfTokenColumn;
1124  int ChangeMaxColumn;
1125 
1126  // If we don't create a replacement for this change, we have to consider
1127  // it to be immovable.
1128  if (!C.CreateReplacement)
1129  ChangeMaxColumn = ChangeMinColumn;
1130  else if (Style.ColumnLimit == 0)
1131  ChangeMaxColumn = INT_MAX;
1132  else if (Style.ColumnLimit >= C.TokenLength)
1133  ChangeMaxColumn = Style.ColumnLimit - C.TokenLength;
1134  else
1135  ChangeMaxColumn = ChangeMinColumn;
1136 
1137  if (I + 1 < Size && Changes[I + 1].ContinuesPPDirective &&
1138  ChangeMaxColumn >= 2) {
1139  ChangeMaxColumn -= 2;
1140  }
1141 
1142  bool WasAlignedWithStartOfNextLine = false;
1143  if (C.NewlinesBefore >= 1) { // A comment on its own line.
1144  const auto CommentColumn =
1145  SourceMgr.getSpellingColumnNumber(C.OriginalWhitespaceRange.getEnd());
1146  for (int J = I + 1; J < Size; ++J) {
1147  if (Changes[J].Tok->is(tok::comment))
1148  continue;
1149 
1150  const auto NextColumn = SourceMgr.getSpellingColumnNumber(
1151  Changes[J].OriginalWhitespaceRange.getEnd());
1152  // The start of the next token was previously aligned with the
1153  // start of this comment.
1154  WasAlignedWithStartOfNextLine =
1155  CommentColumn == NextColumn ||
1156  CommentColumn == NextColumn + Style.IndentWidth;
1157  break;
1158  }
1159  }
1160 
1161  // We don't want to align comments which end a scope, which are here
1162  // identified by most closing braces.
1163  auto DontAlignThisComment = [](const auto *Tok) {
1164  if (Tok->is(tok::semi)) {
1165  Tok = Tok->getPreviousNonComment();
1166  if (!Tok)
1167  return false;
1168  }
1169  if (Tok->is(tok::r_paren)) {
1170  // Back up past the parentheses and a `TT_DoWhile` that may precede.
1171  Tok = Tok->MatchingParen;
1172  if (!Tok)
1173  return false;
1174  Tok = Tok->getPreviousNonComment();
1175  if (!Tok)
1176  return false;
1177  if (Tok->is(TT_DoWhile)) {
1178  const auto *Prev = Tok->getPreviousNonComment();
1179  if (!Prev) {
1180  // A do-while-loop without braces.
1181  return true;
1182  }
1183  Tok = Prev;
1184  }
1185  }
1186 
1187  if (Tok->isNot(tok::r_brace))
1188  return false;
1189 
1190  while (Tok->Previous && Tok->Previous->is(tok::r_brace))
1191  Tok = Tok->Previous;
1192  return Tok->NewlinesBefore > 0;
1193  };
1194 
1195  if (I > 0 && C.NewlinesBefore == 0 &&
1196  DontAlignThisComment(Changes[I - 1].Tok)) {
1197  alignTrailingComments(StartOfSequence, I, MinColumn);
1198  // Reset to initial values, but skip this change for the next alignment
1199  // pass.
1200  MinColumn = 0;
1201  MaxColumn = INT_MAX;
1202  StartOfSequence = I + 1;
1203  } else if (BreakBeforeNext || Newlines > NewLineThreshold ||
1204  (ChangeMinColumn > MaxColumn || ChangeMaxColumn < MinColumn) ||
1205  // Break the comment sequence if the previous line did not end
1206  // in a trailing comment.
1207  (C.NewlinesBefore == 1 && I > 0 &&
1208  !Changes[I - 1].IsTrailingComment) ||
1209  WasAlignedWithStartOfNextLine) {
1210  alignTrailingComments(StartOfSequence, I, MinColumn);
1211  MinColumn = ChangeMinColumn;
1212  MaxColumn = ChangeMaxColumn;
1213  StartOfSequence = I;
1214  } else {
1215  MinColumn = std::max(MinColumn, ChangeMinColumn);
1216  MaxColumn = std::min(MaxColumn, ChangeMaxColumn);
1217  }
1218  BreakBeforeNext = (I == 0) || (C.NewlinesBefore > 1) ||
1219  // Never start a sequence with a comment at the beginning
1220  // of the line.
1221  (C.NewlinesBefore == 1 && StartOfSequence == I);
1222  Newlines = 0;
1223  }
1224  alignTrailingComments(StartOfSequence, Size, MinColumn);
1225 }
1226 
1227 void WhitespaceManager::alignTrailingComments(unsigned Start, unsigned End,
1228  unsigned Column) {
1229  for (unsigned i = Start; i != End; ++i) {
1230  int Shift = 0;
1231  if (Changes[i].IsTrailingComment)
1232  Shift = Column - Changes[i].StartOfTokenColumn;
1233  if (Changes[i].StartOfBlockComment) {
1234  Shift = Changes[i].IndentationOffset +
1235  Changes[i].StartOfBlockComment->StartOfTokenColumn -
1236  Changes[i].StartOfTokenColumn;
1237  }
1238  if (Shift <= 0)
1239  continue;
1240  Changes[i].Spaces += Shift;
1241  if (i + 1 != Changes.size())
1242  Changes[i + 1].PreviousEndOfTokenColumn += Shift;
1243  Changes[i].StartOfTokenColumn += Shift;
1244  }
1245 }
1246 
1247 void WhitespaceManager::alignEscapedNewlines() {
1248  const auto Align = Style.AlignEscapedNewlines;
1249  if (Align == FormatStyle::ENAS_DontAlign)
1250  return;
1251 
1252  const bool WithLastLine = Align == FormatStyle::ENAS_LeftWithLastLine;
1253  const bool AlignLeft = Align == FormatStyle::ENAS_Left || WithLastLine;
1254  const auto MaxColumn = Style.ColumnLimit;
1255  unsigned MaxEndOfLine = AlignLeft ? 0 : MaxColumn;
1256  unsigned StartOfMacro = 0;
1257  for (unsigned i = 1, e = Changes.size(); i < e; ++i) {
1258  Change &C = Changes[i];
1259  if (C.NewlinesBefore == 0 && (!WithLastLine || C.Tok->isNot(tok::eof)))
1260  continue;
1261  const bool InPPDirective = C.ContinuesPPDirective;
1262  const auto BackslashColumn = C.PreviousEndOfTokenColumn + 2;
1263  if (InPPDirective ||
1264  (WithLastLine && (MaxColumn == 0 || BackslashColumn <= MaxColumn))) {
1265  MaxEndOfLine = std::max(BackslashColumn, MaxEndOfLine);
1266  }
1267  if (!InPPDirective) {
1268  alignEscapedNewlines(StartOfMacro + 1, i, MaxEndOfLine);
1269  MaxEndOfLine = AlignLeft ? 0 : MaxColumn;
1270  StartOfMacro = i;
1271  }
1272  }
1273  alignEscapedNewlines(StartOfMacro + 1, Changes.size(), MaxEndOfLine);
1274 }
1275 
1276 void WhitespaceManager::alignEscapedNewlines(unsigned Start, unsigned End,
1277  unsigned Column) {
1278  for (unsigned i = Start; i < End; ++i) {
1279  Change &C = Changes[i];
1280  if (C.NewlinesBefore > 0) {
1281  assert(C.ContinuesPPDirective);
1282  if (C.PreviousEndOfTokenColumn + 1 > Column)
1283  C.EscapedNewlineColumn = 0;
1284  else
1285  C.EscapedNewlineColumn = Column;
1286  }
1287  }
1288 }
1289 
1290 void WhitespaceManager::alignArrayInitializers() {
1291  if (Style.AlignArrayOfStructures == FormatStyle::AIAS_None)
1292  return;
1293 
1294  for (unsigned ChangeIndex = 1U, ChangeEnd = Changes.size();
1295  ChangeIndex < ChangeEnd; ++ChangeIndex) {
1296  auto &C = Changes[ChangeIndex];
1297  if (C.Tok->IsArrayInitializer) {
1298  bool FoundComplete = false;
1299  for (unsigned InsideIndex = ChangeIndex + 1; InsideIndex < ChangeEnd;
1300  ++InsideIndex) {
1301  if (Changes[InsideIndex].Tok == C.Tok->MatchingParen) {
1302  alignArrayInitializers(ChangeIndex, InsideIndex + 1);
1303  ChangeIndex = InsideIndex + 1;
1304  FoundComplete = true;
1305  break;
1306  }
1307  }
1308  if (!FoundComplete)
1309  ChangeIndex = ChangeEnd;
1310  }
1311  }
1312 }
1313 
1314 void WhitespaceManager::alignArrayInitializers(unsigned Start, unsigned End) {
1315 
1316  if (Style.AlignArrayOfStructures == FormatStyle::AIAS_Right)
1317  alignArrayInitializersRightJustified(getCells(Start, End));
1318  else if (Style.AlignArrayOfStructures == FormatStyle::AIAS_Left)
1319  alignArrayInitializersLeftJustified(getCells(Start, End));
1320 }
1321 
1322 void WhitespaceManager::alignArrayInitializersRightJustified(
1323  CellDescriptions &&CellDescs) {
1324  if (!CellDescs.isRectangular())
1325  return;
1326 
1327  const int BracePadding = Style.Cpp11BracedListStyle ? 0 : 1;
1328  auto &Cells = CellDescs.Cells;
1329  // Now go through and fixup the spaces.
1330  auto *CellIter = Cells.begin();
1331  for (auto i = 0U; i < CellDescs.CellCounts[0]; ++i, ++CellIter) {
1332  unsigned NetWidth = 0U;
1333  if (isSplitCell(*CellIter))
1334  NetWidth = getNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces);
1335  auto CellWidth = getMaximumCellWidth(CellIter, NetWidth);
1336 
1337  if (Changes[CellIter->Index].Tok->is(tok::r_brace)) {
1338  // So in here we want to see if there is a brace that falls
1339  // on a line that was split. If so on that line we make sure that
1340  // the spaces in front of the brace are enough.
1341  const auto *Next = CellIter;
1342  do {
1343  const FormatToken *Previous = Changes[Next->Index].Tok->Previous;
1344  if (Previous && Previous->isNot(TT_LineComment)) {
1345  Changes[Next->Index].Spaces = BracePadding;
1346  Changes[Next->Index].NewlinesBefore = 0;
1347  }
1348  Next = Next->NextColumnElement;
1349  } while (Next);
1350  // Unless the array is empty, we need the position of all the
1351  // immediately adjacent cells
1352  if (CellIter != Cells.begin()) {
1353  auto ThisNetWidth =
1354  getNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces);
1355  auto MaxNetWidth = getMaximumNetWidth(
1356  Cells.begin(), CellIter, CellDescs.InitialSpaces,
1357  CellDescs.CellCounts[0], CellDescs.CellCounts.size());
1358  if (ThisNetWidth < MaxNetWidth)
1359  Changes[CellIter->Index].Spaces = (MaxNetWidth - ThisNetWidth);
1360  auto RowCount = 1U;
1361  auto Offset = std::distance(Cells.begin(), CellIter);
1362  for (const auto *Next = CellIter->NextColumnElement; Next;
1363  Next = Next->NextColumnElement) {
1364  if (RowCount >= CellDescs.CellCounts.size())
1365  break;
1366  auto *Start = (Cells.begin() + RowCount * CellDescs.CellCounts[0]);
1367  auto *End = Start + Offset;
1368  ThisNetWidth = getNetWidth(Start, End, CellDescs.InitialSpaces);
1369  if (ThisNetWidth < MaxNetWidth)
1370  Changes[Next->Index].Spaces = (MaxNetWidth - ThisNetWidth);
1371  ++RowCount;
1372  }
1373  }
1374  } else {
1375  auto ThisWidth =
1376  calculateCellWidth(CellIter->Index, CellIter->EndIndex, true) +
1377  NetWidth;
1378  if (Changes[CellIter->Index].NewlinesBefore == 0) {
1379  Changes[CellIter->Index].Spaces = (CellWidth - (ThisWidth + NetWidth));
1380  Changes[CellIter->Index].Spaces += (i > 0) ? 1 : BracePadding;
1381  }
1382  alignToStartOfCell(CellIter->Index, CellIter->EndIndex);
1383  for (const auto *Next = CellIter->NextColumnElement; Next;
1384  Next = Next->NextColumnElement) {
1385  ThisWidth =
1386  calculateCellWidth(Next->Index, Next->EndIndex, true) + NetWidth;
1387  if (Changes[Next->Index].NewlinesBefore == 0) {
1388  Changes[Next->Index].Spaces = (CellWidth - ThisWidth);
1389  Changes[Next->Index].Spaces += (i > 0) ? 1 : BracePadding;
1390  }
1391  alignToStartOfCell(Next->Index, Next->EndIndex);
1392  }
1393  }
1394  }
1395 }
1396 
1397 void WhitespaceManager::alignArrayInitializersLeftJustified(
1398  CellDescriptions &&CellDescs) {
1399 
1400  if (!CellDescs.isRectangular())
1401  return;
1402 
1403  const int BracePadding = Style.Cpp11BracedListStyle ? 0 : 1;
1404  auto &Cells = CellDescs.Cells;
1405  // Now go through and fixup the spaces.
1406  auto *CellIter = Cells.begin();
1407  // The first cell of every row needs to be against the left brace.
1408  for (const auto *Next = CellIter; Next; Next = Next->NextColumnElement) {
1409  auto &Change = Changes[Next->Index];
1410  Change.Spaces =
1411  Change.NewlinesBefore == 0 ? BracePadding : CellDescs.InitialSpaces;
1412  }
1413  ++CellIter;
1414  for (auto i = 1U; i < CellDescs.CellCounts[0]; i++, ++CellIter) {
1415  auto MaxNetWidth = getMaximumNetWidth(
1416  Cells.begin(), CellIter, CellDescs.InitialSpaces,
1417  CellDescs.CellCounts[0], CellDescs.CellCounts.size());
1418  auto ThisNetWidth =
1419  getNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces);
1420  if (Changes[CellIter->Index].NewlinesBefore == 0) {
1421  Changes[CellIter->Index].Spaces =
1422  MaxNetWidth - ThisNetWidth +
1423  (Changes[CellIter->Index].Tok->isNot(tok::r_brace) ? 1
1424  : BracePadding);
1425  }
1426  auto RowCount = 1U;
1427  auto Offset = std::distance(Cells.begin(), CellIter);
1428  for (const auto *Next = CellIter->NextColumnElement; Next;
1429  Next = Next->NextColumnElement) {
1430  if (RowCount >= CellDescs.CellCounts.size())
1431  break;
1432  auto *Start = (Cells.begin() + RowCount * CellDescs.CellCounts[0]);
1433  auto *End = Start + Offset;
1434  auto ThisNetWidth = getNetWidth(Start, End, CellDescs.InitialSpaces);
1435  if (Changes[Next->Index].NewlinesBefore == 0) {
1436  Changes[Next->Index].Spaces =
1437  MaxNetWidth - ThisNetWidth +
1438  (Changes[Next->Index].Tok->isNot(tok::r_brace) ? 1 : BracePadding);
1439  }
1440  ++RowCount;
1441  }
1442  }
1443 }
1444 
1445 bool WhitespaceManager::isSplitCell(const CellDescription &Cell) {
1446  if (Cell.HasSplit)
1447  return true;
1448  for (const auto *Next = Cell.NextColumnElement; Next;
1449  Next = Next->NextColumnElement) {
1450  if (Next->HasSplit)
1451  return true;
1452  }
1453  return false;
1454 }
1455 
1456 WhitespaceManager::CellDescriptions WhitespaceManager::getCells(unsigned Start,
1457  unsigned End) {
1458 
1459  unsigned Depth = 0;
1460  unsigned Cell = 0;
1461  SmallVector<unsigned> CellCounts;
1462  unsigned InitialSpaces = 0;
1463  unsigned InitialTokenLength = 0;
1464  unsigned EndSpaces = 0;
1465  SmallVector<CellDescription> Cells;
1466  const FormatToken *MatchingParen = nullptr;
1467  for (unsigned i = Start; i < End; ++i) {
1468  auto &C = Changes[i];
1469  if (C.Tok->is(tok::l_brace))
1470  ++Depth;
1471  else if (C.Tok->is(tok::r_brace))
1472  --Depth;
1473  if (Depth == 2) {
1474  if (C.Tok->is(tok::l_brace)) {
1475  Cell = 0;
1476  MatchingParen = C.Tok->MatchingParen;
1477  if (InitialSpaces == 0) {
1478  InitialSpaces = C.Spaces + C.TokenLength;
1479  InitialTokenLength = C.TokenLength;
1480  auto j = i - 1;
1481  for (; Changes[j].NewlinesBefore == 0 && j > Start; --j) {
1482  InitialSpaces += Changes[j].Spaces + Changes[j].TokenLength;
1483  InitialTokenLength += Changes[j].TokenLength;
1484  }
1485  if (C.NewlinesBefore == 0) {
1486  InitialSpaces += Changes[j].Spaces + Changes[j].TokenLength;
1487  InitialTokenLength += Changes[j].TokenLength;
1488  }
1489  }
1490  } else if (C.Tok->is(tok::comma)) {
1491  if (!Cells.empty())
1492  Cells.back().EndIndex = i;
1493  if (const auto *Next = C.Tok->getNextNonComment();
1494  Next && Next->isNot(tok::r_brace)) { // dangling comma
1495  ++Cell;
1496  }
1497  }
1498  } else if (Depth == 1) {
1499  if (C.Tok == MatchingParen) {
1500  if (!Cells.empty())
1501  Cells.back().EndIndex = i;
1502  Cells.push_back(CellDescription{i, ++Cell, i + 1, false, nullptr});
1503  CellCounts.push_back(C.Tok->Previous->isNot(tok::comma) ? Cell + 1
1504  : Cell);
1505  // Go to the next non-comment and ensure there is a break in front
1506  const auto *NextNonComment = C.Tok->getNextNonComment();
1507  while (NextNonComment && NextNonComment->is(tok::comma))
1508  NextNonComment = NextNonComment->getNextNonComment();
1509  auto j = i;
1510  while (j < End && Changes[j].Tok != NextNonComment)
1511  ++j;
1512  if (j < End && Changes[j].NewlinesBefore == 0 &&
1513  Changes[j].Tok->isNot(tok::r_brace)) {
1514  Changes[j].NewlinesBefore = 1;
1515  // Account for the added token lengths
1516  Changes[j].Spaces = InitialSpaces - InitialTokenLength;
1517  }
1518  } else if (C.Tok->is(tok::comment) && C.Tok->NewlinesBefore == 0) {
1519  // Trailing comments stay at a space past the last token
1520  C.Spaces = Changes[i - 1].Tok->is(tok::comma) ? 1 : 2;
1521  } else if (C.Tok->is(tok::l_brace)) {
1522  // We need to make sure that the ending braces is aligned to the
1523  // start of our initializer
1524  auto j = i - 1;
1525  for (; j > 0 && !Changes[j].Tok->ArrayInitializerLineStart; --j)
1526  ; // Nothing the loop does the work
1527  EndSpaces = Changes[j].Spaces;
1528  }
1529  } else if (Depth == 0 && C.Tok->is(tok::r_brace)) {
1530  C.NewlinesBefore = 1;
1531  C.Spaces = EndSpaces;
1532  }
1533  if (C.Tok->StartsColumn) {
1534  // This gets us past tokens that have been split over multiple
1535  // lines
1536  bool HasSplit = false;
1537  if (Changes[i].NewlinesBefore > 0) {
1538  // So if we split a line previously and the tail line + this token is
1539  // less then the column limit we remove the split here and just put
1540  // the column start at a space past the comma
1541  //
1542  // FIXME This if branch covers the cases where the column is not
1543  // the first column. This leads to weird pathologies like the formatting
1544  // auto foo = Items{
1545  // Section{
1546  // 0, bar(),
1547  // }
1548  // };
1549  // Well if it doesn't lead to that it's indicative that the line
1550  // breaking should be revisited. Unfortunately alot of other options
1551  // interact with this
1552  auto j = i - 1;
1553  if ((j - 1) > Start && Changes[j].Tok->is(tok::comma) &&
1554  Changes[j - 1].NewlinesBefore > 0) {
1555  --j;
1556  auto LineLimit = Changes[j].Spaces + Changes[j].TokenLength;
1557  if (LineLimit < Style.ColumnLimit) {
1558  Changes[i].NewlinesBefore = 0;
1559  Changes[i].Spaces = 1;
1560  }
1561  }
1562  }
1563  while (Changes[i].NewlinesBefore > 0 && Changes[i].Tok == C.Tok) {
1564  Changes[i].Spaces = InitialSpaces;
1565  ++i;
1566  HasSplit = true;
1567  }
1568  if (Changes[i].Tok != C.Tok)
1569  --i;
1570  Cells.push_back(CellDescription{i, Cell, i, HasSplit, nullptr});
1571  }
1572  }
1573 
1574  return linkCells({Cells, CellCounts, InitialSpaces});
1575 }
1576 
1577 unsigned WhitespaceManager::calculateCellWidth(unsigned Start, unsigned End,
1578  bool WithSpaces) const {
1579  unsigned CellWidth = 0;
1580  for (auto i = Start; i < End; i++) {
1581  if (Changes[i].NewlinesBefore > 0)
1582  CellWidth = 0;
1583  CellWidth += Changes[i].TokenLength;
1584  CellWidth += (WithSpaces ? Changes[i].Spaces : 0);
1585  }
1586  return CellWidth;
1587 }
1588 
1589 void WhitespaceManager::alignToStartOfCell(unsigned Start, unsigned End) {
1590  if ((End - Start) <= 1)
1591  return;
1592  // If the line is broken anywhere in there make sure everything
1593  // is aligned to the parent
1594  for (auto i = Start + 1; i < End; i++)
1595  if (Changes[i].NewlinesBefore > 0)
1596  Changes[i].Spaces = Changes[Start].Spaces;
1597 }
1598 
1599 WhitespaceManager::CellDescriptions
1600 WhitespaceManager::linkCells(CellDescriptions &&CellDesc) {
1601  auto &Cells = CellDesc.Cells;
1602  for (auto *CellIter = Cells.begin(); CellIter != Cells.end(); ++CellIter) {
1603  if (!CellIter->NextColumnElement && (CellIter + 1) != Cells.end()) {
1604  for (auto *NextIter = CellIter + 1; NextIter != Cells.end(); ++NextIter) {
1605  if (NextIter->Cell == CellIter->Cell) {
1606  CellIter->NextColumnElement = &(*NextIter);
1607  break;
1608  }
1609  }
1610  }
1611  }
1612  return std::move(CellDesc);
1613 }
1614 
1615 void WhitespaceManager::generateChanges() {
1616  for (unsigned i = 0, e = Changes.size(); i != e; ++i) {
1617  const Change &C = Changes[i];
1618  if (i > 0) {
1619  auto Last = Changes[i - 1].OriginalWhitespaceRange;
1620  auto New = Changes[i].OriginalWhitespaceRange;
1621  // Do not generate two replacements for the same location. As a special
1622  // case, it is allowed if there is a replacement for the empty range
1623  // between 2 tokens and another non-empty range at the start of the second
1624  // token. We didn't implement logic to combine replacements for 2
1625  // consecutive source ranges into a single replacement, because the
1626  // program works fine without it.
1627  //
1628  // We can't eliminate empty original whitespace ranges. They appear when
1629  // 2 tokens have no whitespace in between in the input. It does not
1630  // matter whether whitespace is to be added. If no whitespace is to be
1631  // added, the replacement will be empty, and it gets eliminated after this
1632  // step in storeReplacement. For example, if the input is `foo();`,
1633  // there will be a replacement for the range between every consecutive
1634  // pair of tokens.
1635  //
1636  // A replacement at the start of a token can be added by
1637  // BreakableStringLiteralUsingOperators::insertBreak when it adds braces
1638  // around the string literal. Say Verilog code is being formatted and the
1639  // first line is to become the next 2 lines.
1640  // x("long string");
1641  // x({"long ",
1642  // "string"});
1643  // There will be a replacement for the empty range between the parenthesis
1644  // and the string and another replacement for the quote character. The
1645  // replacement for the empty range between the parenthesis and the quote
1646  // comes from ContinuationIndenter::addTokenOnCurrentLine when it changes
1647  // the original empty range between the parenthesis and the string to
1648  // another empty one. The replacement for the quote character comes from
1649  // BreakableStringLiteralUsingOperators::insertBreak when it adds the
1650  // brace. In the example, the replacement for the empty range is the same
1651  // as the original text. However, eliminating replacements that are same
1652  // as the original does not help in general. For example, a newline can
1653  // be inserted, causing the first line to become the next 3 lines.
1654  // xxxxxxxxxxx("long string");
1655  // xxxxxxxxxxx(
1656  // {"long ",
1657  // "string"});
1658  // In that case, the empty range between the parenthesis and the string
1659  // will be replaced by a newline and 4 spaces. So we will still have to
1660  // deal with a replacement for an empty source range followed by a
1661  // replacement for a non-empty source range.
1662  if (Last.getBegin() == New.getBegin() &&
1663  (Last.getEnd() != Last.getBegin() ||
1664  New.getEnd() == New.getBegin())) {
1665  continue;
1666  }
1667  }
1668  if (C.CreateReplacement) {
1669  std::string ReplacementText = C.PreviousLinePostfix;
1670  if (C.ContinuesPPDirective) {
1671  appendEscapedNewlineText(ReplacementText, C.NewlinesBefore,
1672  C.PreviousEndOfTokenColumn,
1673  C.EscapedNewlineColumn);
1674  } else {
1675  appendNewlineText(ReplacementText, C.NewlinesBefore);
1676  }
1677  // FIXME: This assert should hold if we computed the column correctly.
1678  // assert((int)C.StartOfTokenColumn >= C.Spaces);
1679  appendIndentText(
1680  ReplacementText, C.Tok->IndentLevel, std::max(0, C.Spaces),
1681  std::max((int)C.StartOfTokenColumn, C.Spaces) - std::max(0, C.Spaces),
1682  C.IsAligned);
1683  ReplacementText.append(C.CurrentLinePrefix);
1684  storeReplacement(C.OriginalWhitespaceRange, ReplacementText);
1685  }
1686  }
1687 }
1688 
1689 void WhitespaceManager::storeReplacement(SourceRange Range, StringRef Text) {
1690  unsigned WhitespaceLength = SourceMgr.getFileOffset(Range.getEnd()) -
1691  SourceMgr.getFileOffset(Range.getBegin());
1692  // Don't create a replacement, if it does not change anything.
1693  if (StringRef(SourceMgr.getCharacterData(Range.getBegin()),
1694  WhitespaceLength) == Text) {
1695  return;
1696  }
1697  auto Err = Replaces.add(tooling::Replacement(
1698  SourceMgr, CharSourceRange::getCharRange(Range), Text));
1699  // FIXME: better error handling. For now, just print an error message in the
1700  // release version.
1701  if (Err) {
1702  llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1703  assert(false);
1704  }
1705 }
1706 
1707 void WhitespaceManager::appendNewlineText(std::string &Text,
1708  unsigned Newlines) {
1709  if (UseCRLF) {
1710  Text.reserve(Text.size() + 2 * Newlines);
1711  for (unsigned i = 0; i < Newlines; ++i)
1712  Text.append("\r\n");
1713  } else {
1714  Text.append(Newlines, '\n');
1715  }
1716 }
1717 
1718 void WhitespaceManager::appendEscapedNewlineText(
1719  std::string &Text, unsigned Newlines, unsigned PreviousEndOfTokenColumn,
1720  unsigned EscapedNewlineColumn) {
1721  if (Newlines > 0) {
1722  unsigned Spaces =
1723  std::max<int>(1, EscapedNewlineColumn - PreviousEndOfTokenColumn - 1);
1724  for (unsigned i = 0; i < Newlines; ++i) {
1725  Text.append(Spaces, ' ');
1726  Text.append(UseCRLF ? "\\\r\n" : "\\\n");
1727  Spaces = std::max<int>(0, EscapedNewlineColumn - 1);
1728  }
1729  }
1730 }
1731 
1732 void WhitespaceManager::appendIndentText(std::string &Text,
1733  unsigned IndentLevel, unsigned Spaces,
1734  unsigned WhitespaceStartColumn,
1735  bool IsAligned) {
1736  switch (Style.UseTab) {
1737  case FormatStyle::UT_Never:
1738  Text.append(Spaces, ' ');
1739  break;
1740  case FormatStyle::UT_Always: {
1741  if (Style.TabWidth) {
1742  unsigned FirstTabWidth =
1743  Style.TabWidth - WhitespaceStartColumn % Style.TabWidth;
1744 
1745  // Insert only spaces when we want to end up before the next tab.
1746  if (Spaces < FirstTabWidth || Spaces == 1) {
1747  Text.append(Spaces, ' ');
1748  break;
1749  }
1750  // Align to the next tab.
1751  Spaces -= FirstTabWidth;
1752  Text.append("\t");
1753 
1754  Text.append(Spaces / Style.TabWidth, '\t');
1755  Text.append(Spaces % Style.TabWidth, ' ');
1756  } else if (Spaces == 1) {
1757  Text.append(Spaces, ' ');
1758  }
1759  break;
1760  }
1762  if (WhitespaceStartColumn == 0) {
1763  unsigned Indentation = IndentLevel * Style.IndentWidth;
1764  Spaces = appendTabIndent(Text, Spaces, Indentation);
1765  }
1766  Text.append(Spaces, ' ');
1767  break;
1769  if (WhitespaceStartColumn == 0)
1770  Spaces = appendTabIndent(Text, Spaces, Spaces);
1771  Text.append(Spaces, ' ');
1772  break;
1774  if (WhitespaceStartColumn == 0) {
1775  unsigned Indentation =
1776  IsAligned ? IndentLevel * Style.IndentWidth : Spaces;
1777  Spaces = appendTabIndent(Text, Spaces, Indentation);
1778  }
1779  Text.append(Spaces, ' ');
1780  break;
1781  }
1782 }
1783 
1784 unsigned WhitespaceManager::appendTabIndent(std::string &Text, unsigned Spaces,
1785  unsigned Indentation) {
1786  // This happens, e.g. when a line in a block comment is indented less than the
1787  // first one.
1788  if (Indentation > Spaces)
1789  Indentation = Spaces;
1790  if (Style.TabWidth) {
1791  unsigned Tabs = Indentation / Style.TabWidth;
1792  Text.append(Tabs, '\t');
1793  Spaces -= Tabs * Style.TabWidth;
1794  }
1795  return Spaces;
1796 }
1797 
1798 } // namespace format
1799 } // namespace clang
int Depth
Definition: ASTDiff.cpp:190
MatchType Type
StringRef P
StringRef Text
Definition: Format.cpp:2977
unsigned Offset
Definition: Format.cpp:2978
SourceRange Range
Definition: SemaObjC.cpp:754
SourceLocation End
StateNode * Previous
WhitespaceManager class manages whitespace around tokens and their replacements.
__DEVICE__ int min(int __a, int __b)
__DEVICE__ int max(int __a, int __b)
static CharSourceRange getCharRange(SourceRange R)
Encodes a location in the source.
SourceLocation getLocWithOffset(IntTy Offset) const
Return a source location with the specified offset from this SourceLocation.
unsigned getFileOffset(SourceLocation SpellingLoc) const
Returns the offset from the start of the file that the specified SourceLocation represents.
const char * getCharacterData(SourceLocation SL, bool *Invalid=nullptr) const
Return a pointer to the start of the specified location in the appropriate spelling MemoryBuffer.
unsigned getSpellingColumnNumber(SourceLocation Loc, bool *Invalid=nullptr) const
bool isBeforeInTranslationUnit(SourceLocation LHS, SourceLocation RHS) const
Determines the order of 2 source locations in the translation unit.
A trivial tuple used to represent a source range.
SourceLocation getEnd() const
SourceLocation getBegin() const
Functor to sort changes in original source order.
bool operator()(const Change &C1, const Change &C2) const
void replaceWhitespaceInToken(const FormatToken &Tok, unsigned Offset, unsigned ReplaceChars, StringRef PreviousPostfix, StringRef CurrentPrefix, bool InPPDirective, unsigned Newlines, int Spaces)
Inserts or replaces whitespace in the middle of a token.
void addUntouchableToken(const FormatToken &Tok, bool InPPDirective)
Adds information about an unchangeable token's whitespace.
static bool inputUsesCRLF(StringRef Text, bool DefaultToCRLF)
Infers whether the input is using CRLF.
llvm::Error addReplacement(const tooling::Replacement &Replacement)
const tooling::Replacements & generateReplacements()
Returns all the Replacements created during formatting.
void replaceWhitespace(FormatToken &Tok, unsigned Newlines, unsigned Spaces, unsigned StartOfTokenColumn, bool IsAligned=false, bool InPPDirective=false)
Replaces the whitespace in front of Tok.
A text replacement.
Definition: Replacement.h:83
Maintains a set of replacements that are conflict-free.
Definition: Replacement.h:212
llvm::Error add(const Replacement &R)
Adds a new replacement R to the current set of replacements.
#define INT_MAX
Definition: limits.h:50
@ MR_ExpandedArg
The token was expanded from a macro argument when formatting the expanded token sequence.
Definition: FormatToken.h:227
static void AlignMatchingTokenSequence(unsigned &StartOfSequence, unsigned &EndOfSequence, unsigned &MinColumn, std::function< bool(const WhitespaceManager::Change &C)> Matches, SmallVector< WhitespaceManager::Change, 16 > &Changes)
static unsigned AlignTokens(const FormatStyle &Style, F &&Matches, SmallVector< WhitespaceManager::Change, 16 > &Changes, unsigned StartAt, const FormatStyle::AlignConsecutiveStyle &ACS={}, bool RightJustify=false)
static void AlignTokenSequence(const FormatStyle &Style, unsigned Start, unsigned End, unsigned Column, bool RightJustify, F &&Matches, SmallVector< WhitespaceManager::Change, 16 > &Changes)
TokenType
Determines the semantic type of a syntactic token, e.g.
Definition: FormatToken.h:205
std::string toString(const til::SExpr *E)
The JSON file list parser is used to communicate input to InstallAPI.
if(T->getSizeExpr()) TRY_TO(TraverseStmt(const_cast< Expr * >(T -> getSizeExpr())))
for(const auto &A :T->param_types())
float __ovld __cnfn distance(float, float)
Returns the distance between p0 and p1.
#define false
Definition: stdbool.h:26
The FormatStyle is used to configure the formatting to follow specific guidelines.
Definition: Format.h:55
@ UT_AlignWithSpaces
Use tabs for line continuation and indentation, and spaces for alignment.
Definition: Format.h:4898
@ UT_ForContinuationAndIndentation
Fill all leading whitespace with tabs, and use spaces for alignment that appears within a line (e....
Definition: Format.h:4895
@ UT_ForIndentation
Use tabs only for indentation.
Definition: Format.h:4892
@ UT_Always
Use tabs whenever we need to fill whitespace that spans at least from one tab stop to the next one.
Definition: Format.h:4901
@ UT_Never
Never use tab.
Definition: Format.h:4890
bool isTableGen() const
Definition: Format.h:3188
@ TCAS_Never
Don't align trailing comments but other formatter applies.
Definition: Format.h:579
@ TCAS_Leave
Leave trailing comments as they are.
Definition: Format.h:561
@ TCAS_Always
Align trailing comments.
Definition: Format.h:570
@ AIAS_Left
Align array column and left justify the columns e.g.
Definition: Format.h:120
@ AIAS_Right
Align array column and right justify the columns e.g.
Definition: Format.h:130
@ AIAS_None
Don't align array initializer columns.
Definition: Format.h:132
@ ENAS_DontAlign
Don't align escaped newlines.
Definition: Format.h:480
@ ENAS_Left
Align escaped newlines as far left as possible.
Definition: Format.h:488
@ ENAS_LeftWithLastLine
Align escaped newlines as far left as possible, using the last line of the preprocessor directive as ...
Definition: Format.h:497
@ PAS_Right
Align pointer to the right.
Definition: Format.h:3567
@ RAS_Right
Align reference to the right.
Definition: Format.h:3739
@ RAS_Pointer
Align reference like PointerAlignment.
Definition: Format.h:3729
A wrapper around a Token storing information about the whitespace characters preceding it.
Definition: FormatToken.h:290
unsigned OriginalColumn
The original 0-based column of this token, including expanded tabs.
Definition: FormatToken.h:500
SourceLocation getStartOfNonWhitespace() const
Returns actual token start location without leading escaped newlines and whitespace.
Definition: FormatToken.h:785
void setDecision(FormatDecision D)
Definition: FormatToken.h:398
unsigned Finalized
If true, this token has been fully formatted (indented and potentially re-formatted inside),...
Definition: FormatToken.h:369
std::optional< MacroExpansion > MacroCtx
Definition: FormatToken.h:591
unsigned NewlinesBefore
The number of newlines immediately before the Token.
Definition: FormatToken.h:459
unsigned IsFirst
Indicates that this is the first token of the file.
Definition: FormatToken.h:329
SourceRange WhitespaceRange
The range of the whitespace immediately preceding the Token.
Definition: FormatToken.h:319
Represents a change before a token, a break inside a token, or the layout of an unchanged token (or w...
Change(const FormatToken &Tok, bool CreateReplacement, SourceRange OriginalWhitespaceRange, int Spaces, unsigned StartOfTokenColumn, unsigned NewlinesBefore, StringRef PreviousLinePostfix, StringRef CurrentLinePrefix, bool IsAligned, bool ContinuesPPDirective, bool IsInsideToken)
Creates a Change.