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 /// \brief 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  ArrayRef<const char *> KnownPrefixes(KnownCStylePrefixes);
50  KnownPrefixes = KnownTextProtoPrefixes;
51 
52  StringRef LongestPrefix;
53  for (StringRef KnownPrefix : KnownPrefixes) {
54  if (Comment.startswith(KnownPrefix)) {
55  size_t PrefixLength = KnownPrefix.size();
56  while (PrefixLength < Comment.size() && Comment[PrefixLength] == ' ')
57  ++PrefixLength;
58  if (PrefixLength > LongestPrefix.size())
59  LongestPrefix = Comment.substr(0, PrefixLength);
60  }
61  }
62  return LongestPrefix;
63 }
64 
65 static BreakableToken::Split getCommentSplit(StringRef Text,
66  unsigned ContentStartColumn,
67  unsigned ColumnLimit,
68  unsigned TabWidth,
70  DEBUG(llvm::dbgs() << "Comment split: \"" << Text << ", " << ColumnLimit
71  << "\", Content start: " << ContentStartColumn << "\n");
72  if (ColumnLimit <= ContentStartColumn + 1)
73  return BreakableToken::Split(StringRef::npos, 0);
74 
75  unsigned MaxSplit = ColumnLimit - ContentStartColumn + 1;
76  unsigned MaxSplitBytes = 0;
77 
78  for (unsigned NumChars = 0;
79  NumChars < MaxSplit && MaxSplitBytes < Text.size();) {
80  unsigned BytesInChar =
81  encoding::getCodePointNumBytes(Text[MaxSplitBytes], Encoding);
82  NumChars +=
83  encoding::columnWidthWithTabs(Text.substr(MaxSplitBytes, BytesInChar),
84  ContentStartColumn, TabWidth, Encoding);
85  MaxSplitBytes += BytesInChar;
86  }
87 
88  StringRef::size_type SpaceOffset = Text.find_last_of(Blanks, MaxSplitBytes);
89 
90  // Do not split before a number followed by a dot: this would be interpreted
91  // as a numbered list, which would prevent re-flowing in subsequent passes.
92  static auto *const kNumberedListRegexp = new llvm::Regex("^[1-9][0-9]?\\.");
93  if (SpaceOffset != StringRef::npos &&
94  kNumberedListRegexp->match(Text.substr(SpaceOffset).ltrim(Blanks)))
95  SpaceOffset = Text.find_last_of(Blanks, SpaceOffset);
96 
97  if (SpaceOffset == StringRef::npos ||
98  // Don't break at leading whitespace.
99  Text.find_last_not_of(Blanks, SpaceOffset) == StringRef::npos) {
100  // Make sure that we don't break at leading whitespace that
101  // reaches past MaxSplit.
102  StringRef::size_type FirstNonWhitespace = Text.find_first_not_of(Blanks);
103  if (FirstNonWhitespace == StringRef::npos)
104  // If the comment is only whitespace, we cannot split.
105  return BreakableToken::Split(StringRef::npos, 0);
106  SpaceOffset = Text.find_first_of(
107  Blanks, std::max<unsigned>(MaxSplitBytes, FirstNonWhitespace));
108  }
109  if (SpaceOffset != StringRef::npos && SpaceOffset != 0) {
110  StringRef BeforeCut = Text.substr(0, SpaceOffset).rtrim(Blanks);
111  StringRef AfterCut = Text.substr(SpaceOffset).ltrim(Blanks);
112  return BreakableToken::Split(BeforeCut.size(),
113  AfterCut.begin() - BeforeCut.end());
114  }
115  return BreakableToken::Split(StringRef::npos, 0);
116 }
117 
118 static BreakableToken::Split
119 getStringSplit(StringRef Text, unsigned UsedColumns, unsigned ColumnLimit,
120  unsigned TabWidth, encoding::Encoding Encoding) {
121  // FIXME: Reduce unit test case.
122  if (Text.empty())
123  return BreakableToken::Split(StringRef::npos, 0);
124  if (ColumnLimit <= UsedColumns)
125  return BreakableToken::Split(StringRef::npos, 0);
126  unsigned MaxSplit = ColumnLimit - UsedColumns;
127  StringRef::size_type SpaceOffset = 0;
128  StringRef::size_type SlashOffset = 0;
129  StringRef::size_type WordStartOffset = 0;
130  StringRef::size_type SplitPoint = 0;
131  for (unsigned Chars = 0;;) {
132  unsigned Advance;
133  if (Text[0] == '\\') {
134  Advance = encoding::getEscapeSequenceLength(Text);
135  Chars += Advance;
136  } else {
137  Advance = encoding::getCodePointNumBytes(Text[0], Encoding);
139  Text.substr(0, Advance), UsedColumns + Chars, TabWidth, Encoding);
140  }
141 
142  if (Chars > MaxSplit || Text.size() <= Advance)
143  break;
144 
145  if (IsBlank(Text[0]))
146  SpaceOffset = SplitPoint;
147  if (Text[0] == '/')
148  SlashOffset = SplitPoint;
149  if (Advance == 1 && !isAlphanumeric(Text[0]))
150  WordStartOffset = SplitPoint;
151 
152  SplitPoint += Advance;
153  Text = Text.substr(Advance);
154  }
155 
156  if (SpaceOffset != 0)
157  return BreakableToken::Split(SpaceOffset + 1, 0);
158  if (SlashOffset != 0)
159  return BreakableToken::Split(SlashOffset + 1, 0);
160  if (WordStartOffset != 0)
161  return BreakableToken::Split(WordStartOffset + 1, 0);
162  if (SplitPoint != 0)
163  return BreakableToken::Split(SplitPoint, 0);
164  return BreakableToken::Split(StringRef::npos, 0);
165 }
166 
168  assert((Token.is(TT_BlockComment) || Token.is(TT_LineComment)) &&
169  "formatting regions are switched by comment tokens");
170  StringRef Content = Token.TokenText.substr(2).ltrim();
171  return Content.startswith("clang-format on") ||
172  Content.startswith("clang-format off");
173 }
174 
175 unsigned
176 BreakableToken::getLengthAfterCompression(unsigned RemainingTokenColumns,
177  Split Split) const {
178  // Example: consider the content
179  // lala lala
180  // - RemainingTokenColumns is the original number of columns, 10;
181  // - Split is (4, 2), denoting the two spaces between the two words;
182  //
183  // We compute the number of columns when the split is compressed into a single
184  // space, like:
185  // lala lala
186  //
187  // FIXME: Correctly measure the length of whitespace in Split.second so it
188  // works with tabs.
189  return RemainingTokenColumns + 1 - Split.second;
190 }
191 
192 unsigned BreakableStringLiteral::getLineCount() const { return 1; }
193 
194 unsigned BreakableStringLiteral::getRangeLength(unsigned LineIndex,
195  unsigned Offset,
196  StringRef::size_type Length,
197  unsigned StartColumn) const {
198  llvm_unreachable("Getting the length of a part of the string literal "
199  "indicates that the code tries to reflow it.");
200 }
201 
202 unsigned
204  unsigned StartColumn) const {
205  return UnbreakableTailLength + Postfix.size() +
206  encoding::columnWidthWithTabs(Line.substr(Offset, StringRef::npos),
208 }
209 
211  bool Break) const {
212  return StartColumn + Prefix.size();
213 }
214 
216  const FormatToken &Tok, unsigned StartColumn, StringRef Prefix,
217  StringRef Postfix, unsigned UnbreakableTailLength, bool InPPDirective,
219  : BreakableToken(Tok, InPPDirective, Encoding, Style),
220  StartColumn(StartColumn), Prefix(Prefix), Postfix(Postfix),
221  UnbreakableTailLength(UnbreakableTailLength) {
222  assert(Tok.TokenText.startswith(Prefix) && Tok.TokenText.endswith(Postfix));
223  Line = Tok.TokenText.substr(
224  Prefix.size(), Tok.TokenText.size() - Prefix.size() - Postfix.size());
225 }
226 
227 BreakableToken::Split BreakableStringLiteral::getSplit(
228  unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit,
229  unsigned ContentStartColumn, llvm::Regex &CommentPragmasRegex) const {
230  return getStringSplit(Line.substr(TailOffset), ContentStartColumn,
231  ColumnLimit - Postfix.size(), Style.TabWidth, Encoding);
232 }
233 
234 void BreakableStringLiteral::insertBreak(unsigned LineIndex,
235  unsigned TailOffset, Split Split,
236  WhitespaceManager &Whitespaces) const {
237  Whitespaces.replaceWhitespaceInToken(
238  Tok, Prefix.size() + TailOffset + Split.first, Split.second, Postfix,
240 }
241 
243  unsigned StartColumn, bool InPPDirective,
245  const FormatStyle &Style)
246  : BreakableToken(Token, InPPDirective, Encoding, Style),
247  StartColumn(StartColumn) {}
248 
249 unsigned BreakableComment::getLineCount() const { return Lines.size(); }
250 
251 BreakableToken::Split
252 BreakableComment::getSplit(unsigned LineIndex, unsigned TailOffset,
253  unsigned ColumnLimit, unsigned ContentStartColumn,
254  llvm::Regex &CommentPragmasRegex) const {
255  // Don't break lines matching the comment pragmas regex.
256  if (CommentPragmasRegex.match(Content[LineIndex]))
257  return Split(StringRef::npos, 0);
258  return getCommentSplit(Content[LineIndex].substr(TailOffset),
259  ContentStartColumn, ColumnLimit, Style.TabWidth,
260  Encoding);
261 }
262 
264  unsigned LineIndex, unsigned TailOffset, Split Split,
265  WhitespaceManager &Whitespaces) const {
266  StringRef Text = Content[LineIndex].substr(TailOffset);
267  // Text is relative to the content line, but Whitespaces operates relative to
268  // the start of the corresponding token, so compute the start of the Split
269  // that needs to be compressed into a single space relative to the start of
270  // its token.
271  unsigned BreakOffsetInToken =
272  Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
273  unsigned CharsToRemove = Split.second;
274  Whitespaces.replaceWhitespaceInToken(
275  tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "", "",
276  /*InPPDirective=*/false, /*Newlines=*/0, /*Spaces=*/1);
277 }
278 
279 const FormatToken &BreakableComment::tokenAt(unsigned LineIndex) const {
280  return Tokens[LineIndex] ? *Tokens[LineIndex] : Tok;
281 }
282 
283 static bool mayReflowContent(StringRef Content) {
284  Content = Content.trim(Blanks);
285  // Lines starting with '@' commonly have special meaning.
286  // Lines starting with '-', '-#', '+' or '*' are bulleted/numbered lists.
287  bool hasSpecialMeaningPrefix = false;
288  for (StringRef Prefix :
289  {"@", "TODO", "FIXME", "XXX", "-# ", "- ", "+ ", "* "}) {
290  if (Content.startswith(Prefix)) {
291  hasSpecialMeaningPrefix = true;
292  break;
293  }
294  }
295 
296  // Numbered lists may also start with a number followed by '.'
297  // To avoid issues if a line starts with a number which is actually the end
298  // of a previous line, we only consider numbers with up to 2 digits.
299  static auto *const kNumberedListRegexp = new llvm::Regex("^[1-9][0-9]?\\. ");
300  hasSpecialMeaningPrefix =
301  hasSpecialMeaningPrefix || kNumberedListRegexp->match(Content);
302 
303  // Simple heuristic for what to reflow: content should contain at least two
304  // characters and either the first or second character must be
305  // non-punctuation.
306  return Content.size() >= 2 && !hasSpecialMeaningPrefix &&
307  !Content.endswith("\\") &&
308  // Note that this is UTF-8 safe, since if isPunctuation(Content[0]) is
309  // true, then the first code point must be 1 byte long.
310  (!isPunctuation(Content[0]) || !isPunctuation(Content[1]));
311 }
312 
314  const FormatToken &Token, unsigned StartColumn,
315  unsigned OriginalStartColumn, bool FirstInLine, bool InPPDirective,
317  : BreakableComment(Token, StartColumn, InPPDirective, Encoding, Style),
318  DelimitersOnNewline(false),
319  UnbreakableTailLength(Token.UnbreakableTailLength) {
320  assert(Tok.is(TT_BlockComment) &&
321  "block comment section must start with a block comment");
322 
323  StringRef TokenText(Tok.TokenText);
324  assert(TokenText.startswith("/*") && TokenText.endswith("*/"));
325  TokenText.substr(2, TokenText.size() - 4).split(Lines, "\n");
326 
327  int IndentDelta = StartColumn - OriginalStartColumn;
328  Content.resize(Lines.size());
329  Content[0] = Lines[0];
330  ContentColumn.resize(Lines.size());
331  // Account for the initial '/*'.
332  ContentColumn[0] = StartColumn + 2;
333  Tokens.resize(Lines.size());
334  for (size_t i = 1; i < Lines.size(); ++i)
335  adjustWhitespace(i, IndentDelta);
336 
337  // Align decorations with the column of the star on the first line,
338  // that is one column after the start "/*".
339  DecorationColumn = StartColumn + 1;
340 
341  // Account for comment decoration patterns like this:
342  //
343  // /*
344  // ** blah blah blah
345  // */
346  if (Lines.size() >= 2 && Content[1].startswith("**") &&
347  static_cast<unsigned>(ContentColumn[1]) == StartColumn) {
348  DecorationColumn = StartColumn;
349  }
350 
351  Decoration = "* ";
352  if (Lines.size() == 1 && !FirstInLine) {
353  // Comments for which FirstInLine is false can start on arbitrary column,
354  // and available horizontal space can be too small to align consecutive
355  // lines with the first one.
356  // FIXME: We could, probably, align them to current indentation level, but
357  // now we just wrap them without stars.
358  Decoration = "";
359  }
360  for (size_t i = 1, e = Lines.size(); i < e && !Decoration.empty(); ++i) {
361  // If the last line is empty, the closing "*/" will have a star.
362  if (i + 1 == e && Content[i].empty())
363  break;
364  if (!Content[i].empty() && i + 1 != e && Decoration.startswith(Content[i]))
365  continue;
366  while (!Content[i].startswith(Decoration))
367  Decoration = Decoration.substr(0, Decoration.size() - 1);
368  }
369 
370  LastLineNeedsDecoration = true;
371  IndentAtLineBreak = ContentColumn[0] + 1;
372  for (size_t i = 1, e = Lines.size(); i < e; ++i) {
373  if (Content[i].empty()) {
374  if (i + 1 == e) {
375  // Empty last line means that we already have a star as a part of the
376  // trailing */. We also need to preserve whitespace, so that */ is
377  // correctly indented.
378  LastLineNeedsDecoration = false;
379  // Align the star in the last '*/' with the stars on the previous lines.
380  if (e >= 2 && !Decoration.empty()) {
381  ContentColumn[i] = DecorationColumn;
382  }
383  } else if (Decoration.empty()) {
384  // For all other lines, set the start column to 0 if they're empty, so
385  // we do not insert trailing whitespace anywhere.
386  ContentColumn[i] = 0;
387  }
388  continue;
389  }
390 
391  // The first line already excludes the star.
392  // The last line excludes the star if LastLineNeedsDecoration is false.
393  // For all other lines, adjust the line to exclude the star and
394  // (optionally) the first whitespace.
395  unsigned DecorationSize = Decoration.startswith(Content[i])
396  ? Content[i].size()
397  : Decoration.size();
398  if (DecorationSize) {
399  ContentColumn[i] = DecorationColumn + DecorationSize;
400  }
401  Content[i] = Content[i].substr(DecorationSize);
402  if (!Decoration.startswith(Content[i]))
403  IndentAtLineBreak =
404  std::min<int>(IndentAtLineBreak, std::max(0, ContentColumn[i]));
405  }
406  IndentAtLineBreak = std::max<unsigned>(IndentAtLineBreak, Decoration.size());
407 
408  // Detect a multiline jsdoc comment and set DelimitersOnNewline in that case.
409  if (Style.Language == FormatStyle::LK_JavaScript ||
410  Style.Language == FormatStyle::LK_Java) {
411  if ((Lines[0] == "*" || Lines[0].startswith("* ")) && Lines.size() > 1) {
412  // This is a multiline jsdoc comment.
413  DelimitersOnNewline = true;
414  } else if (Lines[0].startswith("* ") && Lines.size() == 1) {
415  // Detect a long single-line comment, like:
416  // /** long long long */
417  // Below, '2' is the width of '*/'.
418  unsigned EndColumn =
419  ContentColumn[0] +
421  Style.TabWidth, Encoding) +
422  2;
423  DelimitersOnNewline = EndColumn > Style.ColumnLimit;
424  }
425  }
426 
427  DEBUG({
428  llvm::dbgs() << "IndentAtLineBreak " << IndentAtLineBreak << "\n";
429  llvm::dbgs() << "DelimitersOnNewline " << DelimitersOnNewline << "\n";
430  for (size_t i = 0; i < Lines.size(); ++i) {
431  llvm::dbgs() << i << " |" << Content[i] << "| "
432  << "CC=" << ContentColumn[i] << "| "
433  << "IN=" << (Content[i].data() - Lines[i].data()) << "\n";
434  }
435  });
436 }
437 
438 void BreakableBlockComment::adjustWhitespace(unsigned LineIndex,
439  int IndentDelta) {
440  // When in a preprocessor directive, the trailing backslash in a block comment
441  // is not needed, but can serve a purpose of uniformity with necessary escaped
442  // newlines outside the comment. In this case we remove it here before
443  // trimming the trailing whitespace. The backslash will be re-added later when
444  // inserting a line break.
445  size_t EndOfPreviousLine = Lines[LineIndex - 1].size();
446  if (InPPDirective && Lines[LineIndex - 1].endswith("\\"))
447  --EndOfPreviousLine;
448 
449  // Calculate the end of the non-whitespace text in the previous line.
450  EndOfPreviousLine =
451  Lines[LineIndex - 1].find_last_not_of(Blanks, EndOfPreviousLine);
452  if (EndOfPreviousLine == StringRef::npos)
453  EndOfPreviousLine = 0;
454  else
455  ++EndOfPreviousLine;
456  // Calculate the start of the non-whitespace text in the current line.
457  size_t StartOfLine = Lines[LineIndex].find_first_not_of(Blanks);
458  if (StartOfLine == StringRef::npos)
459  StartOfLine = Lines[LineIndex].rtrim("\r\n").size();
460 
461  StringRef Whitespace = Lines[LineIndex].substr(0, StartOfLine);
462  // Adjust Lines to only contain relevant text.
463  size_t PreviousContentOffset =
464  Content[LineIndex - 1].data() - Lines[LineIndex - 1].data();
465  Content[LineIndex - 1] = Lines[LineIndex - 1].substr(
466  PreviousContentOffset, EndOfPreviousLine - PreviousContentOffset);
467  Content[LineIndex] = Lines[LineIndex].substr(StartOfLine);
468 
469  // Adjust the start column uniformly across all lines.
470  ContentColumn[LineIndex] =
472  IndentDelta;
473 }
474 
475 unsigned BreakableBlockComment::getRangeLength(unsigned LineIndex,
476  unsigned Offset,
477  StringRef::size_type Length,
478  unsigned StartColumn) const {
479  unsigned LineLength =
480  encoding::columnWidthWithTabs(Content[LineIndex].substr(Offset, Length),
481  StartColumn, Style.TabWidth, Encoding);
482  // FIXME: This should go into getRemainingLength instead, but we currently
483  // break tests when putting it there. Investigate how to fix those tests.
484  // The last line gets a "*/" postfix.
485  if (LineIndex + 1 == Lines.size()) {
486  LineLength += 2;
487  // We never need a decoration when breaking just the trailing "*/" postfix.
488  // Note that checking that Length == 0 is not enough, since Length could
489  // also be StringRef::npos.
490  if (Content[LineIndex].substr(Offset, StringRef::npos).empty()) {
491  LineLength -= Decoration.size();
492  }
493  }
494  return LineLength;
495 }
496 
497 unsigned BreakableBlockComment::getRemainingLength(unsigned LineIndex,
498  unsigned Offset,
499  unsigned StartColumn) const {
500  return UnbreakableTailLength +
501  getRangeLength(LineIndex, Offset, StringRef::npos, StartColumn);
502 }
503 
505  bool Break) const {
506  if (Break)
507  return IndentAtLineBreak;
508  return std::max(0, ContentColumn[LineIndex]);
509 }
510 
511 void BreakableBlockComment::insertBreak(unsigned LineIndex, unsigned TailOffset,
512  Split Split,
513  WhitespaceManager &Whitespaces) const {
514  StringRef Text = Content[LineIndex].substr(TailOffset);
515  StringRef Prefix = Decoration;
516  // We need this to account for the case when we have a decoration "* " for all
517  // the lines except for the last one, where the star in "*/" acts as a
518  // decoration.
519  unsigned LocalIndentAtLineBreak = IndentAtLineBreak;
520  if (LineIndex + 1 == Lines.size() &&
521  Text.size() == Split.first + Split.second) {
522  // For the last line we need to break before "*/", but not to add "* ".
523  Prefix = "";
524  if (LocalIndentAtLineBreak >= 2)
525  LocalIndentAtLineBreak -= 2;
526  }
527  // The split offset is from the beginning of the line. Convert it to an offset
528  // from the beginning of the token text.
529  unsigned BreakOffsetInToken =
530  Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
531  unsigned CharsToRemove = Split.second;
532  assert(LocalIndentAtLineBreak >= Prefix.size());
533  Whitespaces.replaceWhitespaceInToken(
534  tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "", Prefix,
535  InPPDirective, /*Newlines=*/1,
536  /*Spaces=*/LocalIndentAtLineBreak - Prefix.size());
537 }
538 
539 BreakableToken::Split
541  llvm::Regex &CommentPragmasRegex) const {
542  if (!mayReflow(LineIndex, CommentPragmasRegex))
543  return Split(StringRef::npos, 0);
544 
545  size_t Trimmed = Content[LineIndex].find_first_not_of(Blanks);
546  return Split(0, Trimmed != StringRef::npos ? Trimmed : 0);
547 }
548 
550  // A break is introduced when we want delimiters on newline.
551  return DelimitersOnNewline &&
552  Lines[0].substr(1).find_first_not_of(Blanks) != StringRef::npos;
553 }
554 
555 void BreakableBlockComment::reflow(unsigned LineIndex,
556  WhitespaceManager &Whitespaces) const {
557  StringRef TrimmedContent = Content[LineIndex].ltrim(Blanks);
558  // Here we need to reflow.
559  assert(Tokens[LineIndex - 1] == Tokens[LineIndex] &&
560  "Reflowing whitespace within a token");
561  // This is the offset of the end of the last line relative to the start of
562  // the token text in the token.
563  unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() +
564  Content[LineIndex - 1].size() -
565  tokenAt(LineIndex).TokenText.data();
566  unsigned WhitespaceLength = TrimmedContent.data() -
567  tokenAt(LineIndex).TokenText.data() -
568  WhitespaceOffsetInToken;
569  Whitespaces.replaceWhitespaceInToken(
570  tokenAt(LineIndex), WhitespaceOffsetInToken,
571  /*ReplaceChars=*/WhitespaceLength, /*PreviousPostfix=*/"",
572  /*CurrentPrefix=*/ReflowPrefix, InPPDirective, /*Newlines=*/0,
573  /*Spaces=*/0);
574 }
575 
577  unsigned LineIndex, WhitespaceManager &Whitespaces) const {
578  if (LineIndex == 0) {
579  if (DelimitersOnNewline) {
580  // Since we're breaking at index 1 below, the break position and the
581  // break length are the same.
582  size_t BreakLength = Lines[0].substr(1).find_first_not_of(Blanks);
583  if (BreakLength != StringRef::npos)
584  insertBreak(LineIndex, 0, Split(1, BreakLength), Whitespaces);
585  }
586  return;
587  }
588  // Here no reflow with the previous line will happen.
589  // Fix the decoration of the line at LineIndex.
590  StringRef Prefix = Decoration;
591  if (Content[LineIndex].empty()) {
592  if (LineIndex + 1 == Lines.size()) {
593  if (!LastLineNeedsDecoration) {
594  // If the last line was empty, we don't need a prefix, as the */ will
595  // line up with the decoration (if it exists).
596  Prefix = "";
597  }
598  } else if (!Decoration.empty()) {
599  // For other empty lines, if we do have a decoration, adapt it to not
600  // contain a trailing whitespace.
601  Prefix = Prefix.substr(0, 1);
602  }
603  } else {
604  if (ContentColumn[LineIndex] == 1) {
605  // This line starts immediately after the decorating *.
606  Prefix = Prefix.substr(0, 1);
607  }
608  }
609  // This is the offset of the end of the last line relative to the start of the
610  // token text in the token.
611  unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() +
612  Content[LineIndex - 1].size() -
613  tokenAt(LineIndex).TokenText.data();
614  unsigned WhitespaceLength = Content[LineIndex].data() -
615  tokenAt(LineIndex).TokenText.data() -
616  WhitespaceOffsetInToken;
617  Whitespaces.replaceWhitespaceInToken(
618  tokenAt(LineIndex), WhitespaceOffsetInToken, WhitespaceLength, "", Prefix,
619  InPPDirective, /*Newlines=*/1, ContentColumn[LineIndex] - Prefix.size());
620 }
621 
622 BreakableToken::Split
624  if (DelimitersOnNewline) {
625  // Replace the trailing whitespace of the last line with a newline.
626  // In case the last line is empty, the ending '*/' is already on its own
627  // line.
628  StringRef Line = Content.back().substr(TailOffset);
629  StringRef TrimmedLine = Line.rtrim(Blanks);
630  if (!TrimmedLine.empty())
631  return Split(TrimmedLine.size(), Line.size() - TrimmedLine.size());
632  }
633  return Split(StringRef::npos, 0);
634 }
635 
636 bool BreakableBlockComment::mayReflow(unsigned LineIndex,
637  llvm::Regex &CommentPragmasRegex) const {
638  // Content[LineIndex] may exclude the indent after the '*' decoration. In that
639  // case, we compute the start of the comment pragma manually.
640  StringRef IndentContent = Content[LineIndex];
641  if (Lines[LineIndex].ltrim(Blanks).startswith("*")) {
642  IndentContent = Lines[LineIndex].ltrim(Blanks).substr(1);
643  }
644  return LineIndex > 0 && !CommentPragmasRegex.match(IndentContent) &&
645  mayReflowContent(Content[LineIndex]) && !Tok.Finalized &&
646  !switchesFormatting(tokenAt(LineIndex));
647 }
648 
650  const FormatToken &Token, unsigned StartColumn,
651  unsigned OriginalStartColumn, bool FirstInLine, bool InPPDirective,
653  : BreakableComment(Token, StartColumn, InPPDirective, Encoding, Style) {
654  assert(Tok.is(TT_LineComment) &&
655  "line comment section must start with a line comment");
656  FormatToken *LineTok = nullptr;
657  for (const FormatToken *CurrentTok = &Tok;
658  CurrentTok && CurrentTok->is(TT_LineComment);
659  CurrentTok = CurrentTok->Next) {
660  LastLineTok = LineTok;
661  StringRef TokenText(CurrentTok->TokenText);
662  assert((TokenText.startswith("//") || TokenText.startswith("#")) &&
663  "unsupported line comment prefix, '//' and '#' are supported");
664  size_t FirstLineIndex = Lines.size();
665  TokenText.split(Lines, "\n");
666  Content.resize(Lines.size());
667  ContentColumn.resize(Lines.size());
668  OriginalContentColumn.resize(Lines.size());
669  Tokens.resize(Lines.size());
670  Prefix.resize(Lines.size());
671  OriginalPrefix.resize(Lines.size());
672  for (size_t i = FirstLineIndex, e = Lines.size(); i < e; ++i) {
673  Lines[i] = Lines[i].ltrim(Blanks);
674  // We need to trim the blanks in case this is not the first line in a
675  // multiline comment. Then the indent is included in Lines[i].
676  StringRef IndentPrefix =
677  getLineCommentIndentPrefix(Lines[i].ltrim(Blanks), Style);
678  assert((TokenText.startswith("//") || TokenText.startswith("#")) &&
679  "unsupported line comment prefix, '//' and '#' are supported");
680  OriginalPrefix[i] = Prefix[i] = IndentPrefix;
681  if (Lines[i].size() > Prefix[i].size() &&
682  isAlphanumeric(Lines[i][Prefix[i].size()])) {
683  if (Prefix[i] == "//")
684  Prefix[i] = "// ";
685  else 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] == "#" &&
695  Prefix[i] = "# ";
696  }
697 
698  Tokens[i] = LineTok;
699  Content[i] = Lines[i].substr(IndentPrefix.size());
700  OriginalContentColumn[i] =
701  StartColumn + encoding::columnWidthWithTabs(OriginalPrefix[i],
702  StartColumn,
703  Style.TabWidth, Encoding);
704  ContentColumn[i] =
705  StartColumn + encoding::columnWidthWithTabs(Prefix[i], StartColumn,
706  Style.TabWidth, Encoding);
707 
708  // Calculate the end of the non-whitespace text in this line.
709  size_t EndOfLine = Content[i].find_last_not_of(Blanks);
710  if (EndOfLine == StringRef::npos)
711  EndOfLine = Content[i].size();
712  else
713  ++EndOfLine;
714  Content[i] = Content[i].substr(0, EndOfLine);
715  }
716  LineTok = CurrentTok->Next;
717  if (CurrentTok->Next && !CurrentTok->Next->ContinuesLineCommentSection) {
718  // A line comment section needs to broken by a line comment that is
719  // preceded by at least two newlines. Note that we put this break here
720  // instead of breaking at a previous stage during parsing, since that
721  // would split the contents of the enum into two unwrapped lines in this
722  // example, which is undesirable:
723  // enum A {
724  // a, // comment about a
725  //
726  // // comment about b
727  // b
728  // };
729  //
730  // FIXME: Consider putting separate line comment sections as children to
731  // the unwrapped line instead.
732  break;
733  }
734  }
735 }
736 
737 unsigned
739  StringRef::size_type Length,
740  unsigned StartColumn) const {
742  Content[LineIndex].substr(Offset, Length), StartColumn, Style.TabWidth,
743  Encoding);
744 }
745 
747  bool Break) const {
748  if (Break)
749  return OriginalContentColumn[LineIndex];
750  return ContentColumn[LineIndex];
751 }
752 
754  unsigned LineIndex, unsigned TailOffset, Split Split,
755  WhitespaceManager &Whitespaces) const {
756  StringRef Text = Content[LineIndex].substr(TailOffset);
757  // Compute the offset of the split relative to the beginning of the token
758  // text.
759  unsigned BreakOffsetInToken =
760  Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
761  unsigned CharsToRemove = Split.second;
762  // Compute the size of the new indent, including the size of the new prefix of
763  // the newly broken line.
764  unsigned IndentAtLineBreak = OriginalContentColumn[LineIndex] +
765  Prefix[LineIndex].size() -
766  OriginalPrefix[LineIndex].size();
767  assert(IndentAtLineBreak >= Prefix[LineIndex].size());
768  Whitespaces.replaceWhitespaceInToken(
769  tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "",
770  Prefix[LineIndex], InPPDirective, /*Newlines=*/1,
771  /*Spaces=*/IndentAtLineBreak - Prefix[LineIndex].size());
772 }
773 
775  unsigned LineIndex, llvm::Regex &CommentPragmasRegex) const {
776  if (!mayReflow(LineIndex, CommentPragmasRegex))
777  return Split(StringRef::npos, 0);
778 
779  size_t Trimmed = Content[LineIndex].find_first_not_of(Blanks);
780 
781  // In a line comment section each line is a separate token; thus, after a
782  // split we replace all whitespace before the current line comment token
783  // (which does not need to be included in the split), plus the start of the
784  // line up to where the content starts.
785  return Split(0, Trimmed != StringRef::npos ? Trimmed : 0);
786 }
787 
788 void BreakableLineCommentSection::reflow(unsigned LineIndex,
789  WhitespaceManager &Whitespaces) const {
790  // Reflow happens between tokens. Replace the whitespace between the
791  // tokens by the empty string.
792  Whitespaces.replaceWhitespace(
793  *Tokens[LineIndex], /*Newlines=*/0, /*Spaces=*/0,
794  /*StartOfTokenColumn=*/StartColumn, /*InPPDirective=*/false);
795  // Replace the indent and prefix of the token with the reflow prefix.
796  unsigned WhitespaceLength =
797  Content[LineIndex].data() - tokenAt(LineIndex).TokenText.data();
798  Whitespaces.replaceWhitespaceInToken(*Tokens[LineIndex],
799  /*Offset=*/0,
800  /*ReplaceChars=*/WhitespaceLength,
801  /*PreviousPostfix=*/"",
802  /*CurrentPrefix=*/ReflowPrefix,
803  /*InPPDirective=*/false,
804  /*Newlines=*/0,
805  /*Spaces=*/0);
806 }
807 
809  unsigned LineIndex, WhitespaceManager &Whitespaces) const {
810  // If this is the first line of a token, we need to inform Whitespace Manager
811  // about it: either adapt the whitespace range preceding it, or mark it as an
812  // untouchable token.
813  // This happens for instance here:
814  // // line 1 \
815  // // line 2
816  if (LineIndex > 0 && Tokens[LineIndex] != Tokens[LineIndex - 1]) {
817  // This is the first line for the current token, but no reflow with the
818  // previous token is necessary. However, we still may need to adjust the
819  // start column. Note that ContentColumn[LineIndex] is the expected
820  // content column after a possible update to the prefix, hence the prefix
821  // length change is included.
822  unsigned LineColumn =
823  ContentColumn[LineIndex] -
824  (Content[LineIndex].data() - Lines[LineIndex].data()) +
825  (OriginalPrefix[LineIndex].size() - Prefix[LineIndex].size());
826 
827  // We always want to create a replacement instead of adding an untouchable
828  // token, even if LineColumn is the same as the original column of the
829  // token. This is because WhitespaceManager doesn't align trailing
830  // comments if they are untouchable.
831  Whitespaces.replaceWhitespace(*Tokens[LineIndex],
832  /*Newlines=*/1,
833  /*Spaces=*/LineColumn,
834  /*StartOfTokenColumn=*/LineColumn,
835  /*InPPDirective=*/false);
836  }
837  if (OriginalPrefix[LineIndex] != Prefix[LineIndex]) {
838  // Adjust the prefix if necessary.
839 
840  // Take care of the space possibly introduced after a decoration.
841  assert(Prefix[LineIndex] == (OriginalPrefix[LineIndex] + " ").str() &&
842  "Expecting a line comment prefix to differ from original by at most "
843  "a space");
844  Whitespaces.replaceWhitespaceInToken(
845  tokenAt(LineIndex), OriginalPrefix[LineIndex].size(), 0, "", "",
846  /*InPPDirective=*/false, /*Newlines=*/0, /*Spaces=*/1);
847  }
848 }
849 
851  if (LastLineTok) {
852  State.NextToken = LastLineTok->Next;
853  }
854 }
855 
857  unsigned LineIndex, llvm::Regex &CommentPragmasRegex) const {
858  // Line comments have the indent as part of the prefix, so we need to
859  // recompute the start of the line.
860  StringRef IndentContent = Content[LineIndex];
861  if (Lines[LineIndex].startswith("//")) {
862  IndentContent = Lines[LineIndex].substr(2);
863  }
864  // FIXME: Decide whether we want to reflow non-regular indents:
865  // Currently, we only reflow when the OriginalPrefix[LineIndex] matches the
866  // OriginalPrefix[LineIndex-1]. That means we don't reflow
867  // // text that protrudes
868  // // into text with different indent
869  // We do reflow in that case in block comments.
870  return LineIndex > 0 && !CommentPragmasRegex.match(IndentContent) &&
871  mayReflowContent(Content[LineIndex]) && !Tok.Finalized &&
872  !switchesFormatting(tokenAt(LineIndex)) &&
873  OriginalPrefix[LineIndex] == OriginalPrefix[LineIndex - 1];
874 }
875 
876 } // namespace format
877 } // 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
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:290
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:1215
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:853
The current state when indenting a unwrapped line.
SmallVector< StringRef, 16 > Content
Should be used for JavaScript.
Definition: Format.h:1217
const AnnotatedLine * Line
unsigned getEscapeSequenceLength(StringRef Text)
Gets the length of an escape sequence inside a C++ string literal.
Definition: Encoding.h:97
Split getSplit(unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit, unsigned ReflowColumn, llvm::Regex &CommentPragmasRegex) const override
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
Split getSplit(unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit, unsigned ReflowColumn, llvm::Regex &CommentPragmasRegex) const override
#define false
Definition: stdbool.h:33
void adaptStartOfLine(unsigned LineIndex, WhitespaceManager &Whitespaces) const override
bool is(tok::TokenKind Kind) const
Definition: FormatToken.h:304
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
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:46
static bool mayReflowContent(StringRef Content)
LanguageKind Language
Language, this format style is targeted at.
Definition: Format.h:1232
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:302
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:1674
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:1227
BreakableToken(const FormatToken &Tok, bool InPPDirective, encoding::Encoding Encoding, const FormatStyle &Style)
StringRef Text
Definition: Format.cpp:1553
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