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";
IsBlank(char C)29 static bool IsBlank(char C) {
30 switch (C) {
31 case ' ':
32 case '\t':
33 case '\v':
34 case '\f':
35 case '\r':
36 return true;
37 default:
38 return false;
39 }
40 }
41
getLineCommentIndentPrefix(StringRef Comment,const FormatStyle & Style)42 static StringRef getLineCommentIndentPrefix(StringRef Comment,
43 const FormatStyle &Style) {
44 static constexpr StringRef KnownCStylePrefixes[] = {"///<", "//!<", "///",
45 "//!", "//:", "//"};
46 static constexpr StringRef KnownTextProtoPrefixes[] = {"####", "###", "##",
47 "//", "#"};
48 ArrayRef<StringRef> KnownPrefixes(KnownCStylePrefixes);
49 if (Style.Language == FormatStyle::LK_TextProto)
50 KnownPrefixes = KnownTextProtoPrefixes;
51
52 assert(std::is_sorted(KnownPrefixes.begin(), KnownPrefixes.end(),
53 [](StringRef Lhs, StringRef Rhs) noexcept {
54 return Lhs.size() > Rhs.size();
55 }));
56
57 for (StringRef KnownPrefix : KnownPrefixes) {
58 if (Comment.startswith(KnownPrefix)) {
59 const auto PrefixLength =
60 Comment.find_first_not_of(' ', KnownPrefix.size());
61 return Comment.substr(0, PrefixLength);
62 }
63 }
64 return {};
65 }
66
67 static BreakableToken::Split
getCommentSplit(StringRef Text,unsigned ContentStartColumn,unsigned ColumnLimit,unsigned TabWidth,encoding::Encoding Encoding,const FormatStyle & Style,bool DecorationEndsWithStar=false)68 getCommentSplit(StringRef Text, unsigned ContentStartColumn,
69 unsigned ColumnLimit, unsigned TabWidth,
70 encoding::Encoding Encoding, const FormatStyle &Style,
71 bool DecorationEndsWithStar = false) {
72 LLVM_DEBUG(llvm::dbgs() << "Comment split: \"" << Text
73 << "\", Column limit: " << ColumnLimit
74 << ", Content start: " << ContentStartColumn << "\n");
75 if (ColumnLimit <= ContentStartColumn + 1)
76 return BreakableToken::Split(StringRef::npos, 0);
77
78 unsigned MaxSplit = ColumnLimit - ContentStartColumn + 1;
79 unsigned MaxSplitBytes = 0;
80
81 for (unsigned NumChars = 0;
82 NumChars < MaxSplit && MaxSplitBytes < Text.size();) {
83 unsigned BytesInChar =
84 encoding::getCodePointNumBytes(Text[MaxSplitBytes], Encoding);
85 NumChars +=
86 encoding::columnWidthWithTabs(Text.substr(MaxSplitBytes, BytesInChar),
87 ContentStartColumn, TabWidth, Encoding);
88 MaxSplitBytes += BytesInChar;
89 }
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 if (Style.Language == FormatStyle::LK_JavaScript) {
95 StringRef::size_type SpaceOffset =
96 Text.find_first_of(Blanks, MaxSplitBytes);
97 if (SpaceOffset != StringRef::npos && SpaceOffset + 1 < Text.size() &&
98 Text[SpaceOffset + 1] == '{') {
99 MaxSplitBytes = SpaceOffset + 1;
100 }
101 }
102
103 StringRef::size_type SpaceOffset = Text.find_last_of(Blanks, MaxSplitBytes);
104
105 static const auto kNumberedListRegexp = llvm::Regex("^[1-9][0-9]?\\.");
106 // Some spaces are unacceptable to break on, rewind past them.
107 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 if (Style.isCpp()) {
114 StringRef::size_type LastNonBlank =
115 Text.find_last_not_of(Blanks, SpaceOffset);
116 if (LastNonBlank != StringRef::npos && Text[LastNonBlank] == '\\') {
117 SpaceOffset = Text.find_last_of(Blanks, LastNonBlank);
118 continue;
119 }
120 }
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 if (kNumberedListRegexp.match(Text.substr(SpaceOffset).ltrim(Blanks))) {
125 SpaceOffset = Text.find_last_of(Blanks, SpaceOffset);
126 continue;
127 }
128
129 // Avoid ever breaking before a @tag or a { in JavaScript.
130 if (Style.Language == FormatStyle::LK_JavaScript &&
131 SpaceOffset + 1 < Text.size() &&
132 (Text[SpaceOffset + 1] == '{' || Text[SpaceOffset + 1] == '@')) {
133 SpaceOffset = Text.find_last_of(Blanks, SpaceOffset);
134 continue;
135 }
136
137 break;
138 }
139
140 if (SpaceOffset == StringRef::npos ||
141 // Don't break at leading whitespace.
142 Text.find_last_not_of(Blanks, SpaceOffset) == StringRef::npos) {
143 // Make sure that we don't break at leading whitespace that
144 // reaches past MaxSplit.
145 StringRef::size_type FirstNonWhitespace = Text.find_first_not_of(Blanks);
146 if (FirstNonWhitespace == StringRef::npos)
147 // If the comment is only whitespace, we cannot split.
148 return BreakableToken::Split(StringRef::npos, 0);
149 SpaceOffset = Text.find_first_of(
150 Blanks, std::max<unsigned>(MaxSplitBytes, FirstNonWhitespace));
151 }
152 if (SpaceOffset != StringRef::npos && SpaceOffset != 0) {
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 if (SpaceOffset == 1 && Text[SpaceOffset - 1] == '*')
158 return BreakableToken::Split(StringRef::npos, 0);
159 StringRef BeforeCut = Text.substr(0, SpaceOffset).rtrim(Blanks);
160 StringRef AfterCut = Text.substr(SpaceOffset);
161 // Don't trim the leading blanks if it would create a */ after the break.
162 if (!DecorationEndsWithStar || AfterCut.size() <= 1 || AfterCut[1] != '/')
163 AfterCut = AfterCut.ltrim(Blanks);
164 return BreakableToken::Split(BeforeCut.size(),
165 AfterCut.begin() - BeforeCut.end());
166 }
167 return BreakableToken::Split(StringRef::npos, 0);
168 }
169
170 static BreakableToken::Split
getStringSplit(StringRef Text,unsigned UsedColumns,unsigned ColumnLimit,unsigned TabWidth,encoding::Encoding Encoding)171 getStringSplit(StringRef Text, unsigned UsedColumns, unsigned ColumnLimit,
172 unsigned TabWidth, encoding::Encoding Encoding) {
173 // FIXME: Reduce unit test case.
174 if (Text.empty())
175 return BreakableToken::Split(StringRef::npos, 0);
176 if (ColumnLimit <= UsedColumns)
177 return BreakableToken::Split(StringRef::npos, 0);
178 unsigned MaxSplit = ColumnLimit - UsedColumns;
179 StringRef::size_type SpaceOffset = 0;
180 StringRef::size_type SlashOffset = 0;
181 StringRef::size_type WordStartOffset = 0;
182 StringRef::size_type SplitPoint = 0;
183 for (unsigned Chars = 0;;) {
184 unsigned Advance;
185 if (Text[0] == '\\') {
186 Advance = encoding::getEscapeSequenceLength(Text);
187 Chars += Advance;
188 } else {
189 Advance = encoding::getCodePointNumBytes(Text[0], Encoding);
190 Chars += encoding::columnWidthWithTabs(
191 Text.substr(0, Advance), UsedColumns + Chars, TabWidth, Encoding);
192 }
193
194 if (Chars > MaxSplit || Text.size() <= Advance)
195 break;
196
197 if (IsBlank(Text[0]))
198 SpaceOffset = SplitPoint;
199 if (Text[0] == '/')
200 SlashOffset = SplitPoint;
201 if (Advance == 1 && !isAlphanumeric(Text[0]))
202 WordStartOffset = SplitPoint;
203
204 SplitPoint += Advance;
205 Text = Text.substr(Advance);
206 }
207
208 if (SpaceOffset != 0)
209 return BreakableToken::Split(SpaceOffset + 1, 0);
210 if (SlashOffset != 0)
211 return BreakableToken::Split(SlashOffset + 1, 0);
212 if (WordStartOffset != 0)
213 return BreakableToken::Split(WordStartOffset + 1, 0);
214 if (SplitPoint != 0)
215 return BreakableToken::Split(SplitPoint, 0);
216 return BreakableToken::Split(StringRef::npos, 0);
217 }
218
switchesFormatting(const FormatToken & Token)219 bool switchesFormatting(const FormatToken &Token) {
220 assert((Token.is(TT_BlockComment) || Token.is(TT_LineComment)) &&
221 "formatting regions are switched by comment tokens");
222 StringRef Content = Token.TokenText.substr(2).ltrim();
223 return Content.startswith("clang-format on") ||
224 Content.startswith("clang-format off");
225 }
226
227 unsigned
getLengthAfterCompression(unsigned RemainingTokenColumns,Split Split) const228 BreakableToken::getLengthAfterCompression(unsigned RemainingTokenColumns,
229 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 return RemainingTokenColumns + 1 - Split.second;
242 }
243
getLineCount() const244 unsigned BreakableStringLiteral::getLineCount() const { return 1; }
245
getRangeLength(unsigned LineIndex,unsigned Offset,StringRef::size_type Length,unsigned StartColumn) const246 unsigned BreakableStringLiteral::getRangeLength(unsigned LineIndex,
247 unsigned Offset,
248 StringRef::size_type Length,
249 unsigned StartColumn) const {
250 llvm_unreachable("Getting the length of a part of the string literal "
251 "indicates that the code tries to reflow it.");
252 }
253
254 unsigned
getRemainingLength(unsigned LineIndex,unsigned Offset,unsigned StartColumn) const255 BreakableStringLiteral::getRemainingLength(unsigned LineIndex, unsigned Offset,
256 unsigned StartColumn) const {
257 return UnbreakableTailLength + Postfix.size() +
258 encoding::columnWidthWithTabs(Line.substr(Offset, StringRef::npos),
259 StartColumn, Style.TabWidth, Encoding);
260 }
261
getContentStartColumn(unsigned LineIndex,bool Break) const262 unsigned BreakableStringLiteral::getContentStartColumn(unsigned LineIndex,
263 bool Break) const {
264 return StartColumn + Prefix.size();
265 }
266
BreakableStringLiteral(const FormatToken & Tok,unsigned StartColumn,StringRef Prefix,StringRef Postfix,unsigned UnbreakableTailLength,bool InPPDirective,encoding::Encoding Encoding,const FormatStyle & Style)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 UnbreakableTailLength(UnbreakableTailLength) {
274 assert(Tok.TokenText.startswith(Prefix) && Tok.TokenText.endswith(Postfix));
275 Line = Tok.TokenText.substr(
276 Prefix.size(), Tok.TokenText.size() - Prefix.size() - Postfix.size());
277 }
278
getSplit(unsigned LineIndex,unsigned TailOffset,unsigned ColumnLimit,unsigned ContentStartColumn,const llvm::Regex & CommentPragmasRegex) const279 BreakableToken::Split BreakableStringLiteral::getSplit(
280 unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit,
281 unsigned ContentStartColumn, const llvm::Regex &CommentPragmasRegex) const {
282 return getStringSplit(Line.substr(TailOffset), ContentStartColumn,
283 ColumnLimit - Postfix.size(), Style.TabWidth, Encoding);
284 }
285
insertBreak(unsigned LineIndex,unsigned TailOffset,Split Split,unsigned ContentIndent,WhitespaceManager & Whitespaces) const286 void BreakableStringLiteral::insertBreak(unsigned LineIndex,
287 unsigned TailOffset, Split Split,
288 unsigned ContentIndent,
289 WhitespaceManager &Whitespaces) const {
290 Whitespaces.replaceWhitespaceInToken(
291 Tok, Prefix.size() + TailOffset + Split.first, Split.second, Postfix,
292 Prefix, InPPDirective, 1, StartColumn);
293 }
294
BreakableComment(const FormatToken & Token,unsigned StartColumn,bool InPPDirective,encoding::Encoding Encoding,const FormatStyle & Style)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 StartColumn(StartColumn) {}
301
getLineCount() const302 unsigned BreakableComment::getLineCount() const { return Lines.size(); }
303
304 BreakableToken::Split
getSplit(unsigned LineIndex,unsigned TailOffset,unsigned ColumnLimit,unsigned ContentStartColumn,const llvm::Regex & CommentPragmasRegex) const305 BreakableComment::getSplit(unsigned LineIndex, unsigned TailOffset,
306 unsigned ColumnLimit, unsigned ContentStartColumn,
307 const llvm::Regex &CommentPragmasRegex) const {
308 // Don't break lines matching the comment pragmas regex.
309 if (CommentPragmasRegex.match(Content[LineIndex]))
310 return Split(StringRef::npos, 0);
311 return getCommentSplit(Content[LineIndex].substr(TailOffset),
312 ContentStartColumn, ColumnLimit, Style.TabWidth,
313 Encoding, Style);
314 }
315
compressWhitespace(unsigned LineIndex,unsigned TailOffset,Split Split,WhitespaceManager & Whitespaces) const316 void BreakableComment::compressWhitespace(
317 unsigned LineIndex, unsigned TailOffset, Split Split,
318 WhitespaceManager &Whitespaces) const {
319 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 unsigned BreakOffsetInToken =
325 Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
326 unsigned CharsToRemove = Split.second;
327 Whitespaces.replaceWhitespaceInToken(
328 tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "", "",
329 /*InPPDirective=*/false, /*Newlines=*/0, /*Spaces=*/1);
330 }
331
tokenAt(unsigned LineIndex) const332 const FormatToken &BreakableComment::tokenAt(unsigned LineIndex) const {
333 return Tokens[LineIndex] ? *Tokens[LineIndex] : Tok;
334 }
335
mayReflowContent(StringRef Content)336 static bool mayReflowContent(StringRef Content) {
337 Content = Content.trim(Blanks);
338 // Lines starting with '@' commonly have special meaning.
339 // Lines starting with '-', '-#', '+' or '*' are bulleted/numbered lists.
340 bool hasSpecialMeaningPrefix = false;
341 for (StringRef Prefix :
342 {"@", "TODO", "FIXME", "XXX", "-# ", "- ", "+ ", "* "}) {
343 if (Content.startswith(Prefix)) {
344 hasSpecialMeaningPrefix = true;
345 break;
346 }
347 }
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 static const auto kNumberedListRegexp = llvm::Regex("^[1-9][0-9]?\\. ");
353 hasSpecialMeaningPrefix =
354 hasSpecialMeaningPrefix || kNumberedListRegexp.match(Content);
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 return Content.size() >= 2 && !hasSpecialMeaningPrefix &&
360 !Content.endswith("\\") &&
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 (!isPunctuation(Content[0]) || !isPunctuation(Content[1]));
364 }
365
BreakableBlockComment(const FormatToken & Token,unsigned StartColumn,unsigned OriginalStartColumn,bool FirstInLine,bool InPPDirective,encoding::Encoding Encoding,const FormatStyle & Style,bool UseCRLF)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 UnbreakableTailLength(Token.UnbreakableTailLength) {
373 assert(Tok.is(TT_BlockComment) &&
374 "block comment section must start with a block comment");
375
376 StringRef TokenText(Tok.TokenText);
377 assert(TokenText.startswith("/*") && TokenText.endswith("*/"));
378 TokenText.substr(2, TokenText.size() - 4)
379 .split(Lines, UseCRLF ? "\r\n" : "\n");
380
381 int IndentDelta = StartColumn - OriginalStartColumn;
382 Content.resize(Lines.size());
383 Content[0] = Lines[0];
384 ContentColumn.resize(Lines.size());
385 // Account for the initial '/*'.
386 ContentColumn[0] = StartColumn + 2;
387 Tokens.resize(Lines.size());
388 for (size_t i = 1; i < Lines.size(); ++i)
389 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 DecorationColumn = StartColumn + 1;
394
395 // Account for comment decoration patterns like this:
396 //
397 // /*
398 // ** blah blah blah
399 // */
400 if (Lines.size() >= 2 && Content[1].startswith("**") &&
401 static_cast<unsigned>(ContentColumn[1]) == StartColumn) {
402 DecorationColumn = StartColumn;
403 }
404
405 Decoration = "* ";
406 if (Lines.size() == 1 && !FirstInLine) {
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 Decoration = "";
413 }
414 for (size_t i = 1, e = Lines.size(); i < e && !Decoration.empty(); ++i) {
415 // If the last line is empty, the closing "*/" will have a star.
416 if (i + 1 == e && Content[i].empty())
417 break;
418 if (!Content[i].empty() && i + 1 != e && Decoration.startswith(Content[i]))
419 continue;
420 while (!Content[i].startswith(Decoration))
421 Decoration = Decoration.substr(0, Decoration.size() - 1);
422 }
423
424 LastLineNeedsDecoration = true;
425 IndentAtLineBreak = ContentColumn[0] + 1;
426 for (size_t i = 1, e = Lines.size(); i < e; ++i) {
427 if (Content[i].empty()) {
428 if (i + 1 == e) {
429 // Empty last line means that we already have a star as a part of the
430 // trailing */. We also need to preserve whitespace, so that */ is
431 // correctly indented.
432 LastLineNeedsDecoration = false;
433 // Align the star in the last '*/' with the stars on the previous lines.
434 if (e >= 2 && !Decoration.empty()) {
435 ContentColumn[i] = DecorationColumn;
436 }
437 } else if (Decoration.empty()) {
438 // For all other lines, set the start column to 0 if they're empty, so
439 // we do not insert trailing whitespace anywhere.
440 ContentColumn[i] = 0;
441 }
442 continue;
443 }
444
445 // The first line already excludes the star.
446 // The last line excludes the star if LastLineNeedsDecoration is false.
447 // For all other lines, adjust the line to exclude the star and
448 // (optionally) the first whitespace.
449 unsigned DecorationSize = Decoration.startswith(Content[i])
450 ? Content[i].size()
451 : Decoration.size();
452 if (DecorationSize) {
453 ContentColumn[i] = DecorationColumn + DecorationSize;
454 }
455 Content[i] = Content[i].substr(DecorationSize);
456 if (!Decoration.startswith(Content[i]))
457 IndentAtLineBreak =
458 std::min<int>(IndentAtLineBreak, std::max(0, ContentColumn[i]));
459 }
460 IndentAtLineBreak = std::max<unsigned>(IndentAtLineBreak, Decoration.size());
461
462 // Detect a multiline jsdoc comment and set DelimitersOnNewline in that case.
463 if (Style.Language == FormatStyle::LK_JavaScript ||
464 Style.Language == FormatStyle::LK_Java) {
465 if ((Lines[0] == "*" || Lines[0].startswith("* ")) && Lines.size() > 1) {
466 // This is a multiline jsdoc comment.
467 DelimitersOnNewline = true;
468 } else if (Lines[0].startswith("* ") && Lines.size() == 1) {
469 // Detect a long single-line comment, like:
470 // /** long long long */
471 // Below, '2' is the width of '*/'.
472 unsigned EndColumn =
473 ContentColumn[0] +
474 encoding::columnWidthWithTabs(Lines[0], ContentColumn[0],
475 Style.TabWidth, Encoding) +
476 2;
477 DelimitersOnNewline = EndColumn > Style.ColumnLimit;
478 }
479 }
480
481 LLVM_DEBUG({
482 llvm::dbgs() << "IndentAtLineBreak " << IndentAtLineBreak << "\n";
483 llvm::dbgs() << "DelimitersOnNewline " << DelimitersOnNewline << "\n";
484 for (size_t i = 0; i < Lines.size(); ++i) {
485 llvm::dbgs() << i << " |" << Content[i] << "| "
486 << "CC=" << ContentColumn[i] << "| "
487 << "IN=" << (Content[i].data() - Lines[i].data()) << "\n";
488 }
489 });
490 }
491
getSplit(unsigned LineIndex,unsigned TailOffset,unsigned ColumnLimit,unsigned ContentStartColumn,const llvm::Regex & CommentPragmasRegex) const492 BreakableToken::Split BreakableBlockComment::getSplit(
493 unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit,
494 unsigned ContentStartColumn, const llvm::Regex &CommentPragmasRegex) const {
495 // Don't break lines matching the comment pragmas regex.
496 if (CommentPragmasRegex.match(Content[LineIndex]))
497 return Split(StringRef::npos, 0);
498 return getCommentSplit(Content[LineIndex].substr(TailOffset),
499 ContentStartColumn, ColumnLimit, Style.TabWidth,
500 Encoding, Style, Decoration.endswith("*"));
501 }
502
adjustWhitespace(unsigned LineIndex,int IndentDelta)503 void BreakableBlockComment::adjustWhitespace(unsigned LineIndex,
504 int IndentDelta) {
505 // When in a preprocessor directive, the trailing backslash in a block comment
506 // is not needed, but can serve a purpose of uniformity with necessary escaped
507 // newlines outside the comment. In this case we remove it here before
508 // trimming the trailing whitespace. The backslash will be re-added later when
509 // inserting a line break.
510 size_t EndOfPreviousLine = Lines[LineIndex - 1].size();
511 if (InPPDirective && Lines[LineIndex - 1].endswith("\\"))
512 --EndOfPreviousLine;
513
514 // Calculate the end of the non-whitespace text in the previous line.
515 EndOfPreviousLine =
516 Lines[LineIndex - 1].find_last_not_of(Blanks, EndOfPreviousLine);
517 if (EndOfPreviousLine == StringRef::npos)
518 EndOfPreviousLine = 0;
519 else
520 ++EndOfPreviousLine;
521 // Calculate the start of the non-whitespace text in the current line.
522 size_t StartOfLine = Lines[LineIndex].find_first_not_of(Blanks);
523 if (StartOfLine == StringRef::npos)
524 StartOfLine = Lines[LineIndex].size();
525
526 StringRef Whitespace = Lines[LineIndex].substr(0, StartOfLine);
527 // Adjust Lines to only contain relevant text.
528 size_t PreviousContentOffset =
529 Content[LineIndex - 1].data() - Lines[LineIndex - 1].data();
530 Content[LineIndex - 1] = Lines[LineIndex - 1].substr(
531 PreviousContentOffset, EndOfPreviousLine - PreviousContentOffset);
532 Content[LineIndex] = Lines[LineIndex].substr(StartOfLine);
533
534 // Adjust the start column uniformly across all lines.
535 ContentColumn[LineIndex] =
536 encoding::columnWidthWithTabs(Whitespace, 0, Style.TabWidth, Encoding) +
537 IndentDelta;
538 }
539
getRangeLength(unsigned LineIndex,unsigned Offset,StringRef::size_type Length,unsigned StartColumn) const540 unsigned BreakableBlockComment::getRangeLength(unsigned LineIndex,
541 unsigned Offset,
542 StringRef::size_type Length,
543 unsigned StartColumn) const {
544 unsigned LineLength =
545 encoding::columnWidthWithTabs(Content[LineIndex].substr(Offset, Length),
546 StartColumn, Style.TabWidth, Encoding);
547 // FIXME: This should go into getRemainingLength instead, but we currently
548 // break tests when putting it there. Investigate how to fix those tests.
549 // The last line gets a "*/" postfix.
550 if (LineIndex + 1 == Lines.size()) {
551 LineLength += 2;
552 // We never need a decoration when breaking just the trailing "*/" postfix.
553 // Note that checking that Length == 0 is not enough, since Length could
554 // also be StringRef::npos.
555 if (Content[LineIndex].substr(Offset, StringRef::npos).empty()) {
556 LineLength -= Decoration.size();
557 }
558 }
559 return LineLength;
560 }
561
getRemainingLength(unsigned LineIndex,unsigned Offset,unsigned StartColumn) const562 unsigned BreakableBlockComment::getRemainingLength(unsigned LineIndex,
563 unsigned Offset,
564 unsigned StartColumn) const {
565 return UnbreakableTailLength +
566 getRangeLength(LineIndex, Offset, StringRef::npos, StartColumn);
567 }
568
getContentStartColumn(unsigned LineIndex,bool Break) const569 unsigned BreakableBlockComment::getContentStartColumn(unsigned LineIndex,
570 bool Break) const {
571 if (Break)
572 return IndentAtLineBreak;
573 return std::max(0, ContentColumn[LineIndex]);
574 }
575
576 const llvm::StringSet<>
577 BreakableBlockComment::ContentIndentingJavadocAnnotations = {
578 "@param", "@return", "@returns", "@throws", "@type", "@template",
579 "@see", "@deprecated", "@define", "@exports", "@mods", "@private",
580 };
581
getContentIndent(unsigned LineIndex) const582 unsigned BreakableBlockComment::getContentIndent(unsigned LineIndex) const {
583 if (Style.Language != FormatStyle::LK_Java &&
584 Style.Language != FormatStyle::LK_JavaScript)
585 return 0;
586 // The content at LineIndex 0 of a comment like:
587 // /** line 0 */
588 // is "* line 0", so we need to skip over the decoration in that case.
589 StringRef ContentWithNoDecoration = Content[LineIndex];
590 if (LineIndex == 0 && ContentWithNoDecoration.startswith("*")) {
591 ContentWithNoDecoration = ContentWithNoDecoration.substr(1).ltrim(Blanks);
592 }
593 StringRef FirstWord = ContentWithNoDecoration.substr(
594 0, ContentWithNoDecoration.find_first_of(Blanks));
595 if (ContentIndentingJavadocAnnotations.find(FirstWord) !=
596 ContentIndentingJavadocAnnotations.end())
597 return Style.ContinuationIndentWidth;
598 return 0;
599 }
600
insertBreak(unsigned LineIndex,unsigned TailOffset,Split Split,unsigned ContentIndent,WhitespaceManager & Whitespaces) const601 void BreakableBlockComment::insertBreak(unsigned LineIndex, unsigned TailOffset,
602 Split Split, unsigned ContentIndent,
603 WhitespaceManager &Whitespaces) const {
604 StringRef Text = Content[LineIndex].substr(TailOffset);
605 StringRef Prefix = Decoration;
606 // We need this to account for the case when we have a decoration "* " for all
607 // the lines except for the last one, where the star in "*/" acts as a
608 // decoration.
609 unsigned LocalIndentAtLineBreak = IndentAtLineBreak;
610 if (LineIndex + 1 == Lines.size() &&
611 Text.size() == Split.first + Split.second) {
612 // For the last line we need to break before "*/", but not to add "* ".
613 Prefix = "";
614 if (LocalIndentAtLineBreak >= 2)
615 LocalIndentAtLineBreak -= 2;
616 }
617 // The split offset is from the beginning of the line. Convert it to an offset
618 // from the beginning of the token text.
619 unsigned BreakOffsetInToken =
620 Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
621 unsigned CharsToRemove = Split.second;
622 assert(LocalIndentAtLineBreak >= Prefix.size());
623 std::string PrefixWithTrailingIndent = std::string(Prefix);
624 PrefixWithTrailingIndent.append(ContentIndent, ' ');
625 Whitespaces.replaceWhitespaceInToken(
626 tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "",
627 PrefixWithTrailingIndent, InPPDirective, /*Newlines=*/1,
628 /*Spaces=*/LocalIndentAtLineBreak + ContentIndent -
629 PrefixWithTrailingIndent.size());
630 }
631
getReflowSplit(unsigned LineIndex,const llvm::Regex & CommentPragmasRegex) const632 BreakableToken::Split BreakableBlockComment::getReflowSplit(
633 unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
634 if (!mayReflow(LineIndex, CommentPragmasRegex))
635 return Split(StringRef::npos, 0);
636
637 // If we're reflowing into a line with content indent, only reflow the next
638 // line if its starting whitespace matches the content indent.
639 size_t Trimmed = Content[LineIndex].find_first_not_of(Blanks);
640 if (LineIndex) {
641 unsigned PreviousContentIndent = getContentIndent(LineIndex - 1);
642 if (PreviousContentIndent && Trimmed != StringRef::npos &&
643 Trimmed != PreviousContentIndent)
644 return Split(StringRef::npos, 0);
645 }
646
647 return Split(0, Trimmed != StringRef::npos ? Trimmed : 0);
648 }
649
introducesBreakBeforeToken() const650 bool BreakableBlockComment::introducesBreakBeforeToken() const {
651 // A break is introduced when we want delimiters on newline.
652 return DelimitersOnNewline &&
653 Lines[0].substr(1).find_first_not_of(Blanks) != StringRef::npos;
654 }
655
reflow(unsigned LineIndex,WhitespaceManager & Whitespaces) const656 void BreakableBlockComment::reflow(unsigned LineIndex,
657 WhitespaceManager &Whitespaces) const {
658 StringRef TrimmedContent = Content[LineIndex].ltrim(Blanks);
659 // Here we need to reflow.
660 assert(Tokens[LineIndex - 1] == Tokens[LineIndex] &&
661 "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 unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() +
665 Content[LineIndex - 1].size() -
666 tokenAt(LineIndex).TokenText.data();
667 unsigned WhitespaceLength = TrimmedContent.data() -
668 tokenAt(LineIndex).TokenText.data() -
669 WhitespaceOffsetInToken;
670 Whitespaces.replaceWhitespaceInToken(
671 tokenAt(LineIndex), WhitespaceOffsetInToken,
672 /*ReplaceChars=*/WhitespaceLength, /*PreviousPostfix=*/"",
673 /*CurrentPrefix=*/ReflowPrefix, InPPDirective, /*Newlines=*/0,
674 /*Spaces=*/0);
675 }
676
adaptStartOfLine(unsigned LineIndex,WhitespaceManager & Whitespaces) const677 void BreakableBlockComment::adaptStartOfLine(
678 unsigned LineIndex, WhitespaceManager &Whitespaces) const {
679 if (LineIndex == 0) {
680 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 size_t BreakLength = Lines[0].substr(1).find_first_not_of(Blanks);
686 if (BreakLength != StringRef::npos)
687 insertBreak(LineIndex, 0, Split(1, BreakLength), /*ContentIndent=*/0,
688 Whitespaces);
689 }
690 return;
691 }
692 // Here no reflow with the previous line will happen.
693 // Fix the decoration of the line at LineIndex.
694 StringRef Prefix = Decoration;
695 if (Content[LineIndex].empty()) {
696 if (LineIndex + 1 == Lines.size()) {
697 if (!LastLineNeedsDecoration) {
698 // If the last line was empty, we don't need a prefix, as the */ will
699 // line up with the decoration (if it exists).
700 Prefix = "";
701 }
702 } else if (!Decoration.empty()) {
703 // For other empty lines, if we do have a decoration, adapt it to not
704 // contain a trailing whitespace.
705 Prefix = Prefix.substr(0, 1);
706 }
707 } else {
708 if (ContentColumn[LineIndex] == 1) {
709 // This line starts immediately after the decorating *.
710 Prefix = Prefix.substr(0, 1);
711 }
712 }
713 // This is the offset of the end of the last line relative to the start of the
714 // token text in the token.
715 unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() +
716 Content[LineIndex - 1].size() -
717 tokenAt(LineIndex).TokenText.data();
718 unsigned WhitespaceLength = Content[LineIndex].data() -
719 tokenAt(LineIndex).TokenText.data() -
720 WhitespaceOffsetInToken;
721 Whitespaces.replaceWhitespaceInToken(
722 tokenAt(LineIndex), WhitespaceOffsetInToken, WhitespaceLength, "", Prefix,
723 InPPDirective, /*Newlines=*/1, ContentColumn[LineIndex] - Prefix.size());
724 }
725
726 BreakableToken::Split
getSplitAfterLastLine(unsigned TailOffset) const727 BreakableBlockComment::getSplitAfterLastLine(unsigned TailOffset) const {
728 if (DelimitersOnNewline) {
729 // Replace the trailing whitespace of the last line with a newline.
730 // In case the last line is empty, the ending '*/' is already on its own
731 // line.
732 StringRef Line = Content.back().substr(TailOffset);
733 StringRef TrimmedLine = Line.rtrim(Blanks);
734 if (!TrimmedLine.empty())
735 return Split(TrimmedLine.size(), Line.size() - TrimmedLine.size());
736 }
737 return Split(StringRef::npos, 0);
738 }
739
mayReflow(unsigned LineIndex,const llvm::Regex & CommentPragmasRegex) const740 bool BreakableBlockComment::mayReflow(
741 unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
742 // Content[LineIndex] may exclude the indent after the '*' decoration. In that
743 // case, we compute the start of the comment pragma manually.
744 StringRef IndentContent = Content[LineIndex];
745 if (Lines[LineIndex].ltrim(Blanks).startswith("*")) {
746 IndentContent = Lines[LineIndex].ltrim(Blanks).substr(1);
747 }
748 return LineIndex > 0 && !CommentPragmasRegex.match(IndentContent) &&
749 mayReflowContent(Content[LineIndex]) && !Tok.Finalized &&
750 !switchesFormatting(tokenAt(LineIndex));
751 }
752
BreakableLineCommentSection(const FormatToken & Token,unsigned StartColumn,bool InPPDirective,encoding::Encoding Encoding,const FormatStyle & Style)753 BreakableLineCommentSection::BreakableLineCommentSection(
754 const FormatToken &Token, unsigned StartColumn, bool InPPDirective,
755 encoding::Encoding Encoding, const FormatStyle &Style)
756 : BreakableComment(Token, StartColumn, InPPDirective, Encoding, Style) {
757 assert(Tok.is(TT_LineComment) &&
758 "line comment section must start with a line comment");
759 FormatToken *LineTok = nullptr;
760 // How many spaces we changed in the first line of the section, this will be
761 // applied in all following lines
762 int FirstLineSpaceChange = 0;
763 for (const FormatToken *CurrentTok = &Tok;
764 CurrentTok && CurrentTok->is(TT_LineComment);
765 CurrentTok = CurrentTok->Next) {
766 LastLineTok = LineTok;
767 StringRef TokenText(CurrentTok->TokenText);
768 assert((TokenText.startswith("//") || TokenText.startswith("#")) &&
769 "unsupported line comment prefix, '//' and '#' are supported");
770 size_t FirstLineIndex = Lines.size();
771 TokenText.split(Lines, "\n");
772 Content.resize(Lines.size());
773 ContentColumn.resize(Lines.size());
774 PrefixSpaceChange.resize(Lines.size());
775 Tokens.resize(Lines.size());
776 Prefix.resize(Lines.size());
777 OriginalPrefix.resize(Lines.size());
778 for (size_t i = FirstLineIndex, e = Lines.size(); i < e; ++i) {
779 Lines[i] = Lines[i].ltrim(Blanks);
780 StringRef IndentPrefix = getLineCommentIndentPrefix(Lines[i], Style);
781 OriginalPrefix[i] = IndentPrefix;
782 const unsigned SpacesInPrefix =
783 std::count(IndentPrefix.begin(), IndentPrefix.end(), ' ');
784
785 // On the first line of the comment section we calculate how many spaces
786 // are to be added or removed, all lines after that just get only the
787 // change and we will not look at the maximum anymore. Additionally to the
788 // actual first line, we calculate that when the non space Prefix changes,
789 // e.g. from "///" to "//".
790 if (i == 0 || OriginalPrefix[i].rtrim(Blanks) !=
791 OriginalPrefix[i - 1].rtrim(Blanks)) {
792 if (SpacesInPrefix < Style.SpacesInLineCommentPrefix.Minimum &&
793 Lines[i].size() > IndentPrefix.size() &&
794 isAlphanumeric(Lines[i][IndentPrefix.size()])) {
795 FirstLineSpaceChange =
796 Style.SpacesInLineCommentPrefix.Minimum - SpacesInPrefix;
797 } else if (SpacesInPrefix > Style.SpacesInLineCommentPrefix.Maximum) {
798 FirstLineSpaceChange =
799 Style.SpacesInLineCommentPrefix.Maximum - SpacesInPrefix;
800 } else {
801 FirstLineSpaceChange = 0;
802 }
803 }
804
805 if (Lines[i].size() != IndentPrefix.size()) {
806 PrefixSpaceChange[i] = FirstLineSpaceChange;
807
808 if (SpacesInPrefix + PrefixSpaceChange[i] <
809 Style.SpacesInLineCommentPrefix.Minimum) {
810 PrefixSpaceChange[i] += Style.SpacesInLineCommentPrefix.Minimum -
811 (SpacesInPrefix + PrefixSpaceChange[i]);
812 }
813
814 assert(Lines[i].size() > IndentPrefix.size());
815 const auto FirstNonSpace = Lines[i][IndentPrefix.size()];
816 const auto AllowsSpaceChange =
817 SpacesInPrefix != 0 ||
818 (isAlphanumeric(FirstNonSpace) ||
819 (FirstNonSpace == '}' && FirstLineSpaceChange != 0));
820
821 if (PrefixSpaceChange[i] > 0 && AllowsSpaceChange) {
822 Prefix[i] = IndentPrefix.str();
823 Prefix[i].append(PrefixSpaceChange[i], ' ');
824 } else if (PrefixSpaceChange[i] < 0 && AllowsSpaceChange) {
825 Prefix[i] = IndentPrefix
826 .drop_back(std::min<std::size_t>(
827 -PrefixSpaceChange[i], SpacesInPrefix))
828 .str();
829 } else {
830 Prefix[i] = IndentPrefix.str();
831 }
832 } else {
833 // If the IndentPrefix is the whole line, there is no content and we
834 // drop just all space
835 Prefix[i] = IndentPrefix.drop_back(SpacesInPrefix).str();
836 }
837
838 Tokens[i] = LineTok;
839 Content[i] = Lines[i].substr(IndentPrefix.size());
840 ContentColumn[i] =
841 StartColumn + encoding::columnWidthWithTabs(Prefix[i], StartColumn,
842 Style.TabWidth, Encoding);
843
844 // Calculate the end of the non-whitespace text in this line.
845 size_t EndOfLine = Content[i].find_last_not_of(Blanks);
846 if (EndOfLine == StringRef::npos)
847 EndOfLine = Content[i].size();
848 else
849 ++EndOfLine;
850 Content[i] = Content[i].substr(0, EndOfLine);
851 }
852 LineTok = CurrentTok->Next;
853 if (CurrentTok->Next && !CurrentTok->Next->ContinuesLineCommentSection) {
854 // A line comment section needs to broken by a line comment that is
855 // preceded by at least two newlines. Note that we put this break here
856 // instead of breaking at a previous stage during parsing, since that
857 // would split the contents of the enum into two unwrapped lines in this
858 // example, which is undesirable:
859 // enum A {
860 // a, // comment about a
861 //
862 // // comment about b
863 // b
864 // };
865 //
866 // FIXME: Consider putting separate line comment sections as children to
867 // the unwrapped line instead.
868 break;
869 }
870 }
871 }
872
873 unsigned
getRangeLength(unsigned LineIndex,unsigned Offset,StringRef::size_type Length,unsigned StartColumn) const874 BreakableLineCommentSection::getRangeLength(unsigned LineIndex, unsigned Offset,
875 StringRef::size_type Length,
876 unsigned StartColumn) const {
877 return encoding::columnWidthWithTabs(
878 Content[LineIndex].substr(Offset, Length), StartColumn, Style.TabWidth,
879 Encoding);
880 }
881
882 unsigned
getContentStartColumn(unsigned LineIndex,bool) const883 BreakableLineCommentSection::getContentStartColumn(unsigned LineIndex,
884 bool /*Break*/) const {
885 return ContentColumn[LineIndex];
886 }
887
insertBreak(unsigned LineIndex,unsigned TailOffset,Split Split,unsigned ContentIndent,WhitespaceManager & Whitespaces) const888 void BreakableLineCommentSection::insertBreak(
889 unsigned LineIndex, unsigned TailOffset, Split Split,
890 unsigned ContentIndent, WhitespaceManager &Whitespaces) const {
891 StringRef Text = Content[LineIndex].substr(TailOffset);
892 // Compute the offset of the split relative to the beginning of the token
893 // text.
894 unsigned BreakOffsetInToken =
895 Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
896 unsigned CharsToRemove = Split.second;
897 Whitespaces.replaceWhitespaceInToken(
898 tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "",
899 Prefix[LineIndex], InPPDirective, /*Newlines=*/1,
900 /*Spaces=*/ContentColumn[LineIndex] - Prefix[LineIndex].size());
901 }
902
getReflowSplit(unsigned LineIndex,const llvm::Regex & CommentPragmasRegex) const903 BreakableComment::Split BreakableLineCommentSection::getReflowSplit(
904 unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
905 if (!mayReflow(LineIndex, CommentPragmasRegex))
906 return Split(StringRef::npos, 0);
907
908 size_t Trimmed = Content[LineIndex].find_first_not_of(Blanks);
909
910 // In a line comment section each line is a separate token; thus, after a
911 // split we replace all whitespace before the current line comment token
912 // (which does not need to be included in the split), plus the start of the
913 // line up to where the content starts.
914 return Split(0, Trimmed != StringRef::npos ? Trimmed : 0);
915 }
916
reflow(unsigned LineIndex,WhitespaceManager & Whitespaces) const917 void BreakableLineCommentSection::reflow(unsigned LineIndex,
918 WhitespaceManager &Whitespaces) const {
919 if (LineIndex > 0 && Tokens[LineIndex] != Tokens[LineIndex - 1]) {
920 // Reflow happens between tokens. Replace the whitespace between the
921 // tokens by the empty string.
922 Whitespaces.replaceWhitespace(
923 *Tokens[LineIndex], /*Newlines=*/0, /*Spaces=*/0,
924 /*StartOfTokenColumn=*/StartColumn, /*IsAligned=*/true,
925 /*InPPDirective=*/false);
926 } else if (LineIndex > 0) {
927 // In case we're reflowing after the '\' in:
928 //
929 // // line comment \
930 // // line 2
931 //
932 // the reflow happens inside the single comment token (it is a single line
933 // comment with an unescaped newline).
934 // Replace the whitespace between the '\' and '//' with the empty string.
935 //
936 // Offset points to after the '\' relative to start of the token.
937 unsigned Offset = Lines[LineIndex - 1].data() +
938 Lines[LineIndex - 1].size() -
939 tokenAt(LineIndex - 1).TokenText.data();
940 // WhitespaceLength is the number of chars between the '\' and the '//' on
941 // the next line.
942 unsigned WhitespaceLength =
943 Lines[LineIndex].data() - tokenAt(LineIndex).TokenText.data() - Offset;
944 Whitespaces.replaceWhitespaceInToken(*Tokens[LineIndex], Offset,
945 /*ReplaceChars=*/WhitespaceLength,
946 /*PreviousPostfix=*/"",
947 /*CurrentPrefix=*/"",
948 /*InPPDirective=*/false,
949 /*Newlines=*/0,
950 /*Spaces=*/0);
951 }
952 // Replace the indent and prefix of the token with the reflow prefix.
953 unsigned Offset =
954 Lines[LineIndex].data() - tokenAt(LineIndex).TokenText.data();
955 unsigned WhitespaceLength =
956 Content[LineIndex].data() - Lines[LineIndex].data();
957 Whitespaces.replaceWhitespaceInToken(*Tokens[LineIndex], Offset,
958 /*ReplaceChars=*/WhitespaceLength,
959 /*PreviousPostfix=*/"",
960 /*CurrentPrefix=*/ReflowPrefix,
961 /*InPPDirective=*/false,
962 /*Newlines=*/0,
963 /*Spaces=*/0);
964 }
965
adaptStartOfLine(unsigned LineIndex,WhitespaceManager & Whitespaces) const966 void BreakableLineCommentSection::adaptStartOfLine(
967 unsigned LineIndex, WhitespaceManager &Whitespaces) const {
968 // If this is the first line of a token, we need to inform Whitespace Manager
969 // about it: either adapt the whitespace range preceding it, or mark it as an
970 // untouchable token.
971 // This happens for instance here:
972 // // line 1 \
973 // // line 2
974 if (LineIndex > 0 && Tokens[LineIndex] != Tokens[LineIndex - 1]) {
975 // This is the first line for the current token, but no reflow with the
976 // previous token is necessary. However, we still may need to adjust the
977 // start column. Note that ContentColumn[LineIndex] is the expected
978 // content column after a possible update to the prefix, hence the prefix
979 // length change is included.
980 unsigned LineColumn =
981 ContentColumn[LineIndex] -
982 (Content[LineIndex].data() - Lines[LineIndex].data()) +
983 (OriginalPrefix[LineIndex].size() - Prefix[LineIndex].size());
984
985 // We always want to create a replacement instead of adding an untouchable
986 // token, even if LineColumn is the same as the original column of the
987 // token. This is because WhitespaceManager doesn't align trailing
988 // comments if they are untouchable.
989 Whitespaces.replaceWhitespace(*Tokens[LineIndex],
990 /*Newlines=*/1,
991 /*Spaces=*/LineColumn,
992 /*StartOfTokenColumn=*/LineColumn,
993 /*IsAligned=*/true,
994 /*InPPDirective=*/false);
995 }
996 if (OriginalPrefix[LineIndex] != Prefix[LineIndex]) {
997 // Adjust the prefix if necessary.
998 const auto SpacesToRemove = -std::min(PrefixSpaceChange[LineIndex], 0);
999 const auto SpacesToAdd = std::max(PrefixSpaceChange[LineIndex], 0);
1000 Whitespaces.replaceWhitespaceInToken(
1001 tokenAt(LineIndex), OriginalPrefix[LineIndex].size() - SpacesToRemove,
1002 /*ReplaceChars=*/SpacesToRemove, "", "", /*InPPDirective=*/false,
1003 /*Newlines=*/0, /*Spaces=*/SpacesToAdd);
1004 }
1005 }
1006
updateNextToken(LineState & State) const1007 void BreakableLineCommentSection::updateNextToken(LineState &State) const {
1008 if (LastLineTok) {
1009 State.NextToken = LastLineTok->Next;
1010 }
1011 }
1012
mayReflow(unsigned LineIndex,const llvm::Regex & CommentPragmasRegex) const1013 bool BreakableLineCommentSection::mayReflow(
1014 unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
1015 // Line comments have the indent as part of the prefix, so we need to
1016 // recompute the start of the line.
1017 StringRef IndentContent = Content[LineIndex];
1018 if (Lines[LineIndex].startswith("//")) {
1019 IndentContent = Lines[LineIndex].substr(2);
1020 }
1021 // FIXME: Decide whether we want to reflow non-regular indents:
1022 // Currently, we only reflow when the OriginalPrefix[LineIndex] matches the
1023 // OriginalPrefix[LineIndex-1]. That means we don't reflow
1024 // // text that protrudes
1025 // // into text with different indent
1026 // We do reflow in that case in block comments.
1027 return LineIndex > 0 && !CommentPragmasRegex.match(IndentContent) &&
1028 mayReflowContent(Content[LineIndex]) && !Tok.Finalized &&
1029 !switchesFormatting(tokenAt(LineIndex)) &&
1030 OriginalPrefix[LineIndex] == OriginalPrefix[LineIndex - 1];
1031 }
1032
1033 } // namespace format
1034 } // namespace clang
1035