Coverage Report

Created: 2022-07-16 07:03

/Users/buildslave/jenkins/workspace/coverage/llvm-project/clang/lib/Format/BreakableToken.cpp
Line
Count
Source (jump to first uncovered line)
1
//===--- BreakableToken.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
/// Contains implementation of BreakableToken class and classes derived
11
/// from it.
12
///
13
//===----------------------------------------------------------------------===//
14
15
#include "BreakableToken.h"
16
#include "ContinuationIndenter.h"
17
#include "clang/Basic/CharInfo.h"
18
#include "clang/Format/Format.h"
19
#include "llvm/ADT/STLExtras.h"
20
#include "llvm/Support/Debug.h"
21
#include <algorithm>
22
23
#define DEBUG_TYPE "format-token-breaker"
24
25
namespace clang {
26
namespace format {
27
28
static constexpr StringRef Blanks = " \t\v\f\r";
29
13.6k
static bool IsBlank(char C) {
30
13.6k
  switch (C) {
31
1.96k
  case ' ':
32
1.97k
  case '\t':
33
1.97k
  case '\v':
34
1.97k
  case '\f':
35
1.97k
  case '\r':
36
1.97k
    return true;
37
11.7k
  default:
38
11.7k
    return false;
39
13.6k
  }
40
13.6k
}
41
42
static StringRef getLineCommentIndentPrefix(StringRef Comment,
43
15.6k
                                            const FormatStyle &Style) {
44
15.6k
  static constexpr StringRef KnownCStylePrefixes[] = {"///<", "//!<", "///",
45
15.6k
                                                      "//!",  "//:",  "//"};
46
15.6k
  static constexpr StringRef KnownTextProtoPrefixes[] = {"####", "###", "##",
47
15.6k
                                                         "//", "#"};
48
15.6k
  ArrayRef<StringRef> KnownPrefixes(KnownCStylePrefixes);
49
15.6k
  if (Style.Language == FormatStyle::LK_TextProto)
50
264
    KnownPrefixes = KnownTextProtoPrefixes;
51
52
15.6k
  assert(std::is_sorted(KnownPrefixes.begin(), KnownPrefixes.end(),
53
15.6k
                        [](StringRef Lhs, StringRef Rhs) noexcept {
54
15.6k
                          return Lhs.size() > Rhs.size();
55
15.6k
                        }));
56
57
92.5k
  for (StringRef KnownPrefix : KnownPrefixes) {
58
92.5k
    if (Comment.startswith(KnownPrefix)) {
59
15.6k
      const auto PrefixLength =
60
15.6k
          Comment.find_first_not_of(' ', KnownPrefix.size());
61
15.6k
      return Comment.substr(0, PrefixLength);
62
15.6k
    }
63
92.5k
  }
64
0
  return {};
65
15.6k
}
66
67
static BreakableToken::Split
68
getCommentSplit(StringRef Text, unsigned ContentStartColumn,
69
                unsigned ColumnLimit, unsigned TabWidth,
70
                encoding::Encoding Encoding, const FormatStyle &Style,
71
3.85k
                bool DecorationEndsWithStar = false) {
72
3.85k
  LLVM_DEBUG(llvm::dbgs() << "Comment split: \"" << Text
73
3.85k
                          << "\", Column limit: " << ColumnLimit
74
3.85k
                          << ", Content start: " << ContentStartColumn << "\n");
75
3.85k
  if (ColumnLimit <= ContentStartColumn + 1)
76
721
    return BreakableToken::Split(StringRef::npos, 0);
77
78
3.13k
  unsigned MaxSplit = ColumnLimit - ContentStartColumn + 1;
79
3.13k
  unsigned MaxSplitBytes = 0;
80
81
3.13k
  for (unsigned NumChars = 0;
82
65.7k
       NumChars < MaxSplit && 
MaxSplitBytes < Text.size()62.7k
;) {
83
62.6k
    unsigned BytesInChar =
84
62.6k
        encoding::getCodePointNumBytes(Text[MaxSplitBytes], Encoding);
85
62.6k
    NumChars +=
86
62.6k
        encoding::columnWidthWithTabs(Text.substr(MaxSplitBytes, BytesInChar),
87
62.6k
                                      ContentStartColumn, TabWidth, Encoding);
88
62.6k
    MaxSplitBytes += BytesInChar;
89
62.6k
  }
90
91
  // In JavaScript, some @tags can be followed by {, and machinery that parses
92
  // these comments will fail to understand the comment if followed by a line
93
  // break. So avoid ever breaking before a {.
94
3.13k
  if (Style.isJavaScript()) {
95
117
    StringRef::size_type SpaceOffset =
96
117
        Text.find_first_of(Blanks, MaxSplitBytes);
97
117
    if (SpaceOffset != StringRef::npos && 
SpaceOffset + 1 < Text.size()55
&&
98
117
        
Text[SpaceOffset + 1] == '{'40
) {
99
18
      MaxSplitBytes = SpaceOffset + 1;
100
18
    }
101
117
  }
102
103
3.13k
  StringRef::size_type SpaceOffset = Text.find_last_of(Blanks, MaxSplitBytes);
104
105
3.13k
  static const auto kNumberedListRegexp = llvm::Regex("^[1-9][0-9]?\\.");
106
  // Some spaces are unacceptable to break on, rewind past them.
107
3.25k
  while (SpaceOffset != StringRef::npos) {
108
    // If a line-comment ends with `\`, the next line continues the comment,
109
    // whether or not it starts with `//`. This is confusing and triggers
110
    // -Wcomment.
111
    // Avoid introducing multiline comments by not allowing a break right
112
    // after '\'.
113
2.05k
    if (Style.isCpp()) {
114
1.87k
      StringRef::size_type LastNonBlank =
115
1.87k
          Text.find_last_not_of(Blanks, SpaceOffset);
116
1.87k
      if (LastNonBlank != StringRef::npos && 
Text[LastNonBlank] == '\\'1.85k
) {
117
38
        SpaceOffset = Text.find_last_of(Blanks, LastNonBlank);
118
38
        continue;
119
38
      }
120
1.87k
    }
121
122
    // Do not split before a number followed by a dot: this would be interpreted
123
    // as a numbered list, which would prevent re-flowing in subsequent passes.
124
2.01k
    if (kNumberedListRegexp.match(Text.substr(SpaceOffset).ltrim(Blanks))) {
125
4
      SpaceOffset = Text.find_last_of(Blanks, SpaceOffset);
126
4
      continue;
127
4
    }
128
129
    // Avoid ever breaking before a @tag or a { in JavaScript.
130
2.01k
    if (Style.isJavaScript() && 
SpaceOffset + 1 < Text.size()129
&&
131
2.01k
        
(119
Text[SpaceOffset + 1] == '{'119
||
Text[SpaceOffset + 1] == '@'57
)) {
132
79
      SpaceOffset = Text.find_last_of(Blanks, SpaceOffset);
133
79
      continue;
134
79
    }
135
136
1.93k
    break;
137
2.01k
  }
138
139
3.13k
  if (SpaceOffset == StringRef::npos ||
140
      // Don't break at leading whitespace.
141
3.13k
      
Text.find_last_not_of(Blanks, SpaceOffset) == StringRef::npos1.93k
) {
142
    // Make sure that we don't break at leading whitespace that
143
    // reaches past MaxSplit.
144
1.22k
    StringRef::size_type FirstNonWhitespace = Text.find_first_not_of(Blanks);
145
1.22k
    if (FirstNonWhitespace == StringRef::npos) {
146
      // If the comment is only whitespace, we cannot split.
147
14
      return BreakableToken::Split(StringRef::npos, 0);
148
14
    }
149
1.21k
    SpaceOffset = Text.find_first_of(
150
1.21k
        Blanks, std::max<unsigned>(MaxSplitBytes, FirstNonWhitespace));
151
1.21k
  }
152
3.11k
  if (SpaceOffset != StringRef::npos && 
SpaceOffset != 02.58k
) {
153
    // adaptStartOfLine will break after lines starting with /** if the comment
154
    // is broken anywhere. Avoid emitting this break twice here.
155
    // Example: in /** longtextcomesherethatbreaks */ (with ColumnLimit 20) will
156
    // insert a break after /**, so this code must not insert the same break.
157
2.58k
    if (SpaceOffset == 1 && 
Text[SpaceOffset - 1] == '*'3
)
158
2
      return BreakableToken::Split(StringRef::npos, 0);
159
2.58k
    StringRef BeforeCut = Text.substr(0, SpaceOffset).rtrim(Blanks);
160
2.58k
    StringRef AfterCut = Text.substr(SpaceOffset);
161
    // Don't trim the leading blanks if it would create a */ after the break.
162
2.58k
    if (!DecorationEndsWithStar || 
AfterCut.size() <= 16
||
AfterCut[1] != '/'6
)
163
2.57k
      AfterCut = AfterCut.ltrim(Blanks);
164
2.58k
    return BreakableToken::Split(BeforeCut.size(),
165
2.58k
                                 AfterCut.begin() - BeforeCut.end());
166
2.58k
  }
167
535
  return BreakableToken::Split(StringRef::npos, 0);
168
3.11k
}
169
170
static BreakableToken::Split
171
getStringSplit(StringRef Text, unsigned UsedColumns, unsigned ColumnLimit,
172
1.28k
               unsigned TabWidth, encoding::Encoding Encoding) {
173
  // FIXME: Reduce unit test case.
174
1.28k
  if (Text.empty())
175
0
    return BreakableToken::Split(StringRef::npos, 0);
176
1.28k
  if (ColumnLimit <= UsedColumns)
177
5
    return BreakableToken::Split(StringRef::npos, 0);
178
1.27k
  unsigned MaxSplit = ColumnLimit - UsedColumns;
179
1.27k
  StringRef::size_type SpaceOffset = 0;
180
1.27k
  StringRef::size_type SlashOffset = 0;
181
1.27k
  StringRef::size_type WordStartOffset = 0;
182
1.27k
  StringRef::size_type SplitPoint = 0;
183
14.9k
  for (unsigned Chars = 0;;) {
184
14.9k
    unsigned Advance;
185
14.9k
    if (Text[0] == '\\') {
186
35
      Advance = encoding::getEscapeSequenceLength(Text);
187
35
      Chars += Advance;
188
14.9k
    } else {
189
14.9k
      Advance = encoding::getCodePointNumBytes(Text[0], Encoding);
190
14.9k
      Chars += encoding::columnWidthWithTabs(
191
14.9k
          Text.substr(0, Advance), UsedColumns + Chars, TabWidth, Encoding);
192
14.9k
    }
193
194
14.9k
    if (Chars > MaxSplit || 
Text.size() <= Advance13.7k
)
195
1.27k
      break;
196
197
13.6k
    if (IsBlank(Text[0]))
198
1.97k
      SpaceOffset = SplitPoint;
199
13.6k
    if (Text[0] == '/')
200
22
      SlashOffset = SplitPoint;
201
13.6k
    if (Advance == 1 && 
!isAlphanumeric(Text[0])13.5k
)
202
2.22k
      WordStartOffset = SplitPoint;
203
204
13.6k
    SplitPoint += Advance;
205
13.6k
    Text = Text.substr(Advance);
206
13.6k
  }
207
208
1.27k
  if (SpaceOffset != 0)
209
688
    return BreakableToken::Split(SpaceOffset + 1, 0);
210
590
  if (SlashOffset != 0)
211
12
    return BreakableToken::Split(SlashOffset + 1, 0);
212
578
  if (WordStartOffset != 0)
213
14
    return BreakableToken::Split(WordStartOffset + 1, 0);
214
564
  if (SplitPoint != 0)
215
544
    return BreakableToken::Split(SplitPoint, 0);
216
20
  return BreakableToken::Split(StringRef::npos, 0);
217
564
}
218
219
19.1k
bool switchesFormatting(const FormatToken &Token) {
220
19.1k
  assert((Token.is(TT_BlockComment) || Token.is(TT_LineComment)) &&
221
19.1k
         "formatting regions are switched by comment tokens");
222
0
  StringRef Content = Token.TokenText.substr(2).ltrim();
223
19.1k
  return Content.startswith("clang-format on") ||
224
19.1k
         
Content.startswith("clang-format off")19.0k
;
225
19.1k
}
226
227
unsigned
228
BreakableToken::getLengthAfterCompression(unsigned RemainingTokenColumns,
229
1.60k
                                          Split Split) const {
230
  // Example: consider the content
231
  // lala  lala
232
  // - RemainingTokenColumns is the original number of columns, 10;
233
  // - Split is (4, 2), denoting the two spaces between the two words;
234
  //
235
  // We compute the number of columns when the split is compressed into a single
236
  // space, like:
237
  // lala lala
238
  //
239
  // FIXME: Correctly measure the length of whitespace in Split.second so it
240
  // works with tabs.
241
1.60k
  return RemainingTokenColumns + 1 - Split.second;
242
1.60k
}
243
244
20.7k
unsigned BreakableStringLiteral::getLineCount() const { return 1; }
245
246
unsigned BreakableStringLiteral::getRangeLength(unsigned LineIndex,
247
                                                unsigned Offset,
248
                                                StringRef::size_type Length,
249
0
                                                unsigned StartColumn) const {
250
0
  llvm_unreachable("Getting the length of a part of the string literal "
251
0
                   "indicates that the code tries to reflow it.");
252
0
}
253
254
unsigned
255
BreakableStringLiteral::getRemainingLength(unsigned LineIndex, unsigned Offset,
256
11.6k
                                           unsigned StartColumn) const {
257
11.6k
  return UnbreakableTailLength + Postfix.size() +
258
11.6k
         encoding::columnWidthWithTabs(Line.substr(Offset), StartColumn,
259
11.6k
                                       Style.TabWidth, Encoding);
260
11.6k
}
261
262
unsigned BreakableStringLiteral::getContentStartColumn(unsigned LineIndex,
263
11.6k
                                                       bool Break) const {
264
11.6k
  return StartColumn + Prefix.size();
265
11.6k
}
266
267
BreakableStringLiteral::BreakableStringLiteral(
268
    const FormatToken &Tok, unsigned StartColumn, StringRef Prefix,
269
    StringRef Postfix, unsigned UnbreakableTailLength, bool InPPDirective,
270
    encoding::Encoding Encoding, const FormatStyle &Style)
271
    : BreakableToken(Tok, InPPDirective, Encoding, Style),
272
      StartColumn(StartColumn), Prefix(Prefix), Postfix(Postfix),
273
10.3k
      UnbreakableTailLength(UnbreakableTailLength) {
274
10.3k
  assert(Tok.TokenText.startswith(Prefix) && Tok.TokenText.endswith(Postfix));
275
0
  Line = Tok.TokenText.substr(
276
10.3k
      Prefix.size(), Tok.TokenText.size() - Prefix.size() - Postfix.size());
277
10.3k
}
278
279
BreakableToken::Split BreakableStringLiteral::getSplit(
280
    unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit,
281
1.28k
    unsigned ContentStartColumn, const llvm::Regex &CommentPragmasRegex) const {
282
1.28k
  return getStringSplit(Line.substr(TailOffset), ContentStartColumn,
283
1.28k
                        ColumnLimit - Postfix.size(), Style.TabWidth, Encoding);
284
1.28k
}
285
286
void BreakableStringLiteral::insertBreak(unsigned LineIndex,
287
                                         unsigned TailOffset, Split Split,
288
                                         unsigned ContentIndent,
289
126
                                         WhitespaceManager &Whitespaces) const {
290
126
  Whitespaces.replaceWhitespaceInToken(
291
126
      Tok, Prefix.size() + TailOffset + Split.first, Split.second, Postfix,
292
126
      Prefix, InPPDirective, 1, StartColumn);
293
126
}
294
295
BreakableComment::BreakableComment(const FormatToken &Token,
296
                                   unsigned StartColumn, bool InPPDirective,
297
                                   encoding::Encoding Encoding,
298
                                   const FormatStyle &Style)
299
    : BreakableToken(Token, InPPDirective, Encoding, Style),
300
16.5k
      StartColumn(StartColumn) {}
301
302
33.0k
unsigned BreakableComment::getLineCount() const { return Lines.size(); }
303
304
BreakableToken::Split
305
BreakableComment::getSplit(unsigned LineIndex, unsigned TailOffset,
306
                           unsigned ColumnLimit, unsigned ContentStartColumn,
307
2.83k
                           const llvm::Regex &CommentPragmasRegex) const {
308
  // Don't break lines matching the comment pragmas regex.
309
2.83k
  if (CommentPragmasRegex.match(Content[LineIndex]))
310
0
    return Split(StringRef::npos, 0);
311
2.83k
  return getCommentSplit(Content[LineIndex].substr(TailOffset),
312
2.83k
                         ContentStartColumn, ColumnLimit, Style.TabWidth,
313
2.83k
                         Encoding, Style);
314
2.83k
}
315
316
void BreakableComment::compressWhitespace(
317
    unsigned LineIndex, unsigned TailOffset, Split Split,
318
32
    WhitespaceManager &Whitespaces) const {
319
32
  StringRef Text = Content[LineIndex].substr(TailOffset);
320
  // Text is relative to the content line, but Whitespaces operates relative to
321
  // the start of the corresponding token, so compute the start of the Split
322
  // that needs to be compressed into a single space relative to the start of
323
  // its token.
324
32
  unsigned BreakOffsetInToken =
325
32
      Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
326
32
  unsigned CharsToRemove = Split.second;
327
32
  Whitespaces.replaceWhitespaceInToken(
328
32
      tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "", "",
329
32
      /*InPPDirective=*/false, /*Newlines=*/0, /*Spaces=*/1);
330
32
}
331
332
4.04k
const FormatToken &BreakableComment::tokenAt(unsigned LineIndex) const {
333
4.04k
  return Tokens[LineIndex] ? 
*Tokens[LineIndex]1.46k
:
Tok2.57k
;
334
4.04k
}
335
336
813
static bool mayReflowContent(StringRef Content) {
337
813
  Content = Content.trim(Blanks);
338
  // Lines starting with '@' commonly have special meaning.
339
  // Lines starting with '-', '-#', '+' or '*' are bulleted/numbered lists.
340
813
  bool hasSpecialMeaningPrefix = false;
341
813
  for (StringRef Prefix :
342
6.37k
       {"@", "TODO", "FIXME", "XXX", "-# ", "- ", "+ ", "* "}) {
343
6.37k
    if (Content.startswith(Prefix)) {
344
30
      hasSpecialMeaningPrefix = true;
345
30
      break;
346
30
    }
347
6.37k
  }
348
349
  // Numbered lists may also start with a number followed by '.'
350
  // To avoid issues if a line starts with a number which is actually the end
351
  // of a previous line, we only consider numbers with up to 2 digits.
352
813
  static const auto kNumberedListRegexp = llvm::Regex("^[1-9][0-9]?\\. ");
353
813
  hasSpecialMeaningPrefix =
354
813
      hasSpecialMeaningPrefix || 
kNumberedListRegexp.match(Content)783
;
355
356
  // Simple heuristic for what to reflow: content should contain at least two
357
  // characters and either the first or second character must be
358
  // non-punctuation.
359
813
  return Content.size() >= 2 && 
!hasSpecialMeaningPrefix694
&&
360
813
         
!Content.endswith("\\")662
&&
361
         // Note that this is UTF-8 safe, since if isPunctuation(Content[0]) is
362
         // true, then the first code point must be 1 byte long.
363
813
         
(662
!isPunctuation(Content[0])662
||
!isPunctuation(Content[1])6
);
364
813
}
365
366
BreakableBlockComment::BreakableBlockComment(
367
    const FormatToken &Token, unsigned StartColumn,
368
    unsigned OriginalStartColumn, bool FirstInLine, bool InPPDirective,
369
    encoding::Encoding Encoding, const FormatStyle &Style, bool UseCRLF)
370
    : BreakableComment(Token, StartColumn, InPPDirective, Encoding, Style),
371
      DelimitersOnNewline(false),
372
2.70k
      UnbreakableTailLength(Token.UnbreakableTailLength) {
373
2.70k
  assert(Tok.is(TT_BlockComment) &&
374
2.70k
         "block comment section must start with a block comment");
375
376
0
  StringRef TokenText(Tok.TokenText);
377
2.70k
  assert(TokenText.startswith("/*") && TokenText.endswith("*/"));
378
0
  TokenText.substr(2, TokenText.size() - 4)
379
2.70k
      .split(Lines, UseCRLF ? 
"\r\n"8
:
"\n"2.69k
);
380
381
2.70k
  int IndentDelta = StartColumn - OriginalStartColumn;
382
2.70k
  Content.resize(Lines.size());
383
2.70k
  Content[0] = Lines[0];
384
2.70k
  ContentColumn.resize(Lines.size());
385
  // Account for the initial '/*'.
386
2.70k
  ContentColumn[0] = StartColumn + 2;
387
2.70k
  Tokens.resize(Lines.size());
388
3.90k
  for (size_t i = 1; i < Lines.size(); 
++i1.20k
)
389
1.20k
    adjustWhitespace(i, IndentDelta);
390
391
  // Align decorations with the column of the star on the first line,
392
  // that is one column after the start "/*".
393
2.70k
  DecorationColumn = StartColumn + 1;
394
395
  // Account for comment decoration patterns like this:
396
  //
397
  // /*
398
  // ** blah blah blah
399
  // */
400
2.70k
  if (Lines.size() >= 2 && 
Content[1].startswith("**")577
&&
401
2.70k
      
static_cast<unsigned>(ContentColumn[1]) == StartColumn10
) {
402
6
    DecorationColumn = StartColumn;
403
6
  }
404
405
2.70k
  Decoration = "* ";
406
2.70k
  if (Lines.size() == 1 && 
!FirstInLine2.12k
) {
407
    // Comments for which FirstInLine is false can start on arbitrary column,
408
    // and available horizontal space can be too small to align consecutive
409
    // lines with the first one.
410
    // FIXME: We could, probably, align them to current indentation level, but
411
    // now we just wrap them without stars.
412
1.62k
    Decoration = "";
413
1.62k
  }
414
3.36k
  for (size_t i = 1, e = Content.size(); i < e && 
!Decoration.empty()1.06k
;
++i663
) {
415
880
    const StringRef &Text = Content[i];
416
880
    if (i + 1 == e) {
417
      // If the last line is empty, the closing "*/" will have a star.
418
396
      if (Text.empty())
419
217
        break;
420
484
    } else if (!Text.empty() && 
Decoration.startswith(Text)470
) {
421
19
      continue;
422
19
    }
423
1.21k
    
while (644
!Text.startswith(Decoration))
424
570
      Decoration = Decoration.drop_back(1);
425
644
  }
426
427
2.70k
  LastLineNeedsDecoration = true;
428
2.70k
  IndentAtLineBreak = ContentColumn[0] + 1;
429
3.90k
  for (size_t i = 1, e = Lines.size(); i < e; 
++i1.20k
) {
430
1.20k
    if (Content[i].empty()) {
431
405
      if (i + 1 == e) {
432
        // Empty last line means that we already have a star as a part of the
433
        // trailing */. We also need to preserve whitespace, so that */ is
434
        // correctly indented.
435
389
        LastLineNeedsDecoration = false;
436
        // Align the star in the last '*/' with the stars on the previous lines.
437
389
        if (e >= 2 && !Decoration.empty())
438
217
          ContentColumn[i] = DecorationColumn;
439
389
      } else 
if (16
Decoration.empty()16
) {
440
        // For all other lines, set the start column to 0 if they're empty, so
441
        // we do not insert trailing whitespace anywhere.
442
16
        ContentColumn[i] = 0;
443
16
      }
444
405
      continue;
445
405
    }
446
447
    // The first line already excludes the star.
448
    // The last line excludes the star if LastLineNeedsDecoration is false.
449
    // For all other lines, adjust the line to exclude the star and
450
    // (optionally) the first whitespace.
451
800
    unsigned DecorationSize = Decoration.startswith(Content[i])
452
800
                                  ? 
Content[i].size()21
453
800
                                  : 
Decoration.size()779
;
454
800
    if (DecorationSize)
455
387
      ContentColumn[i] = DecorationColumn + DecorationSize;
456
800
    Content[i] = Content[i].substr(DecorationSize);
457
800
    if (!Decoration.startswith(Content[i])) {
458
771
      IndentAtLineBreak =
459
771
          std::min<int>(IndentAtLineBreak, std::max(0, ContentColumn[i]));
460
771
    }
461
800
  }
462
2.70k
  IndentAtLineBreak = std::max<unsigned>(IndentAtLineBreak, Decoration.size());
463
464
  // Detect a multiline jsdoc comment and set DelimitersOnNewline in that case.
465
2.70k
  if (Style.isJavaScript() || 
Style.Language == FormatStyle::LK_Java2.52k
) {
466
200
    if ((Lines[0] == "*" || 
Lines[0].startswith("* ")110
) &&
Lines.size() > 1150
) {
467
      // This is a multiline jsdoc comment.
468
102
      DelimitersOnNewline = true;
469
102
    } else 
if (98
Lines[0].startswith("* ")98
&&
Lines.size() == 148
) {
470
      // Detect a long single-line comment, like:
471
      // /** long long long */
472
      // Below, '2' is the width of '*/'.
473
48
      unsigned EndColumn =
474
48
          ContentColumn[0] +
475
48
          encoding::columnWidthWithTabs(Lines[0], ContentColumn[0],
476
48
                                        Style.TabWidth, Encoding) +
477
48
          2;
478
48
      DelimitersOnNewline = EndColumn > Style.ColumnLimit;
479
48
    }
480
200
  }
481
482
2.70k
  LLVM_DEBUG({
483
2.70k
    llvm::dbgs() << "IndentAtLineBreak " << IndentAtLineBreak << "\n";
484
2.70k
    llvm::dbgs() << "DelimitersOnNewline " << DelimitersOnNewline << "\n";
485
2.70k
    for (size_t i = 0; i < Lines.size(); ++i) {
486
2.70k
      llvm::dbgs() << i << " |" << Content[i] << "| "
487
2.70k
                   << "CC=" << ContentColumn[i] << "| "
488
2.70k
                   << "IN=" << (Content[i].data() - Lines[i].data()) << "\n";
489
2.70k
    }
490
2.70k
  });
491
2.70k
}
492
493
BreakableToken::Split BreakableBlockComment::getSplit(
494
    unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit,
495
1.02k
    unsigned ContentStartColumn, const llvm::Regex &CommentPragmasRegex) const {
496
  // Don't break lines matching the comment pragmas regex.
497
1.02k
  if (CommentPragmasRegex.match(Content[LineIndex]))
498
6
    return Split(StringRef::npos, 0);
499
1.01k
  return getCommentSplit(Content[LineIndex].substr(TailOffset),
500
1.01k
                         ContentStartColumn, ColumnLimit, Style.TabWidth,
501
1.01k
                         Encoding, Style, Decoration.endswith("*"));
502
1.02k
}
503
504
void BreakableBlockComment::adjustWhitespace(unsigned LineIndex,
505
1.20k
                                             int IndentDelta) {
506
  // When in a preprocessor directive, the trailing backslash in a block comment
507
  // is not needed, but can serve a purpose of uniformity with necessary escaped
508
  // newlines outside the comment. In this case we remove it here before
509
  // trimming the trailing whitespace. The backslash will be re-added later when
510
  // inserting a line break.
511
1.20k
  size_t EndOfPreviousLine = Lines[LineIndex - 1].size();
512
1.20k
  if (InPPDirective && 
Lines[LineIndex - 1].endswith("\\")22
)
513
14
    --EndOfPreviousLine;
514
515
  // Calculate the end of the non-whitespace text in the previous line.
516
1.20k
  EndOfPreviousLine =
517
1.20k
      Lines[LineIndex - 1].find_last_not_of(Blanks, EndOfPreviousLine);
518
1.20k
  if (EndOfPreviousLine == StringRef::npos)
519
262
    EndOfPreviousLine = 0;
520
943
  else
521
943
    ++EndOfPreviousLine;
522
  // Calculate the start of the non-whitespace text in the current line.
523
1.20k
  size_t StartOfLine = Lines[LineIndex].find_first_not_of(Blanks);
524
1.20k
  if (StartOfLine == StringRef::npos)
525
405
    StartOfLine = Lines[LineIndex].size();
526
527
1.20k
  StringRef Whitespace = Lines[LineIndex].substr(0, StartOfLine);
528
  // Adjust Lines to only contain relevant text.
529
1.20k
  size_t PreviousContentOffset =
530
1.20k
      Content[LineIndex - 1].data() - Lines[LineIndex - 1].data();
531
1.20k
  Content[LineIndex - 1] = Lines[LineIndex - 1].substr(
532
1.20k
      PreviousContentOffset, EndOfPreviousLine - PreviousContentOffset);
533
1.20k
  Content[LineIndex] = Lines[LineIndex].substr(StartOfLine);
534
535
  // Adjust the start column uniformly across all lines.
536
1.20k
  ContentColumn[LineIndex] =
537
1.20k
      encoding::columnWidthWithTabs(Whitespace, 0, Style.TabWidth, Encoding) +
538
1.20k
      IndentDelta;
539
1.20k
}
540
541
unsigned BreakableBlockComment::getRangeLength(unsigned LineIndex,
542
                                               unsigned Offset,
543
                                               StringRef::size_type Length,
544
5.30k
                                               unsigned StartColumn) const {
545
5.30k
  return encoding::columnWidthWithTabs(
546
5.30k
      Content[LineIndex].substr(Offset, Length), StartColumn, Style.TabWidth,
547
5.30k
      Encoding);
548
5.30k
}
549
550
unsigned BreakableBlockComment::getRemainingLength(unsigned LineIndex,
551
                                                   unsigned Offset,
552
4.73k
                                                   unsigned StartColumn) const {
553
4.73k
  unsigned LineLength =
554
4.73k
      UnbreakableTailLength +
555
4.73k
      getRangeLength(LineIndex, Offset, StringRef::npos, StartColumn);
556
4.73k
  if (LineIndex + 1 == Lines.size()) {
557
3.11k
    LineLength += 2;
558
    // We never need a decoration when breaking just the trailing "*/" postfix.
559
3.11k
    bool HasRemainingText = Offset < Content[LineIndex].size();
560
3.11k
    if (!HasRemainingText) {
561
657
      bool HasDecoration = Lines[LineIndex].ltrim().startswith(Decoration);
562
657
      if (HasDecoration)
563
420
        LineLength -= Decoration.size();
564
657
    }
565
3.11k
  }
566
4.73k
  return LineLength;
567
4.73k
}
568
569
unsigned BreakableBlockComment::getContentStartColumn(unsigned LineIndex,
570
4.28k
                                                      bool Break) const {
571
4.28k
  if (Break)
572
447
    return IndentAtLineBreak;
573
3.83k
  return std::max(0, ContentColumn[LineIndex]);
574
4.28k
}
575
576
const llvm::StringSet<>
577
    BreakableBlockComment::ContentIndentingJavadocAnnotations = {
578
        "@param", "@return",     "@returns", "@throws",  "@type", "@template",
579
        "@see",   "@deprecated", "@define",  "@exports", "@mods", "@private",
580
};
581
582
490
unsigned BreakableBlockComment::getContentIndent(unsigned LineIndex) const {
583
490
  if (Style.Language != FormatStyle::LK_Java && 
!Style.isJavaScript()468
)
584
403
    return 0;
585
  // The content at LineIndex 0 of a comment like:
586
  // /** line 0 */
587
  // is "* line 0", so we need to skip over the decoration in that case.
588
87
  StringRef ContentWithNoDecoration = Content[LineIndex];
589
87
  if (LineIndex == 0 && 
ContentWithNoDecoration.startswith("*")35
)
590
35
    ContentWithNoDecoration = ContentWithNoDecoration.substr(1).ltrim(Blanks);
591
87
  StringRef FirstWord = ContentWithNoDecoration.substr(
592
87
      0, ContentWithNoDecoration.find_first_of(Blanks));
593
87
  if (ContentIndentingJavadocAnnotations.find(FirstWord) !=
594
87
      ContentIndentingJavadocAnnotations.end()) {
595
56
    return Style.ContinuationIndentWidth;
596
56
  }
597
31
  return 0;
598
87
}
599
600
void BreakableBlockComment::insertBreak(unsigned LineIndex, unsigned TailOffset,
601
                                        Split Split, unsigned ContentIndent,
602
164
                                        WhitespaceManager &Whitespaces) const {
603
164
  StringRef Text = Content[LineIndex].substr(TailOffset);
604
164
  StringRef Prefix = Decoration;
605
  // We need this to account for the case when we have a decoration "* " for all
606
  // the lines except for the last one, where the star in "*/" acts as a
607
  // decoration.
608
164
  unsigned LocalIndentAtLineBreak = IndentAtLineBreak;
609
164
  if (LineIndex + 1 == Lines.size() &&
610
164
      
Text.size() == Split.first + Split.second69
) {
611
    // For the last line we need to break before "*/", but not to add "* ".
612
17
    Prefix = "";
613
17
    if (LocalIndentAtLineBreak >= 2)
614
17
      LocalIndentAtLineBreak -= 2;
615
17
  }
616
  // The split offset is from the beginning of the line. Convert it to an offset
617
  // from the beginning of the token text.
618
164
  unsigned BreakOffsetInToken =
619
164
      Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
620
164
  unsigned CharsToRemove = Split.second;
621
164
  assert(LocalIndentAtLineBreak >= Prefix.size());
622
0
  std::string PrefixWithTrailingIndent = std::string(Prefix);
623
164
  PrefixWithTrailingIndent.append(ContentIndent, ' ');
624
164
  Whitespaces.replaceWhitespaceInToken(
625
164
      tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "",
626
164
      PrefixWithTrailingIndent, InPPDirective, /*Newlines=*/1,
627
164
      /*Spaces=*/LocalIndentAtLineBreak + ContentIndent -
628
164
          PrefixWithTrailingIndent.size());
629
164
}
630
631
BreakableToken::Split BreakableBlockComment::getReflowSplit(
632
208
    unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
633
208
  if (!mayReflow(LineIndex, CommentPragmasRegex))
634
119
    return Split(StringRef::npos, 0);
635
636
  // If we're reflowing into a line with content indent, only reflow the next
637
  // line if its starting whitespace matches the content indent.
638
89
  size_t Trimmed = Content[LineIndex].find_first_not_of(Blanks);
639
89
  if (LineIndex) {
640
89
    unsigned PreviousContentIndent = getContentIndent(LineIndex - 1);
641
89
    if (PreviousContentIndent && 
Trimmed != StringRef::npos6
&&
642
89
        
Trimmed != PreviousContentIndent6
) {
643
4
      return Split(StringRef::npos, 0);
644
4
    }
645
89
  }
646
647
85
  return Split(0, Trimmed != StringRef::npos ? Trimmed : 
00
);
648
89
}
649
650
2.70k
bool BreakableBlockComment::introducesBreakBeforeToken() const {
651
  // A break is introduced when we want delimiters on newline.
652
2.70k
  return DelimitersOnNewline &&
653
2.70k
         
Lines[0].substr(1).find_first_not_of(Blanks) != StringRef::npos125
;
654
2.70k
}
655
656
void BreakableBlockComment::reflow(unsigned LineIndex,
657
29
                                   WhitespaceManager &Whitespaces) const {
658
29
  StringRef TrimmedContent = Content[LineIndex].ltrim(Blanks);
659
  // Here we need to reflow.
660
29
  assert(Tokens[LineIndex - 1] == Tokens[LineIndex] &&
661
29
         "Reflowing whitespace within a token");
662
  // This is the offset of the end of the last line relative to the start of
663
  // the token text in the token.
664
0
  unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() +
665
29
                                     Content[LineIndex - 1].size() -
666
29
                                     tokenAt(LineIndex).TokenText.data();
667
29
  unsigned WhitespaceLength = TrimmedContent.data() -
668
29
                              tokenAt(LineIndex).TokenText.data() -
669
29
                              WhitespaceOffsetInToken;
670
29
  Whitespaces.replaceWhitespaceInToken(
671
29
      tokenAt(LineIndex), WhitespaceOffsetInToken,
672
29
      /*ReplaceChars=*/WhitespaceLength, /*PreviousPostfix=*/"",
673
29
      /*CurrentPrefix=*/ReflowPrefix, InPPDirective, /*Newlines=*/0,
674
29
      /*Spaces=*/0);
675
29
}
676
677
void BreakableBlockComment::adaptStartOfLine(
678
1.47k
    unsigned LineIndex, WhitespaceManager &Whitespaces) const {
679
1.47k
  if (LineIndex == 0) {
680
958
    if (DelimitersOnNewline) {
681
      // Since we're breaking at index 1 below, the break position and the
682
      // break length are the same.
683
      // Note: this works because getCommentSplit is careful never to split at
684
      // the beginning of a line.
685
55
      size_t BreakLength = Lines[0].substr(1).find_first_not_of(Blanks);
686
55
      if (BreakLength != StringRef::npos) {
687
12
        insertBreak(LineIndex, 0, Split(1, BreakLength), /*ContentIndent=*/0,
688
12
                    Whitespaces);
689
12
      }
690
55
    }
691
958
    return;
692
958
  }
693
  // Here no reflow with the previous line will happen.
694
  // Fix the decoration of the line at LineIndex.
695
514
  StringRef Prefix = Decoration;
696
514
  if (Content[LineIndex].empty()) {
697
193
    if (LineIndex + 1 == Lines.size()) {
698
176
      if (!LastLineNeedsDecoration) {
699
        // If the last line was empty, we don't need a prefix, as the */ will
700
        // line up with the decoration (if it exists).
701
175
        Prefix = "";
702
175
      }
703
176
    } else 
if (17
!Decoration.empty()17
) {
704
      // For other empty lines, if we do have a decoration, adapt it to not
705
      // contain a trailing whitespace.
706
9
      Prefix = Prefix.substr(0, 1);
707
9
    }
708
321
  } else if (ContentColumn[LineIndex] == 1) {
709
    // This line starts immediately after the decorating *.
710
9
    Prefix = Prefix.substr(0, 1);
711
9
  }
712
  // This is the offset of the end of the last line relative to the start of the
713
  // token text in the token.
714
514
  unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() +
715
514
                                     Content[LineIndex - 1].size() -
716
514
                                     tokenAt(LineIndex).TokenText.data();
717
514
  unsigned WhitespaceLength = Content[LineIndex].data() -
718
514
                              tokenAt(LineIndex).TokenText.data() -
719
514
                              WhitespaceOffsetInToken;
720
514
  Whitespaces.replaceWhitespaceInToken(
721
514
      tokenAt(LineIndex), WhitespaceOffsetInToken, WhitespaceLength, "", Prefix,
722
514
      InPPDirective, /*Newlines=*/1, ContentColumn[LineIndex] - Prefix.size());
723
514
}
724
725
BreakableToken::Split
726
2.70k
BreakableBlockComment::getSplitAfterLastLine(unsigned TailOffset) const {
727
2.70k
  if (DelimitersOnNewline) {
728
    // Replace the trailing whitespace of the last line with a newline.
729
    // In case the last line is empty, the ending '*/' is already on its own
730
    // line.
731
125
    StringRef Line = Content.back().substr(TailOffset);
732
125
    StringRef TrimmedLine = Line.rtrim(Blanks);
733
125
    if (!TrimmedLine.empty())
734
12
      return Split(TrimmedLine.size(), Line.size() - TrimmedLine.size());
735
125
  }
736
2.68k
  return Split(StringRef::npos, 0);
737
2.70k
}
738
739
bool BreakableBlockComment::mayReflow(
740
208
    unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
741
  // Content[LineIndex] may exclude the indent after the '*' decoration. In that
742
  // case, we compute the start of the comment pragma manually.
743
208
  StringRef IndentContent = Content[LineIndex];
744
208
  if (Lines[LineIndex].ltrim(Blanks).startswith("*"))
745
91
    IndentContent = Lines[LineIndex].ltrim(Blanks).substr(1);
746
208
  return LineIndex > 0 && !CommentPragmasRegex.match(IndentContent) &&
747
208
         
mayReflowContent(Content[LineIndex])206
&&
!Tok.Finalized89
&&
748
208
         
!switchesFormatting(tokenAt(LineIndex))89
;
749
208
}
750
751
BreakableLineCommentSection::BreakableLineCommentSection(
752
    const FormatToken &Token, unsigned StartColumn, bool InPPDirective,
753
    encoding::Encoding Encoding, const FormatStyle &Style)
754
13.8k
    : BreakableComment(Token, StartColumn, InPPDirective, Encoding, Style) {
755
13.8k
  assert(Tok.is(TT_LineComment) &&
756
13.8k
         "line comment section must start with a line comment");
757
0
  FormatToken *LineTok = nullptr;
758
13.8k
  const int Minimum = Style.SpacesInLineCommentPrefix.Minimum;
759
  // How many spaces we changed in the first line of the section, this will be
760
  // applied in all following lines
761
13.8k
  int FirstLineSpaceChange = 0;
762
13.8k
  for (const FormatToken *CurrentTok = &Tok;
763
20.3k
       CurrentTok && 
CurrentTok->is(TT_LineComment)15.5k
;
764
15.5k
       
CurrentTok = CurrentTok->Next6.55k
) {
765
15.5k
    LastLineTok = LineTok;
766
15.5k
    StringRef TokenText(CurrentTok->TokenText);
767
15.5k
    assert((TokenText.startswith("//") || TokenText.startswith("#")) &&
768
15.5k
           "unsupported line comment prefix, '//' and '#' are supported");
769
0
    size_t FirstLineIndex = Lines.size();
770
15.5k
    TokenText.split(Lines, "\n");
771
15.5k
    Content.resize(Lines.size());
772
15.5k
    ContentColumn.resize(Lines.size());
773
15.5k
    PrefixSpaceChange.resize(Lines.size());
774
15.5k
    Tokens.resize(Lines.size());
775
15.5k
    Prefix.resize(Lines.size());
776
15.5k
    OriginalPrefix.resize(Lines.size());
777
31.1k
    for (size_t i = FirstLineIndex, e = Lines.size(); i < e; 
++i15.6k
) {
778
15.6k
      Lines[i] = Lines[i].ltrim(Blanks);
779
15.6k
      StringRef IndentPrefix = getLineCommentIndentPrefix(Lines[i], Style);
780
15.6k
      OriginalPrefix[i] = IndentPrefix;
781
15.6k
      const int SpacesInPrefix = llvm::count(IndentPrefix, ' ');
782
783
      // This lambda also considers multibyte character that is not handled in
784
      // functions like isPunctuation provided by CharInfo.
785
15.6k
      const auto NoSpaceBeforeFirstCommentChar = [&]() {
786
13.7k
        assert(Lines[i].size() > IndentPrefix.size());
787
0
        const char FirstCommentChar = Lines[i][IndentPrefix.size()];
788
13.7k
        const unsigned FirstCharByteSize =
789
13.7k
            encoding::getCodePointNumBytes(FirstCommentChar, Encoding);
790
13.7k
        if (encoding::columnWidth(
791
13.7k
                Lines[i].substr(IndentPrefix.size(), FirstCharByteSize),
792
13.7k
                Encoding) != 1) {
793
8
          return false;
794
8
        }
795
        // In C-like comments, add a space before #. For example this is useful
796
        // to preserve the relative indentation when commenting out code with
797
        // #includes.
798
        //
799
        // In languages using # as the comment leader such as proto, don't
800
        // add a space to support patterns like:
801
        // #########
802
        // # section
803
        // #########
804
13.7k
        if (FirstCommentChar == '#' && 
!TokenText.startswith("#")20
)
805
12
          return false;
806
13.7k
        return FirstCommentChar == '\\' || 
isPunctuation(FirstCommentChar)13.7k
||
807
13.7k
               
isHorizontalWhitespace(FirstCommentChar)13.2k
;
808
13.7k
      };
809
810
      // On the first line of the comment section we calculate how many spaces
811
      // are to be added or removed, all lines after that just get only the
812
      // change and we will not look at the maximum anymore. Additionally to the
813
      // actual first line, we calculate that when the non space Prefix changes,
814
      // e.g. from "///" to "//".
815
15.6k
      if (i == 0 || OriginalPrefix[i].rtrim(Blanks) !=
816
13.8k
                        OriginalPrefix[i - 1].rtrim(Blanks)) {
817
13.8k
        if (SpacesInPrefix < Minimum && 
Lines[i].size() > IndentPrefix.size()2.31k
&&
818
13.8k
            
!NoSpaceBeforeFirstCommentChar()250
) {
819
176
          FirstLineSpaceChange = Minimum - SpacesInPrefix;
820
13.6k
        } else if (static_cast<unsigned>(SpacesInPrefix) >
821
13.6k
                   Style.SpacesInLineCommentPrefix.Maximum) {
822
80
          FirstLineSpaceChange =
823
80
              Style.SpacesInLineCommentPrefix.Maximum - SpacesInPrefix;
824
13.5k
        } else {
825
13.5k
          FirstLineSpaceChange = 0;
826
13.5k
        }
827
13.8k
      }
828
829
15.6k
      if (Lines[i].size() != IndentPrefix.size()) {
830
13.5k
        PrefixSpaceChange[i] = FirstLineSpaceChange;
831
832
13.5k
        if (SpacesInPrefix + PrefixSpaceChange[i] < Minimum) {
833
102
          PrefixSpaceChange[i] +=
834
102
              Minimum - (SpacesInPrefix + PrefixSpaceChange[i]);
835
102
        }
836
837
13.5k
        assert(Lines[i].size() > IndentPrefix.size());
838
0
        const auto FirstNonSpace = Lines[i][IndentPrefix.size()];
839
13.5k
        const bool IsFormatComment = LineTok && 
switchesFormatting(*LineTok)1.78k
;
840
13.5k
        const bool LineRequiresLeadingSpace =
841
13.5k
            !NoSpaceBeforeFirstCommentChar() ||
842
13.5k
            
(461
FirstNonSpace == '}'461
&&
FirstLineSpaceChange != 048
);
843
13.5k
        const bool AllowsSpaceChange =
844
13.5k
            !IsFormatComment &&
845
13.5k
            
(13.5k
SpacesInPrefix != 013.5k
||
LineRequiresLeadingSpace355
);
846
847
13.5k
        if (PrefixSpaceChange[i] > 0 && 
AllowsSpaceChange362
) {
848
272
          Prefix[i] = IndentPrefix.str();
849
272
          Prefix[i].append(PrefixSpaceChange[i], ' ');
850
13.2k
        } else if (PrefixSpaceChange[i] < 0 && 
AllowsSpaceChange136
) {
851
136
          Prefix[i] = IndentPrefix
852
136
                          .drop_back(std::min<std::size_t>(
853
136
                              -PrefixSpaceChange[i], SpacesInPrefix))
854
136
                          .str();
855
13.1k
        } else {
856
13.1k
          Prefix[i] = IndentPrefix.str();
857
13.1k
        }
858
13.5k
      } else {
859
        // If the IndentPrefix is the whole line, there is no content and we
860
        // drop just all space
861
2.07k
        Prefix[i] = IndentPrefix.drop_back(SpacesInPrefix).str();
862
2.07k
      }
863
864
0
      Tokens[i] = LineTok;
865
15.6k
      Content[i] = Lines[i].substr(IndentPrefix.size());
866
15.6k
      ContentColumn[i] =
867
15.6k
          StartColumn + encoding::columnWidthWithTabs(Prefix[i], StartColumn,
868
15.6k
                                                      Style.TabWidth, Encoding);
869
870
      // Calculate the end of the non-whitespace text in this line.
871
15.6k
      size_t EndOfLine = Content[i].find_last_not_of(Blanks);
872
15.6k
      if (EndOfLine == StringRef::npos)
873
2.07k
        EndOfLine = Content[i].size();
874
13.5k
      else
875
13.5k
        ++EndOfLine;
876
15.6k
      Content[i] = Content[i].substr(0, EndOfLine);
877
15.6k
    }
878
15.5k
    LineTok = CurrentTok->Next;
879
15.5k
    if (CurrentTok->Next && 
!CurrentTok->Next->ContinuesLineCommentSection10.8k
) {
880
      // A line comment section needs to broken by a line comment that is
881
      // preceded by at least two newlines. Note that we put this break here
882
      // instead of breaking at a previous stage during parsing, since that
883
      // would split the contents of the enum into two unwrapped lines in this
884
      // example, which is undesirable:
885
      // enum A {
886
      //   a, // comment about a
887
      //
888
      //   // comment about b
889
      //   b
890
      // };
891
      //
892
      // FIXME: Consider putting separate line comment sections as children to
893
      // the unwrapped line instead.
894
9.03k
      break;
895
9.03k
    }
896
15.5k
  }
897
13.8k
}
898
899
unsigned
900
BreakableLineCommentSection::getRangeLength(unsigned LineIndex, unsigned Offset,
901
                                            StringRef::size_type Length,
902
19.4k
                                            unsigned StartColumn) const {
903
19.4k
  return encoding::columnWidthWithTabs(
904
19.4k
      Content[LineIndex].substr(Offset, Length), StartColumn, Style.TabWidth,
905
19.4k
      Encoding);
906
19.4k
}
907
908
unsigned
909
BreakableLineCommentSection::getContentStartColumn(unsigned LineIndex,
910
16.2k
                                                   bool /*Break*/) const {
911
16.2k
  return ContentColumn[LineIndex];
912
16.2k
}
913
914
void BreakableLineCommentSection::insertBreak(
915
    unsigned LineIndex, unsigned TailOffset, Split Split,
916
477
    unsigned ContentIndent, WhitespaceManager &Whitespaces) const {
917
477
  StringRef Text = Content[LineIndex].substr(TailOffset);
918
  // Compute the offset of the split relative to the beginning of the token
919
  // text.
920
477
  unsigned BreakOffsetInToken =
921
477
      Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
922
477
  unsigned CharsToRemove = Split.second;
923
477
  Whitespaces.replaceWhitespaceInToken(
924
477
      tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "",
925
477
      Prefix[LineIndex], InPPDirective, /*Newlines=*/1,
926
477
      /*Spaces=*/ContentColumn[LineIndex] - Prefix[LineIndex].size());
927
477
}
928
929
BreakableComment::Split BreakableLineCommentSection::getReflowSplit(
930
607
    unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
931
607
  if (!mayReflow(LineIndex, CommentPragmasRegex))
932
92
    return Split(StringRef::npos, 0);
933
934
515
  size_t Trimmed = Content[LineIndex].find_first_not_of(Blanks);
935
936
  // In a line comment section each line is a separate token; thus, after a
937
  // split we replace all whitespace before the current line comment token
938
  // (which does not need to be included in the split), plus the start of the
939
  // line up to where the content starts.
940
515
  return Split(0, Trimmed != StringRef::npos ? Trimmed : 
00
);
941
607
}
942
943
void BreakableLineCommentSection::reflow(unsigned LineIndex,
944
222
                                         WhitespaceManager &Whitespaces) const {
945
222
  if (LineIndex > 0 && Tokens[LineIndex] != Tokens[LineIndex - 1]) {
946
    // Reflow happens between tokens. Replace the whitespace between the
947
    // tokens by the empty string.
948
221
    Whitespaces.replaceWhitespace(
949
221
        *Tokens[LineIndex], /*Newlines=*/0, /*Spaces=*/0,
950
221
        /*StartOfTokenColumn=*/StartColumn, /*IsAligned=*/true,
951
221
        /*InPPDirective=*/false);
952
221
  } else 
if (1
LineIndex > 01
) {
953
    // In case we're reflowing after the '\' in:
954
    //
955
    //   // line comment \
956
    //   // line 2
957
    //
958
    // the reflow happens inside the single comment token (it is a single line
959
    // comment with an unescaped newline).
960
    // Replace the whitespace between the '\' and '//' with the empty string.
961
    //
962
    // Offset points to after the '\' relative to start of the token.
963
1
    unsigned Offset = Lines[LineIndex - 1].data() +
964
1
                      Lines[LineIndex - 1].size() -
965
1
                      tokenAt(LineIndex - 1).TokenText.data();
966
    // WhitespaceLength is the number of chars between the '\' and the '//' on
967
    // the next line.
968
1
    unsigned WhitespaceLength =
969
1
        Lines[LineIndex].data() - tokenAt(LineIndex).TokenText.data() - Offset;
970
1
    Whitespaces.replaceWhitespaceInToken(*Tokens[LineIndex], Offset,
971
1
                                         /*ReplaceChars=*/WhitespaceLength,
972
1
                                         /*PreviousPostfix=*/"",
973
1
                                         /*CurrentPrefix=*/"",
974
1
                                         /*InPPDirective=*/false,
975
1
                                         /*Newlines=*/0,
976
1
                                         /*Spaces=*/0);
977
1
  }
978
  // Replace the indent and prefix of the token with the reflow prefix.
979
222
  unsigned Offset =
980
222
      Lines[LineIndex].data() - tokenAt(LineIndex).TokenText.data();
981
222
  unsigned WhitespaceLength =
982
222
      Content[LineIndex].data() - Lines[LineIndex].data();
983
222
  Whitespaces.replaceWhitespaceInToken(*Tokens[LineIndex], Offset,
984
222
                                       /*ReplaceChars=*/WhitespaceLength,
985
222
                                       /*PreviousPostfix=*/"",
986
222
                                       /*CurrentPrefix=*/ReflowPrefix,
987
222
                                       /*InPPDirective=*/false,
988
222
                                       /*Newlines=*/0,
989
222
                                       /*Spaces=*/0);
990
222
}
991
992
void BreakableLineCommentSection::adaptStartOfLine(
993
3.61k
    unsigned LineIndex, WhitespaceManager &Whitespaces) const {
994
  // If this is the first line of a token, we need to inform Whitespace Manager
995
  // about it: either adapt the whitespace range preceding it, or mark it as an
996
  // untouchable token.
997
  // This happens for instance here:
998
  // // line 1 \
999
  // // line 2
1000
3.61k
  if (LineIndex > 0 && 
Tokens[LineIndex] != Tokens[LineIndex - 1]611
) {
1001
    // This is the first line for the current token, but no reflow with the
1002
    // previous token is necessary. However, we still may need to adjust the
1003
    // start column. Note that ContentColumn[LineIndex] is the expected
1004
    // content column after a possible update to the prefix, hence the prefix
1005
    // length change is included.
1006
608
    unsigned LineColumn =
1007
608
        ContentColumn[LineIndex] -
1008
608
        (Content[LineIndex].data() - Lines[LineIndex].data()) +
1009
608
        (OriginalPrefix[LineIndex].size() - Prefix[LineIndex].size());
1010
1011
    // We always want to create a replacement instead of adding an untouchable
1012
    // token, even if LineColumn is the same as the original column of the
1013
    // token. This is because WhitespaceManager doesn't align trailing
1014
    // comments if they are untouchable.
1015
608
    Whitespaces.replaceWhitespace(*Tokens[LineIndex],
1016
608
                                  /*Newlines=*/1,
1017
608
                                  /*Spaces=*/LineColumn,
1018
608
                                  /*StartOfTokenColumn=*/LineColumn,
1019
608
                                  /*IsAligned=*/true,
1020
608
                                  /*InPPDirective=*/false);
1021
608
  }
1022
3.61k
  if (OriginalPrefix[LineIndex] != Prefix[LineIndex]) {
1023
    // Adjust the prefix if necessary.
1024
185
    const auto SpacesToRemove = -std::min(PrefixSpaceChange[LineIndex], 0);
1025
185
    const auto SpacesToAdd = std::max(PrefixSpaceChange[LineIndex], 0);
1026
185
    Whitespaces.replaceWhitespaceInToken(
1027
185
        tokenAt(LineIndex), OriginalPrefix[LineIndex].size() - SpacesToRemove,
1028
185
        /*ReplaceChars=*/SpacesToRemove, "", "", /*InPPDirective=*/false,
1029
185
        /*Newlines=*/0, /*Spaces=*/SpacesToAdd);
1030
185
  }
1031
3.61k
}
1032
1033
13.8k
void BreakableLineCommentSection::updateNextToken(LineState &State) const {
1034
13.8k
  if (LastLineTok)
1035
744
    State.NextToken = LastLineTok->Next;
1036
13.8k
}
1037
1038
bool BreakableLineCommentSection::mayReflow(
1039
607
    unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
1040
  // Line comments have the indent as part of the prefix, so we need to
1041
  // recompute the start of the line.
1042
607
  StringRef IndentContent = Content[LineIndex];
1043
607
  if (Lines[LineIndex].startswith("//"))
1044
597
    IndentContent = Lines[LineIndex].substr(2);
1045
  // FIXME: Decide whether we want to reflow non-regular indents:
1046
  // Currently, we only reflow when the OriginalPrefix[LineIndex] matches the
1047
  // OriginalPrefix[LineIndex-1]. That means we don't reflow
1048
  // // text that protrudes
1049
  // //    into text with different indent
1050
  // We do reflow in that case in block comments.
1051
607
  return LineIndex > 0 && !CommentPragmasRegex.match(IndentContent) &&
1052
607
         mayReflowContent(Content[LineIndex]) && 
!Tok.Finalized571
&&
1053
607
         
!switchesFormatting(tokenAt(LineIndex))571
&&
1054
607
         
OriginalPrefix[LineIndex] == OriginalPrefix[LineIndex - 1]571
;
1055
607
}
1056
1057
} // namespace format
1058
} // namespace clang