clang  7.0.0svn
BreakableToken.cpp
Go to the documentation of this file.
1 //===--- BreakableToken.cpp - Format C++ code -----------------------------===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 ///
10 /// \file
11 /// Contains implementation of BreakableToken class and classes derived
12 /// from it.
13 ///
14 //===----------------------------------------------------------------------===//
15 
16 #include "BreakableToken.h"
17 #include "ContinuationIndenter.h"
18 #include "clang/Basic/CharInfo.h"
19 #include "clang/Format/Format.h"
20 #include "llvm/ADT/STLExtras.h"
21 #include "llvm/Support/Debug.h"
22 #include <algorithm>
23 
24 #define DEBUG_TYPE "format-token-breaker"
25 
26 namespace clang {
27 namespace format {
28 
29 static const char *const Blanks = " \t\v\f\r";
30 static bool IsBlank(char C) {
31  switch (C) {
32  case ' ':
33  case '\t':
34  case '\v':
35  case '\f':
36  case '\r':
37  return true;
38  default:
39  return false;
40  }
41 }
42 
43 static StringRef getLineCommentIndentPrefix(StringRef Comment,
44  const FormatStyle &Style) {
45  static const char *const KnownCStylePrefixes[] = {"///<", "//!<", "///", "//",
46  "//!"};
47  static const char *const KnownTextProtoPrefixes[] = {"//", "#", "##", "###",
48  "####"};
49  ArrayRef<const char *> KnownPrefixes(KnownCStylePrefixes);
51  KnownPrefixes = KnownTextProtoPrefixes;
52 
53  StringRef LongestPrefix;
54  for (StringRef KnownPrefix : KnownPrefixes) {
55  if (Comment.startswith(KnownPrefix)) {
56  size_t PrefixLength = KnownPrefix.size();
57  while (PrefixLength < Comment.size() && Comment[PrefixLength] == ' ')
58  ++PrefixLength;
59  if (PrefixLength > LongestPrefix.size())
60  LongestPrefix = Comment.substr(0, PrefixLength);
61  }
62  }
63  return LongestPrefix;
64 }
65 
66 static BreakableToken::Split getCommentSplit(StringRef Text,
67  unsigned ContentStartColumn,
68  unsigned ColumnLimit,
69  unsigned TabWidth,
71  LLVM_DEBUG(llvm::dbgs() << "Comment split: \"" << Text << ", " << ColumnLimit
72  << "\", Content start: " << ContentStartColumn
73  << "\n");
74  if (ColumnLimit <= ContentStartColumn + 1)
75  return BreakableToken::Split(StringRef::npos, 0);
76 
77  unsigned MaxSplit = ColumnLimit - ContentStartColumn + 1;
78  unsigned MaxSplitBytes = 0;
79 
80  for (unsigned NumChars = 0;
81  NumChars < MaxSplit && MaxSplitBytes < Text.size();) {
82  unsigned BytesInChar =
83  encoding::getCodePointNumBytes(Text[MaxSplitBytes], Encoding);
84  NumChars +=
85  encoding::columnWidthWithTabs(Text.substr(MaxSplitBytes, BytesInChar),
86  ContentStartColumn, TabWidth, Encoding);
87  MaxSplitBytes += BytesInChar;
88  }
89 
90  StringRef::size_type SpaceOffset = Text.find_last_of(Blanks, MaxSplitBytes);
91 
92  // Do not split before a number followed by a dot: this would be interpreted
93  // as a numbered list, which would prevent re-flowing in subsequent passes.
94  static auto *const kNumberedListRegexp = new llvm::Regex("^[1-9][0-9]?\\.");
95  if (SpaceOffset != StringRef::npos &&
96  kNumberedListRegexp->match(Text.substr(SpaceOffset).ltrim(Blanks)))
97  SpaceOffset = Text.find_last_of(Blanks, SpaceOffset);
98 
99  if (SpaceOffset == StringRef::npos ||
100  // Don't break at leading whitespace.
101  Text.find_last_not_of(Blanks, SpaceOffset) == StringRef::npos) {
102  // Make sure that we don't break at leading whitespace that
103  // reaches past MaxSplit.
104  StringRef::size_type FirstNonWhitespace = Text.find_first_not_of(Blanks);
105  if (FirstNonWhitespace == StringRef::npos)
106  // If the comment is only whitespace, we cannot split.
107  return BreakableToken::Split(StringRef::npos, 0);
108  SpaceOffset = Text.find_first_of(
109  Blanks, std::max<unsigned>(MaxSplitBytes, FirstNonWhitespace));
110  }
111  if (SpaceOffset != StringRef::npos && SpaceOffset != 0) {
112  StringRef BeforeCut = Text.substr(0, SpaceOffset).rtrim(Blanks);
113  StringRef AfterCut = Text.substr(SpaceOffset).ltrim(Blanks);
114  return BreakableToken::Split(BeforeCut.size(),
115  AfterCut.begin() - BeforeCut.end());
116  }
117  return BreakableToken::Split(StringRef::npos, 0);
118 }
119 
120 static BreakableToken::Split
121 getStringSplit(StringRef Text, unsigned UsedColumns, unsigned ColumnLimit,
122  unsigned TabWidth, encoding::Encoding Encoding) {
123  // FIXME: Reduce unit test case.
124  if (Text.empty())
125  return BreakableToken::Split(StringRef::npos, 0);
126  if (ColumnLimit <= UsedColumns)
127  return BreakableToken::Split(StringRef::npos, 0);
128  unsigned MaxSplit = ColumnLimit - UsedColumns;
129  StringRef::size_type SpaceOffset = 0;
130  StringRef::size_type SlashOffset = 0;
131  StringRef::size_type WordStartOffset = 0;
132  StringRef::size_type SplitPoint = 0;
133  for (unsigned Chars = 0;;) {
134  unsigned Advance;
135  if (Text[0] == '\\') {
136  Advance = encoding::getEscapeSequenceLength(Text);
137  Chars += Advance;
138  } else {
139  Advance = encoding::getCodePointNumBytes(Text[0], Encoding);
141  Text.substr(0, Advance), UsedColumns + Chars, TabWidth, Encoding);
142  }
143 
144  if (Chars > MaxSplit || Text.size() <= Advance)
145  break;
146 
147  if (IsBlank(Text[0]))
148  SpaceOffset = SplitPoint;
149  if (Text[0] == '/')
150  SlashOffset = SplitPoint;
151  if (Advance == 1 && !isAlphanumeric(Text[0]))
152  WordStartOffset = SplitPoint;
153 
154  SplitPoint += Advance;
155  Text = Text.substr(Advance);
156  }
157 
158  if (SpaceOffset != 0)
159  return BreakableToken::Split(SpaceOffset + 1, 0);
160  if (SlashOffset != 0)
161  return BreakableToken::Split(SlashOffset + 1, 0);
162  if (WordStartOffset != 0)
163  return BreakableToken::Split(WordStartOffset + 1, 0);
164  if (SplitPoint != 0)
165  return BreakableToken::Split(SplitPoint, 0);
166  return BreakableToken::Split(StringRef::npos, 0);
167 }
168 
170  assert((Token.is(TT_BlockComment) || Token.is(TT_LineComment)) &&
171  "formatting regions are switched by comment tokens");
172  StringRef Content = Token.TokenText.substr(2).ltrim();
173  return Content.startswith("clang-format on") ||
174  Content.startswith("clang-format off");
175 }
176 
177 unsigned
178 BreakableToken::getLengthAfterCompression(unsigned RemainingTokenColumns,
179  Split Split) const {
180  // Example: consider the content
181  // lala lala
182  // - RemainingTokenColumns is the original number of columns, 10;
183  // - Split is (4, 2), denoting the two spaces between the two words;
184  //
185  // We compute the number of columns when the split is compressed into a single
186  // space, like:
187  // lala lala
188  //
189  // FIXME: Correctly measure the length of whitespace in Split.second so it
190  // works with tabs.
191  return RemainingTokenColumns + 1 - Split.second;
192 }
193 
194 unsigned BreakableStringLiteral::getLineCount() const { return 1; }
195 
196 unsigned BreakableStringLiteral::getRangeLength(unsigned LineIndex,
197  unsigned Offset,
198  StringRef::size_type Length,
199  unsigned StartColumn) const {
200  llvm_unreachable("Getting the length of a part of the string literal "
201  "indicates that the code tries to reflow it.");
202 }
203 
204 unsigned
206  unsigned StartColumn) const {
207  return UnbreakableTailLength + Postfix.size() +
208  encoding::columnWidthWithTabs(Line.substr(Offset, StringRef::npos),
210 }
211 
213  bool Break) const {
214  return StartColumn + Prefix.size();
215 }
216 
218  const FormatToken &Tok, unsigned StartColumn, StringRef Prefix,
219  StringRef Postfix, unsigned UnbreakableTailLength, bool InPPDirective,
221  : BreakableToken(Tok, InPPDirective, Encoding, Style),
222  StartColumn(StartColumn), Prefix(Prefix), Postfix(Postfix),
223  UnbreakableTailLength(UnbreakableTailLength) {
224  assert(Tok.TokenText.startswith(Prefix) && Tok.TokenText.endswith(Postfix));
225  Line = Tok.TokenText.substr(
226  Prefix.size(), Tok.TokenText.size() - Prefix.size() - Postfix.size());
227 }
228 
229 BreakableToken::Split BreakableStringLiteral::getSplit(
230  unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit,
231  unsigned ContentStartColumn, llvm::Regex &CommentPragmasRegex) const {
232  return getStringSplit(Line.substr(TailOffset), ContentStartColumn,
233  ColumnLimit - Postfix.size(), Style.TabWidth, Encoding);
234 }
235 
236 void BreakableStringLiteral::insertBreak(unsigned LineIndex,
237  unsigned TailOffset, Split Split,
238  WhitespaceManager &Whitespaces) const {
239  Whitespaces.replaceWhitespaceInToken(
240  Tok, Prefix.size() + TailOffset + Split.first, Split.second, Postfix,
242 }
243 
245  unsigned StartColumn, bool InPPDirective,
247  const FormatStyle &Style)
248  : BreakableToken(Token, InPPDirective, Encoding, Style),
249  StartColumn(StartColumn) {}
250 
251 unsigned BreakableComment::getLineCount() const { return Lines.size(); }
252 
253 BreakableToken::Split
254 BreakableComment::getSplit(unsigned LineIndex, unsigned TailOffset,
255  unsigned ColumnLimit, unsigned ContentStartColumn,
256  llvm::Regex &CommentPragmasRegex) const {
257  // Don't break lines matching the comment pragmas regex.
258  if (CommentPragmasRegex.match(Content[LineIndex]))
259  return Split(StringRef::npos, 0);
260  return getCommentSplit(Content[LineIndex].substr(TailOffset),
261  ContentStartColumn, ColumnLimit, Style.TabWidth,
262  Encoding);
263 }
264 
266  unsigned LineIndex, unsigned TailOffset, Split Split,
267  WhitespaceManager &Whitespaces) const {
268  StringRef Text = Content[LineIndex].substr(TailOffset);
269  // Text is relative to the content line, but Whitespaces operates relative to
270  // the start of the corresponding token, so compute the start of the Split
271  // that needs to be compressed into a single space relative to the start of
272  // its token.
273  unsigned BreakOffsetInToken =
274  Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
275  unsigned CharsToRemove = Split.second;
276  Whitespaces.replaceWhitespaceInToken(
277  tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "", "",
278  /*InPPDirective=*/false, /*Newlines=*/0, /*Spaces=*/1);
279 }
280 
281 const FormatToken &BreakableComment::tokenAt(unsigned LineIndex) const {
282  return Tokens[LineIndex] ? *Tokens[LineIndex] : Tok;
283 }
284 
285 static bool mayReflowContent(StringRef Content) {
286  Content = Content.trim(Blanks);
287  // Lines starting with '@' commonly have special meaning.
288  // Lines starting with '-', '-#', '+' or '*' are bulleted/numbered lists.
289  bool hasSpecialMeaningPrefix = false;
290  for (StringRef Prefix :
291  {"@", "TODO", "FIXME", "XXX", "-# ", "- ", "+ ", "* "}) {
292  if (Content.startswith(Prefix)) {
293  hasSpecialMeaningPrefix = true;
294  break;
295  }
296  }
297 
298  // Numbered lists may also start with a number followed by '.'
299  // To avoid issues if a line starts with a number which is actually the end
300  // of a previous line, we only consider numbers with up to 2 digits.
301  static auto *const kNumberedListRegexp = new llvm::Regex("^[1-9][0-9]?\\. ");
302  hasSpecialMeaningPrefix =
303  hasSpecialMeaningPrefix || kNumberedListRegexp->match(Content);
304 
305  // Simple heuristic for what to reflow: content should contain at least two
306  // characters and either the first or second character must be
307  // non-punctuation.
308  return Content.size() >= 2 && !hasSpecialMeaningPrefix &&
309  !Content.endswith("\\") &&
310  // Note that this is UTF-8 safe, since if isPunctuation(Content[0]) is
311  // true, then the first code point must be 1 byte long.
312  (!isPunctuation(Content[0]) || !isPunctuation(Content[1]));
313 }
314 
316  const FormatToken &Token, unsigned StartColumn,
317  unsigned OriginalStartColumn, bool FirstInLine, bool InPPDirective,
319  : BreakableComment(Token, StartColumn, InPPDirective, Encoding, Style),
320  DelimitersOnNewline(false),
321  UnbreakableTailLength(Token.UnbreakableTailLength) {
322  assert(Tok.is(TT_BlockComment) &&
323  "block comment section must start with a block comment");
324 
325  StringRef TokenText(Tok.TokenText);
326  assert(TokenText.startswith("/*") && TokenText.endswith("*/"));
327  TokenText.substr(2, TokenText.size() - 4).split(Lines, "\n");
328 
329  int IndentDelta = StartColumn - OriginalStartColumn;
330  Content.resize(Lines.size());
331  Content[0] = Lines[0];
332  ContentColumn.resize(Lines.size());
333  // Account for the initial '/*'.
334  ContentColumn[0] = StartColumn + 2;
335  Tokens.resize(Lines.size());
336  for (size_t i = 1; i < Lines.size(); ++i)
337  adjustWhitespace(i, IndentDelta);
338 
339  // Align decorations with the column of the star on the first line,
340  // that is one column after the start "/*".
341  DecorationColumn = StartColumn + 1;
342 
343  // Account for comment decoration patterns like this:
344  //
345  // /*
346  // ** blah blah blah
347  // */
348  if (Lines.size() >= 2 && Content[1].startswith("**") &&
349  static_cast<unsigned>(ContentColumn[1]) == StartColumn) {
350  DecorationColumn = StartColumn;
351  }
352 
353  Decoration = "* ";
354  if (Lines.size() == 1 && !FirstInLine) {
355  // Comments for which FirstInLine is false can start on arbitrary column,
356  // and available horizontal space can be too small to align consecutive
357  // lines with the first one.
358  // FIXME: We could, probably, align them to current indentation level, but
359  // now we just wrap them without stars.
360  Decoration = "";
361  }
362  for (size_t i = 1, e = Lines.size(); i < e && !Decoration.empty(); ++i) {
363  // If the last line is empty, the closing "*/" will have a star.
364  if (i + 1 == e && Content[i].empty())
365  break;
366  if (!Content[i].empty() && i + 1 != e && Decoration.startswith(Content[i]))
367  continue;
368  while (!Content[i].startswith(Decoration))
369  Decoration = Decoration.substr(0, Decoration.size() - 1);
370  }
371 
372  LastLineNeedsDecoration = true;
373  IndentAtLineBreak = ContentColumn[0] + 1;
374  for (size_t i = 1, e = Lines.size(); i < e; ++i) {
375  if (Content[i].empty()) {
376  if (i + 1 == e) {
377  // Empty last line means that we already have a star as a part of the
378  // trailing */. We also need to preserve whitespace, so that */ is
379  // correctly indented.
380  LastLineNeedsDecoration = false;
381  // Align the star in the last '*/' with the stars on the previous lines.
382  if (e >= 2 && !Decoration.empty()) {
383  ContentColumn[i] = DecorationColumn;
384  }
385  } else if (Decoration.empty()) {
386  // For all other lines, set the start column to 0 if they're empty, so
387  // we do not insert trailing whitespace anywhere.
388  ContentColumn[i] = 0;
389  }
390  continue;
391  }
392 
393  // The first line already excludes the star.
394  // The last line excludes the star if LastLineNeedsDecoration is false.
395  // For all other lines, adjust the line to exclude the star and
396  // (optionally) the first whitespace.
397  unsigned DecorationSize = Decoration.startswith(Content[i])
398  ? Content[i].size()
399  : Decoration.size();
400  if (DecorationSize) {
401  ContentColumn[i] = DecorationColumn + DecorationSize;
402  }
403  Content[i] = Content[i].substr(DecorationSize);
404  if (!Decoration.startswith(Content[i]))
405  IndentAtLineBreak =
406  std::min<int>(IndentAtLineBreak, std::max(0, ContentColumn[i]));
407  }
408  IndentAtLineBreak = std::max<unsigned>(IndentAtLineBreak, Decoration.size());
409 
410  // Detect a multiline jsdoc comment and set DelimitersOnNewline in that case.
411  if (Style.Language == FormatStyle::LK_JavaScript ||
412  Style.Language == FormatStyle::LK_Java) {
413  if ((Lines[0] == "*" || Lines[0].startswith("* ")) && Lines.size() > 1) {
414  // This is a multiline jsdoc comment.
415  DelimitersOnNewline = true;
416  } else if (Lines[0].startswith("* ") && Lines.size() == 1) {
417  // Detect a long single-line comment, like:
418  // /** long long long */
419  // Below, '2' is the width of '*/'.
420  unsigned EndColumn =
421  ContentColumn[0] +
423  Style.TabWidth, Encoding) +
424  2;
425  DelimitersOnNewline = EndColumn > Style.ColumnLimit;
426  }
427  }
428 
429  LLVM_DEBUG({
430  llvm::dbgs() << "IndentAtLineBreak " << IndentAtLineBreak << "\n";
431  llvm::dbgs() << "DelimitersOnNewline " << DelimitersOnNewline << "\n";
432  for (size_t i = 0; i < Lines.size(); ++i) {
433  llvm::dbgs() << i << " |" << Content[i] << "| "
434  << "CC=" << ContentColumn[i] << "| "
435  << "IN=" << (Content[i].data() - Lines[i].data()) << "\n";
436  }
437  });
438 }
439 
440 void BreakableBlockComment::adjustWhitespace(unsigned LineIndex,
441  int IndentDelta) {
442  // When in a preprocessor directive, the trailing backslash in a block comment
443  // is not needed, but can serve a purpose of uniformity with necessary escaped
444  // newlines outside the comment. In this case we remove it here before
445  // trimming the trailing whitespace. The backslash will be re-added later when
446  // inserting a line break.
447  size_t EndOfPreviousLine = Lines[LineIndex - 1].size();
448  if (InPPDirective && Lines[LineIndex - 1].endswith("\\"))
449  --EndOfPreviousLine;
450 
451  // Calculate the end of the non-whitespace text in the previous line.
452  EndOfPreviousLine =
453  Lines[LineIndex - 1].find_last_not_of(Blanks, EndOfPreviousLine);
454  if (EndOfPreviousLine == StringRef::npos)
455  EndOfPreviousLine = 0;
456  else
457  ++EndOfPreviousLine;
458  // Calculate the start of the non-whitespace text in the current line.
459  size_t StartOfLine = Lines[LineIndex].find_first_not_of(Blanks);
460  if (StartOfLine == StringRef::npos)
461  StartOfLine = Lines[LineIndex].rtrim("\r\n").size();
462 
463  StringRef Whitespace = Lines[LineIndex].substr(0, StartOfLine);
464  // Adjust Lines to only contain relevant text.
465  size_t PreviousContentOffset =
466  Content[LineIndex - 1].data() - Lines[LineIndex - 1].data();
467  Content[LineIndex - 1] = Lines[LineIndex - 1].substr(
468  PreviousContentOffset, EndOfPreviousLine - PreviousContentOffset);
469  Content[LineIndex] = Lines[LineIndex].substr(StartOfLine);
470 
471  // Adjust the start column uniformly across all lines.
472  ContentColumn[LineIndex] =
474  IndentDelta;
475 }
476 
477 unsigned BreakableBlockComment::getRangeLength(unsigned LineIndex,
478  unsigned Offset,
479  StringRef::size_type Length,
480  unsigned StartColumn) const {
481  unsigned LineLength =
482  encoding::columnWidthWithTabs(Content[LineIndex].substr(Offset, Length),
483  StartColumn, Style.TabWidth, Encoding);
484  // FIXME: This should go into getRemainingLength instead, but we currently
485  // break tests when putting it there. Investigate how to fix those tests.
486  // The last line gets a "*/" postfix.
487  if (LineIndex + 1 == Lines.size()) {
488  LineLength += 2;
489  // We never need a decoration when breaking just the trailing "*/" postfix.
490  // Note that checking that Length == 0 is not enough, since Length could
491  // also be StringRef::npos.
492  if (Content[LineIndex].substr(Offset, StringRef::npos).empty()) {
493  LineLength -= Decoration.size();
494  }
495  }
496  return LineLength;
497 }
498 
499 unsigned BreakableBlockComment::getRemainingLength(unsigned LineIndex,
500  unsigned Offset,
501  unsigned StartColumn) const {
502  return UnbreakableTailLength +
503  getRangeLength(LineIndex, Offset, StringRef::npos, StartColumn);
504 }
505 
507  bool Break) const {
508  if (Break)
509  return IndentAtLineBreak;
510  return std::max(0, ContentColumn[LineIndex]);
511 }
512 
513 void BreakableBlockComment::insertBreak(unsigned LineIndex, unsigned TailOffset,
514  Split Split,
515  WhitespaceManager &Whitespaces) const {
516  StringRef Text = Content[LineIndex].substr(TailOffset);
517  StringRef Prefix = Decoration;
518  // We need this to account for the case when we have a decoration "* " for all
519  // the lines except for the last one, where the star in "*/" acts as a
520  // decoration.
521  unsigned LocalIndentAtLineBreak = IndentAtLineBreak;
522  if (LineIndex + 1 == Lines.size() &&
523  Text.size() == Split.first + Split.second) {
524  // For the last line we need to break before "*/", but not to add "* ".
525  Prefix = "";
526  if (LocalIndentAtLineBreak >= 2)
527  LocalIndentAtLineBreak -= 2;
528  }
529  // The split offset is from the beginning of the line. Convert it to an offset
530  // from the beginning of the token text.
531  unsigned BreakOffsetInToken =
532  Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
533  unsigned CharsToRemove = Split.second;
534  assert(LocalIndentAtLineBreak >= Prefix.size());
535  Whitespaces.replaceWhitespaceInToken(
536  tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "", Prefix,
537  InPPDirective, /*Newlines=*/1,
538  /*Spaces=*/LocalIndentAtLineBreak - Prefix.size());
539 }
540 
541 BreakableToken::Split
543  llvm::Regex &CommentPragmasRegex) const {
544  if (!mayReflow(LineIndex, CommentPragmasRegex))
545  return Split(StringRef::npos, 0);
546 
547  size_t Trimmed = Content[LineIndex].find_first_not_of(Blanks);
548  return Split(0, Trimmed != StringRef::npos ? Trimmed : 0);
549 }
550 
552  // A break is introduced when we want delimiters on newline.
553  return DelimitersOnNewline &&
554  Lines[0].substr(1).find_first_not_of(Blanks) != StringRef::npos;
555 }
556 
557 void BreakableBlockComment::reflow(unsigned LineIndex,
558  WhitespaceManager &Whitespaces) const {
559  StringRef TrimmedContent = Content[LineIndex].ltrim(Blanks);
560  // Here we need to reflow.
561  assert(Tokens[LineIndex - 1] == Tokens[LineIndex] &&
562  "Reflowing whitespace within a token");
563  // This is the offset of the end of the last line relative to the start of
564  // the token text in the token.
565  unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() +
566  Content[LineIndex - 1].size() -
567  tokenAt(LineIndex).TokenText.data();
568  unsigned WhitespaceLength = TrimmedContent.data() -
569  tokenAt(LineIndex).TokenText.data() -
570  WhitespaceOffsetInToken;
571  Whitespaces.replaceWhitespaceInToken(
572  tokenAt(LineIndex), WhitespaceOffsetInToken,
573  /*ReplaceChars=*/WhitespaceLength, /*PreviousPostfix=*/"",
574  /*CurrentPrefix=*/ReflowPrefix, InPPDirective, /*Newlines=*/0,
575  /*Spaces=*/0);
576 }
577 
579  unsigned LineIndex, WhitespaceManager &Whitespaces) const {
580  if (LineIndex == 0) {
581  if (DelimitersOnNewline) {
582  // Since we're breaking at index 1 below, the break position and the
583  // break length are the same.
584  size_t BreakLength = Lines[0].substr(1).find_first_not_of(Blanks);
585  if (BreakLength != StringRef::npos)
586  insertBreak(LineIndex, 0, Split(1, BreakLength), Whitespaces);
587  }
588  return;
589  }
590  // Here no reflow with the previous line will happen.
591  // Fix the decoration of the line at LineIndex.
592  StringRef Prefix = Decoration;
593  if (Content[LineIndex].empty()) {
594  if (LineIndex + 1 == Lines.size()) {
595  if (!LastLineNeedsDecoration) {
596  // If the last line was empty, we don't need a prefix, as the */ will
597  // line up with the decoration (if it exists).
598  Prefix = "";
599  }
600  } else if (!Decoration.empty()) {
601  // For other empty lines, if we do have a decoration, adapt it to not
602  // contain a trailing whitespace.
603  Prefix = Prefix.substr(0, 1);
604  }
605  } else {
606  if (ContentColumn[LineIndex] == 1) {
607  // This line starts immediately after the decorating *.
608  Prefix = Prefix.substr(0, 1);
609  }
610  }
611  // This is the offset of the end of the last line relative to the start of the
612  // token text in the token.
613  unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() +
614  Content[LineIndex - 1].size() -
615  tokenAt(LineIndex).TokenText.data();
616  unsigned WhitespaceLength = Content[LineIndex].data() -
617  tokenAt(LineIndex).TokenText.data() -
618  WhitespaceOffsetInToken;
619  Whitespaces.replaceWhitespaceInToken(
620  tokenAt(LineIndex), WhitespaceOffsetInToken, WhitespaceLength, "", Prefix,
621  InPPDirective, /*Newlines=*/1, ContentColumn[LineIndex] - Prefix.size());
622 }
623 
624 BreakableToken::Split
626  if (DelimitersOnNewline) {
627  // Replace the trailing whitespace of the last line with a newline.
628  // In case the last line is empty, the ending '*/' is already on its own
629  // line.
630  StringRef Line = Content.back().substr(TailOffset);
631  StringRef TrimmedLine = Line.rtrim(Blanks);
632  if (!TrimmedLine.empty())
633  return Split(TrimmedLine.size(), Line.size() - TrimmedLine.size());
634  }
635  return Split(StringRef::npos, 0);
636 }
637 
638 bool BreakableBlockComment::mayReflow(unsigned LineIndex,
639  llvm::Regex &CommentPragmasRegex) const {
640  // Content[LineIndex] may exclude the indent after the '*' decoration. In that
641  // case, we compute the start of the comment pragma manually.
642  StringRef IndentContent = Content[LineIndex];
643  if (Lines[LineIndex].ltrim(Blanks).startswith("*")) {
644  IndentContent = Lines[LineIndex].ltrim(Blanks).substr(1);
645  }
646  return LineIndex > 0 && !CommentPragmasRegex.match(IndentContent) &&
647  mayReflowContent(Content[LineIndex]) && !Tok.Finalized &&
648  !switchesFormatting(tokenAt(LineIndex));
649 }
650 
652  const FormatToken &Token, unsigned StartColumn,
653  unsigned OriginalStartColumn, bool FirstInLine, bool InPPDirective,
655  : BreakableComment(Token, StartColumn, InPPDirective, Encoding, Style) {
656  assert(Tok.is(TT_LineComment) &&
657  "line comment section must start with a line comment");
658  FormatToken *LineTok = nullptr;
659  for (const FormatToken *CurrentTok = &Tok;
660  CurrentTok && CurrentTok->is(TT_LineComment);
661  CurrentTok = CurrentTok->Next) {
662  LastLineTok = LineTok;
663  StringRef TokenText(CurrentTok->TokenText);
664  assert((TokenText.startswith("//") || TokenText.startswith("#")) &&
665  "unsupported line comment prefix, '//' and '#' are supported");
666  size_t FirstLineIndex = Lines.size();
667  TokenText.split(Lines, "\n");
668  Content.resize(Lines.size());
669  ContentColumn.resize(Lines.size());
670  OriginalContentColumn.resize(Lines.size());
671  Tokens.resize(Lines.size());
672  Prefix.resize(Lines.size());
673  OriginalPrefix.resize(Lines.size());
674  for (size_t i = FirstLineIndex, e = Lines.size(); i < e; ++i) {
675  Lines[i] = Lines[i].ltrim(Blanks);
676  // We need to trim the blanks in case this is not the first line in a
677  // multiline comment. Then the indent is included in Lines[i].
678  StringRef IndentPrefix =
679  getLineCommentIndentPrefix(Lines[i].ltrim(Blanks), Style);
680  assert((TokenText.startswith("//") || TokenText.startswith("#")) &&
681  "unsupported line comment prefix, '//' and '#' are supported");
682  OriginalPrefix[i] = Prefix[i] = IndentPrefix;
683  if (Lines[i].size() > Prefix[i].size() &&
684  isAlphanumeric(Lines[i][Prefix[i].size()])) {
685  if (Prefix[i] == "//")
686  Prefix[i] = "// ";
687  else if (Prefix[i] == "///")
688  Prefix[i] = "/// ";
689  else if (Prefix[i] == "//!")
690  Prefix[i] = "//! ";
691  else if (Prefix[i] == "///<")
692  Prefix[i] = "///< ";
693  else if (Prefix[i] == "//!<")
694  Prefix[i] = "//!< ";
695  else if (Prefix[i] == "#" &&
697  Prefix[i] = "# ";
698  }
699 
700  Tokens[i] = LineTok;
701  Content[i] = Lines[i].substr(IndentPrefix.size());
702  OriginalContentColumn[i] =
703  StartColumn + encoding::columnWidthWithTabs(OriginalPrefix[i],
704  StartColumn,
705  Style.TabWidth, Encoding);
706  ContentColumn[i] =
707  StartColumn + encoding::columnWidthWithTabs(Prefix[i], StartColumn,
708  Style.TabWidth, Encoding);
709 
710  // Calculate the end of the non-whitespace text in this line.
711  size_t EndOfLine = Content[i].find_last_not_of(Blanks);
712  if (EndOfLine == StringRef::npos)
713  EndOfLine = Content[i].size();
714  else
715  ++EndOfLine;
716  Content[i] = Content[i].substr(0, EndOfLine);
717  }
718  LineTok = CurrentTok->Next;
719  if (CurrentTok->Next && !CurrentTok->Next->ContinuesLineCommentSection) {
720  // A line comment section needs to broken by a line comment that is
721  // preceded by at least two newlines. Note that we put this break here
722  // instead of breaking at a previous stage during parsing, since that
723  // would split the contents of the enum into two unwrapped lines in this
724  // example, which is undesirable:
725  // enum A {
726  // a, // comment about a
727  //
728  // // comment about b
729  // b
730  // };
731  //
732  // FIXME: Consider putting separate line comment sections as children to
733  // the unwrapped line instead.
734  break;
735  }
736  }
737 }
738 
739 unsigned
741  StringRef::size_type Length,
742  unsigned StartColumn) const {
744  Content[LineIndex].substr(Offset, Length), StartColumn, Style.TabWidth,
745  Encoding);
746 }
747 
749  bool Break) const {
750  if (Break)
751  return OriginalContentColumn[LineIndex];
752  return ContentColumn[LineIndex];
753 }
754 
756  unsigned LineIndex, unsigned TailOffset, Split Split,
757  WhitespaceManager &Whitespaces) const {
758  StringRef Text = Content[LineIndex].substr(TailOffset);
759  // Compute the offset of the split relative to the beginning of the token
760  // text.
761  unsigned BreakOffsetInToken =
762  Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
763  unsigned CharsToRemove = Split.second;
764  // Compute the size of the new indent, including the size of the new prefix of
765  // the newly broken line.
766  unsigned IndentAtLineBreak = OriginalContentColumn[LineIndex] +
767  Prefix[LineIndex].size() -
768  OriginalPrefix[LineIndex].size();
769  assert(IndentAtLineBreak >= Prefix[LineIndex].size());
770  Whitespaces.replaceWhitespaceInToken(
771  tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "",
772  Prefix[LineIndex], InPPDirective, /*Newlines=*/1,
773  /*Spaces=*/IndentAtLineBreak - Prefix[LineIndex].size());
774 }
775 
777  unsigned LineIndex, llvm::Regex &CommentPragmasRegex) const {
778  if (!mayReflow(LineIndex, CommentPragmasRegex))
779  return Split(StringRef::npos, 0);
780 
781  size_t Trimmed = Content[LineIndex].find_first_not_of(Blanks);
782 
783  // In a line comment section each line is a separate token; thus, after a
784  // split we replace all whitespace before the current line comment token
785  // (which does not need to be included in the split), plus the start of the
786  // line up to where the content starts.
787  return Split(0, Trimmed != StringRef::npos ? Trimmed : 0);
788 }
789 
790 void BreakableLineCommentSection::reflow(unsigned LineIndex,
791  WhitespaceManager &Whitespaces) const {
792  if (LineIndex > 0 && Tokens[LineIndex] != Tokens[LineIndex - 1]) {
793  // Reflow happens between tokens. Replace the whitespace between the
794  // tokens by the empty string.
795  Whitespaces.replaceWhitespace(
796  *Tokens[LineIndex], /*Newlines=*/0, /*Spaces=*/0,
797  /*StartOfTokenColumn=*/StartColumn, /*InPPDirective=*/false);
798  } else if (LineIndex > 0) {
799  // In case we're reflowing after the '\' in:
800  //
801  // // line comment \
802  // // line 2
803  //
804  // the reflow happens inside the single comment token (it is a single line
805  // comment with an unescaped newline).
806  // Replace the whitespace between the '\' and '//' with the empty string.
807  //
808  // Offset points to after the '\' relative to start of the token.
809  unsigned Offset = Lines[LineIndex - 1].data() +
810  Lines[LineIndex - 1].size() -
811  tokenAt(LineIndex - 1).TokenText.data();
812  // WhitespaceLength is the number of chars between the '\' and the '//' on
813  // the next line.
814  unsigned WhitespaceLength =
815  Lines[LineIndex].data() - tokenAt(LineIndex).TokenText.data() - Offset;
816  Whitespaces.replaceWhitespaceInToken(*Tokens[LineIndex],
817  Offset,
818  /*ReplaceChars=*/WhitespaceLength,
819  /*PreviousPostfix=*/"",
820  /*CurrentPrefix=*/"",
821  /*InPPDirective=*/false,
822  /*Newlines=*/0,
823  /*Spaces=*/0);
824 
825  }
826  // Replace the indent and prefix of the token with the reflow prefix.
827  unsigned Offset =
828  Lines[LineIndex].data() - tokenAt(LineIndex).TokenText.data();
829  unsigned WhitespaceLength =
830  Content[LineIndex].data() - Lines[LineIndex].data();
831  Whitespaces.replaceWhitespaceInToken(*Tokens[LineIndex],
832  Offset,
833  /*ReplaceChars=*/WhitespaceLength,
834  /*PreviousPostfix=*/"",
835  /*CurrentPrefix=*/ReflowPrefix,
836  /*InPPDirective=*/false,
837  /*Newlines=*/0,
838  /*Spaces=*/0);
839 }
840 
842  unsigned LineIndex, WhitespaceManager &Whitespaces) const {
843  // If this is the first line of a token, we need to inform Whitespace Manager
844  // about it: either adapt the whitespace range preceding it, or mark it as an
845  // untouchable token.
846  // This happens for instance here:
847  // // line 1 \
848  // // line 2
849  if (LineIndex > 0 && Tokens[LineIndex] != Tokens[LineIndex - 1]) {
850  // This is the first line for the current token, but no reflow with the
851  // previous token is necessary. However, we still may need to adjust the
852  // start column. Note that ContentColumn[LineIndex] is the expected
853  // content column after a possible update to the prefix, hence the prefix
854  // length change is included.
855  unsigned LineColumn =
856  ContentColumn[LineIndex] -
857  (Content[LineIndex].data() - Lines[LineIndex].data()) +
858  (OriginalPrefix[LineIndex].size() - Prefix[LineIndex].size());
859 
860  // We always want to create a replacement instead of adding an untouchable
861  // token, even if LineColumn is the same as the original column of the
862  // token. This is because WhitespaceManager doesn't align trailing
863  // comments if they are untouchable.
864  Whitespaces.replaceWhitespace(*Tokens[LineIndex],
865  /*Newlines=*/1,
866  /*Spaces=*/LineColumn,
867  /*StartOfTokenColumn=*/LineColumn,
868  /*InPPDirective=*/false);
869  }
870  if (OriginalPrefix[LineIndex] != Prefix[LineIndex]) {
871  // Adjust the prefix if necessary.
872 
873  // Take care of the space possibly introduced after a decoration.
874  assert(Prefix[LineIndex] == (OriginalPrefix[LineIndex] + " ").str() &&
875  "Expecting a line comment prefix to differ from original by at most "
876  "a space");
877  Whitespaces.replaceWhitespaceInToken(
878  tokenAt(LineIndex), OriginalPrefix[LineIndex].size(), 0, "", "",
879  /*InPPDirective=*/false, /*Newlines=*/0, /*Spaces=*/1);
880  }
881 }
882 
884  if (LastLineTok) {
885  State.NextToken = LastLineTok->Next;
886  }
887 }
888 
890  unsigned LineIndex, llvm::Regex &CommentPragmasRegex) const {
891  // Line comments have the indent as part of the prefix, so we need to
892  // recompute the start of the line.
893  StringRef IndentContent = Content[LineIndex];
894  if (Lines[LineIndex].startswith("//")) {
895  IndentContent = Lines[LineIndex].substr(2);
896  }
897  // FIXME: Decide whether we want to reflow non-regular indents:
898  // Currently, we only reflow when the OriginalPrefix[LineIndex] matches the
899  // OriginalPrefix[LineIndex-1]. That means we don't reflow
900  // // text that protrudes
901  // // into text with different indent
902  // We do reflow in that case in block comments.
903  return LineIndex > 0 && !CommentPragmasRegex.match(IndentContent) &&
904  mayReflowContent(Content[LineIndex]) && !Tok.Finalized &&
905  !switchesFormatting(tokenAt(LineIndex)) &&
906  OriginalPrefix[LineIndex] == OriginalPrefix[LineIndex - 1];
907 }
908 
909 } // namespace format
910 } // namespace clang
unsigned getLineCount() const override
bool switchesFormatting(const FormatToken &Token)
Checks if Token switches formatting, like /* clang-format off.
Declares BreakableToken, BreakableStringLiteral, BreakableComment, BreakableBlockComment and Breakabl...
Split getSplitAfterLastLine(unsigned TailOffset) const override
static const char *const Blanks
void insertBreak(unsigned LineIndex, unsigned TailOffset, Split Split, WhitespaceManager &Whitespaces) const override
void adaptStartOfLine(unsigned LineIndex, WhitespaceManager &Whitespaces) const override
bool introducesBreakBeforeToken() const override
unsigned getRemainingLength(unsigned LineIndex, unsigned Offset, unsigned StartColumn) const override
void reflow(unsigned LineIndex, WhitespaceManager &Whitespaces) const override
Split getSplit(unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit, unsigned ContentStartColumn, llvm::Regex &CommentPragmasRegex) const override
const FormatToken & tokenAt(unsigned LineIndex) const
BreakableBlockComment(const FormatToken &Token, unsigned StartColumn, unsigned OriginalStartColumn, bool FirstInLine, bool InPPDirective, encoding::Encoding Encoding, const FormatStyle &Style)
FormatToken * Next
The next token in the unwrapped line.
Definition: FormatToken.h:296
static StringRef getLineCommentIndentPrefix(StringRef Comment, const FormatStyle &Style)
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.
SmallVector< int, 16 > ContentColumn
Split getReflowSplit(unsigned LineIndex, llvm::Regex &CommentPragmasRegex) const override
LineState State
Should be used for Java.
Definition: Format.h:1184
void updateNextToken(LineState &State) const override
unsigned getContentStartColumn(unsigned LineIndex, bool Break) const override
Token - This structure provides full information about a lexed token.
Definition: Token.h:35
unsigned columnWidthWithTabs(StringRef Text, unsigned StartColumn, unsigned TabWidth, Encoding Encoding)
Returns the number of columns required to display the Text, starting from the StartColumn on a termin...
Definition: Encoding.h:62
Manages the whitespaces around tokens and their replacements.
uint32_t Offset
Definition: CacheTokens.cpp:43
const FormatToken & Tok
bool mayReflow(unsigned LineIndex, llvm::Regex &CommentPragmasRegex) const override
Split getReflowSplit(unsigned LineIndex, llvm::Regex &CommentPragmasRegex) const override
unsigned ColumnLimit
The column limit.
Definition: Format.h:885
The current state when indenting a unwrapped line.
SmallVector< StringRef, 16 > Content
Should be used for JavaScript.
Definition: Format.h:1186
const AnnotatedLine * Line
unsigned getEscapeSequenceLength(StringRef Text)
Gets the length of an escape sequence inside a C++ string literal.
Definition: Encoding.h:97
A wrapper around a Token storing information about the whitespace characters preceding it...
Definition: FormatToken.h:123
LLVM_READONLY bool isAlphanumeric(unsigned char c)
Return true if this character is an ASCII letter or digit: [a-zA-Z0-9].
Definition: CharInfo.h:118
unsigned getCodePointNumBytes(char FirstChar, Encoding Encoding)
Gets the number of bytes in a sequence representing a single codepoint and starting with FirstChar in...
Definition: Encoding.h:78
void replaceWhitespace(FormatToken &Tok, unsigned Newlines, unsigned Spaces, unsigned StartOfTokenColumn, bool InPPDirective=false)
Replaces the whitespace in front of Tok.
SmallVector< FormatToken *, 16 > Tokens
#define false
Definition: stdbool.h:33
void adaptStartOfLine(unsigned LineIndex, WhitespaceManager &Whitespaces) const override
bool is(tok::TokenKind Kind) const
Definition: FormatToken.h:310
const bool InPPDirective
Various functions to configurably format source code.
unsigned getRangeLength(unsigned LineIndex, unsigned Offset, StringRef::size_type Length, unsigned StartColumn) const override
void compressWhitespace(unsigned LineIndex, unsigned TailOffset, Split Split, WhitespaceManager &Whitespaces) const override
BreakableComment(const FormatToken &Token, unsigned StartColumn, bool InPPDirective, encoding::Encoding Encoding, const FormatStyle &Style)
Creates a breakable token for a comment.
unsigned getRangeLength(unsigned LineIndex, unsigned Offset, StringRef::size_type Length, unsigned StartColumn) const override
StringRef TokenText
The raw text of the token.
Definition: FormatToken.h:171
void insertBreak(unsigned LineIndex, unsigned TailOffset, Split Split, WhitespaceManager &Whitespaces) const override
Split getSplit(unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit, unsigned ContentStartColumn, llvm::Regex &CommentPragmasRegex) const override
BreakableStringLiteral(const FormatToken &Tok, unsigned StartColumn, StringRef Prefix, StringRef Postfix, unsigned UnbreakableTailLength, bool InPPDirective, encoding::Encoding Encoding, const FormatStyle &Style)
Creates a breakable token for a single line string literal.
BreakableLineCommentSection(const FormatToken &Token, unsigned StartColumn, unsigned OriginalStartColumn, bool FirstInLine, bool InPPDirective, encoding::Encoding Encoding, const FormatStyle &Style)
The FormatStyle is used to configure the formatting to follow specific guidelines.
Definition: Format.h:48
static bool mayReflowContent(StringRef Content)
LanguageKind Language
Language, this format style is targeted at.
Definition: Format.h:1201
Dataflow Directional Tag Classes.
unsigned getContentStartColumn(unsigned LineIndex, bool Break) const override
bool Finalized
If true, this token has been fully formatted (indented and potentially re-formatted inside)...
Definition: FormatToken.h:308
unsigned getLengthAfterCompression(unsigned RemainingTokenColumns, Split Split) const
Returns the number of columns needed to format RemainingTokenColumns, assuming that Split is within t...
unsigned getRangeLength(unsigned LineIndex, unsigned Offset, StringRef::size_type Length, unsigned StartColumn) const override
unsigned TabWidth
The number of columns used for tab stops.
Definition: Format.h:1657
unsigned getRemainingLength(unsigned LineIndex, unsigned Offset, unsigned StartColumn) const override
void reflow(unsigned LineIndex, WhitespaceManager &Whitespaces) const override
bool mayReflow(unsigned LineIndex, llvm::Regex &CommentPragmasRegex) const override
static bool IsBlank(char C)
void insertBreak(unsigned LineIndex, unsigned TailOffset, Split Split, WhitespaceManager &Whitespaces) const override
LLVM_READONLY bool isPunctuation(unsigned char c)
Return true if this character is an ASCII punctuation character.
Definition: CharInfo.h:132
unsigned getContentStartColumn(unsigned LineIndex, bool Break) const override
static BreakableToken::Split getStringSplit(StringRef Text, unsigned UsedColumns, unsigned ColumnLimit, unsigned TabWidth, encoding::Encoding Encoding)
__DEVICE__ int max(int __a, int __b)
unsigned getLineCount() const override
Should be used for Protocol Buffer messages in text format (https://developers.google.com/protocol-buffers/).
Definition: Format.h:1196
BreakableToken(const FormatToken &Tok, bool InPPDirective, encoding::Encoding Encoding, const FormatStyle &Style)
StringRef Text
Definition: Format.cpp:1605
FormatToken * NextToken
The token that needs to be next formatted.
static BreakableToken::Split getCommentSplit(StringRef Text, unsigned ContentStartColumn, unsigned ColumnLimit, unsigned TabWidth, encoding::Encoding Encoding)
This file implements an indenter that manages the indentation of continuations.
SmallVector< StringRef, 16 > Lines
const encoding::Encoding Encoding
const FormatStyle & Style