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
25namespace clang {
26namespace format {
27
28static constexpr StringRef Blanks = " \t\v\f\r";
29
30static StringRef getLineCommentIndentPrefix(StringRef Comment,
31 const FormatStyle &Style) {
32 static constexpr StringRef KnownCStylePrefixes[] = {"///<", "//!<", "///",
33 "//!", "//:", "//"};
34 static constexpr StringRef KnownTextProtoPrefixes[] = {"####", "###", "##",
35 "//", "#"};
36 ArrayRef<StringRef> KnownPrefixes(KnownCStylePrefixes);
37 if (Style.isTextProto())
38 KnownPrefixes = KnownTextProtoPrefixes;
39
40 assert(
41 llvm::is_sorted(KnownPrefixes, [](StringRef Lhs, StringRef Rhs) noexcept {
42 return Lhs.size() > Rhs.size();
43 }));
44
45 for (StringRef KnownPrefix : KnownPrefixes) {
46 if (Comment.starts_with(Prefix: KnownPrefix)) {
47 const auto PrefixLength =
48 Comment.find_first_not_of(C: ' ', From: KnownPrefix.size());
49 return Comment.substr(Start: 0, N: PrefixLength);
50 }
51 }
52 return {};
53}
54
55static BreakableToken::Split
56getCommentSplit(StringRef Text, unsigned ContentStartColumn,
57 unsigned ColumnLimit, unsigned TabWidth,
58 encoding::Encoding Encoding, const FormatStyle &Style,
59 bool DecorationEndsWithStar = false) {
60 LLVM_DEBUG(llvm::dbgs() << "Comment split: \"" << Text
61 << "\", Column limit: " << ColumnLimit
62 << ", Content start: " << ContentStartColumn << "\n");
63 if (ColumnLimit <= ContentStartColumn + 1)
64 return BreakableToken::Split(StringRef::npos, 0);
65
66 unsigned MaxSplit = ColumnLimit - ContentStartColumn + 1;
67 unsigned MaxSplitBytes = 0;
68
69 for (unsigned NumChars = 0;
70 NumChars < MaxSplit && MaxSplitBytes < Text.size();) {
71 unsigned BytesInChar =
72 encoding::getCodePointNumBytes(FirstChar: Text[MaxSplitBytes], Encoding);
73 NumChars += encoding::columnWidthWithTabs(
74 Text: Text.substr(Start: MaxSplitBytes, N: BytesInChar), StartColumn: ContentStartColumn + NumChars,
75 TabWidth, Encoding);
76 MaxSplitBytes += BytesInChar;
77 }
78
79 // In JavaScript, some @tags can be followed by {, and machinery that parses
80 // these comments will fail to understand the comment if followed by a line
81 // break. So avoid ever breaking before a {.
82 if (Style.isJavaScript()) {
83 StringRef::size_type SpaceOffset =
84 Text.find_first_of(Chars: Blanks, From: MaxSplitBytes);
85 if (SpaceOffset != StringRef::npos && SpaceOffset + 1 < Text.size() &&
86 Text[SpaceOffset + 1] == '{') {
87 MaxSplitBytes = SpaceOffset + 1;
88 }
89 }
90
91 StringRef::size_type SpaceOffset = Text.find_last_of(Chars: Blanks, From: MaxSplitBytes);
92
93 static const auto kNumberedListRegexp = llvm::Regex("^[1-9][0-9]?\\.");
94 // Some spaces are unacceptable to break on, rewind past them.
95 while (SpaceOffset != StringRef::npos) {
96 // If a line-comment ends with `\`, the next line continues the comment,
97 // whether or not it starts with `//`. This is confusing and triggers
98 // -Wcomment.
99 // Avoid introducing multiline comments by not allowing a break right
100 // after '\'.
101 if (Style.isCpp()) {
102 StringRef::size_type LastNonBlank =
103 Text.find_last_not_of(Chars: Blanks, From: SpaceOffset);
104 if (LastNonBlank != StringRef::npos && Text[LastNonBlank] == '\\') {
105 SpaceOffset = Text.find_last_of(Chars: Blanks, From: LastNonBlank);
106 continue;
107 }
108 }
109
110 // Do not split before a number followed by a dot: this would be interpreted
111 // as a numbered list, which would prevent re-flowing in subsequent passes.
112 if (kNumberedListRegexp.match(String: Text.substr(Start: SpaceOffset).ltrim(Chars: Blanks))) {
113 SpaceOffset = Text.find_last_of(Chars: Blanks, From: SpaceOffset);
114 continue;
115 }
116
117 // Avoid ever breaking before a @tag or a { in JavaScript.
118 if (Style.isJavaScript() && SpaceOffset + 1 < Text.size() &&
119 (Text[SpaceOffset + 1] == '{' || Text[SpaceOffset + 1] == '@')) {
120 SpaceOffset = Text.find_last_of(Chars: Blanks, From: SpaceOffset);
121 continue;
122 }
123
124 break;
125 }
126
127 if (SpaceOffset == StringRef::npos ||
128 // Don't break at leading whitespace.
129 Text.find_last_not_of(Chars: Blanks, From: SpaceOffset) == StringRef::npos) {
130 // Make sure that we don't break at leading whitespace that
131 // reaches past MaxSplit.
132 StringRef::size_type FirstNonWhitespace = Text.find_first_not_of(Chars: Blanks);
133 if (FirstNonWhitespace == StringRef::npos) {
134 // If the comment is only whitespace, we cannot split.
135 return BreakableToken::Split(StringRef::npos, 0);
136 }
137 SpaceOffset = Text.find_first_of(
138 Chars: Blanks, From: std::max<unsigned>(a: MaxSplitBytes, b: FirstNonWhitespace));
139 }
140 if (SpaceOffset != StringRef::npos && SpaceOffset != 0) {
141 // adaptStartOfLine will break after lines starting with /** if the comment
142 // is broken anywhere. Avoid emitting this break twice here.
143 // Example: in /** longtextcomesherethatbreaks */ (with ColumnLimit 20) will
144 // insert a break after /**, so this code must not insert the same break.
145 if (SpaceOffset == 1 && Text[SpaceOffset - 1] == '*')
146 return BreakableToken::Split(StringRef::npos, 0);
147 StringRef BeforeCut = Text.substr(Start: 0, N: SpaceOffset).rtrim(Chars: Blanks);
148 StringRef AfterCut = Text.substr(Start: SpaceOffset);
149 if (!DecorationEndsWithStar)
150 AfterCut = AfterCut.ltrim(Chars: Blanks);
151 return BreakableToken::Split(BeforeCut.size(),
152 AfterCut.begin() - BeforeCut.end());
153 }
154 return BreakableToken::Split(StringRef::npos, 0);
155}
156
157static BreakableToken::Split
158getStringSplit(StringRef Text, unsigned UsedColumns, unsigned ColumnLimit,
159 unsigned TabWidth, encoding::Encoding Encoding) {
160 // FIXME: Reduce unit test case.
161 if (Text.empty())
162 return BreakableToken::Split(StringRef::npos, 0);
163 if (ColumnLimit <= UsedColumns)
164 return BreakableToken::Split(StringRef::npos, 0);
165 unsigned MaxSplit = ColumnLimit - UsedColumns;
166 StringRef::size_type SpaceOffset = 0;
167 StringRef::size_type SlashOffset = 0;
168 StringRef::size_type WordStartOffset = 0;
169 StringRef::size_type SplitPoint = 0;
170 for (unsigned Chars = 0;;) {
171 unsigned Advance;
172 if (Text[0] == '\\') {
173 Advance = encoding::getEscapeSequenceLength(Text);
174 Chars += Advance;
175 } else {
176 Advance = encoding::getCodePointNumBytes(FirstChar: Text[0], Encoding);
177 Chars += encoding::columnWidthWithTabs(
178 Text: Text.substr(Start: 0, N: Advance), StartColumn: UsedColumns + Chars, TabWidth, Encoding);
179 }
180
181 if (Chars > MaxSplit || Text.size() <= Advance)
182 break;
183
184 if (Blanks.contains(C: Text[0]))
185 SpaceOffset = SplitPoint;
186 if (Text[0] == '/')
187 SlashOffset = SplitPoint;
188 if (Advance == 1 && !isAlphanumeric(c: Text[0]))
189 WordStartOffset = SplitPoint;
190
191 SplitPoint += Advance;
192 Text = Text.substr(Start: Advance);
193 }
194
195 if (SpaceOffset != 0)
196 return BreakableToken::Split(SpaceOffset + 1, 0);
197 if (SlashOffset != 0)
198 return BreakableToken::Split(SlashOffset + 1, 0);
199 if (WordStartOffset != 0)
200 return BreakableToken::Split(WordStartOffset + 1, 0);
201 if (SplitPoint != 0)
202 return BreakableToken::Split(SplitPoint, 0);
203 return BreakableToken::Split(StringRef::npos, 0);
204}
205
206bool switchesFormatting(const FormatToken &Token) {
207 assert((Token.is(TT_BlockComment) || Token.is(TT_LineComment)) &&
208 "formatting regions are switched by comment tokens");
209 StringRef Content = Token.TokenText.substr(Start: 2).ltrim();
210 return Content.starts_with(Prefix: "clang-format on") ||
211 Content.starts_with(Prefix: "clang-format off");
212}
213
214unsigned
215BreakableToken::getLengthAfterCompression(unsigned RemainingTokenColumns,
216 Split Split) const {
217 // Example: consider the content
218 // lala lala
219 // - RemainingTokenColumns is the original number of columns, 10;
220 // - Split is (4, 2), denoting the two spaces between the two words;
221 //
222 // We compute the number of columns when the split is compressed into a single
223 // space, like:
224 // lala lala
225 //
226 // FIXME: Correctly measure the length of whitespace in Split.second so it
227 // works with tabs.
228 return RemainingTokenColumns + 1 - Split.second;
229}
230
231unsigned BreakableStringLiteral::getLineCount() const { return 1; }
232
233unsigned BreakableStringLiteral::getRangeLength(unsigned LineIndex,
234 unsigned Offset,
235 StringRef::size_type Length,
236 unsigned StartColumn) const {
237 llvm_unreachable("Getting the length of a part of the string literal "
238 "indicates that the code tries to reflow it.");
239}
240
241unsigned
242BreakableStringLiteral::getRemainingLength(unsigned LineIndex, unsigned Offset,
243 unsigned StartColumn) const {
244 return UnbreakableTailLength + Postfix.size() +
245 encoding::columnWidthWithTabs(Text: Line.substr(Start: Offset), StartColumn,
246 TabWidth: Style.TabWidth, Encoding);
247}
248
249unsigned BreakableStringLiteral::getContentStartColumn(unsigned LineIndex,
250 bool Break) const {
251 return StartColumn + Prefix.size();
252}
253
254BreakableStringLiteral::BreakableStringLiteral(
255 const FormatToken &Tok, unsigned StartColumn, StringRef Prefix,
256 StringRef Postfix, unsigned UnbreakableTailLength, bool InPPDirective,
257 encoding::Encoding Encoding, const FormatStyle &Style)
258 : BreakableToken(Tok, InPPDirective, Encoding, Style),
259 StartColumn(StartColumn), Prefix(Prefix), Postfix(Postfix),
260 UnbreakableTailLength(UnbreakableTailLength) {
261 assert(Tok.TokenText.starts_with(Prefix) && Tok.TokenText.ends_with(Postfix));
262 Line = Tok.TokenText.substr(
263 Start: Prefix.size(), N: Tok.TokenText.size() - Prefix.size() - Postfix.size());
264}
265
266BreakableToken::Split BreakableStringLiteral::getSplit(
267 unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit,
268 unsigned ContentStartColumn, const llvm::Regex &CommentPragmasRegex) const {
269 return getStringSplit(Text: Line.substr(Start: TailOffset), UsedColumns: ContentStartColumn,
270 ColumnLimit: ColumnLimit - Postfix.size(), TabWidth: Style.TabWidth, Encoding);
271}
272
273void BreakableStringLiteral::insertBreak(unsigned LineIndex,
274 unsigned TailOffset, Split Split,
275 unsigned ContentIndent,
276 WhitespaceManager &Whitespaces) const {
277 Whitespaces.replaceWhitespaceInToken(
278 Tok, Offset: Prefix.size() + TailOffset + Split.first, ReplaceChars: Split.second, PreviousPostfix: Postfix,
279 CurrentPrefix: Prefix, InPPDirective, Newlines: 1, Spaces: StartColumn);
280}
281
282BreakableStringLiteralUsingOperators::BreakableStringLiteralUsingOperators(
283 const FormatToken &Tok, QuoteStyleType QuoteStyle, bool UnindentPlus,
284 unsigned StartColumn, unsigned UnbreakableTailLength, bool InPPDirective,
285 encoding::Encoding Encoding, const FormatStyle &Style)
286 : BreakableStringLiteral(
287 Tok, StartColumn, /*Prefix=*/QuoteStyle == SingleQuotes ? "'"
288 : QuoteStyle == AtDoubleQuotes ? "@\""
289 : "\"",
290 /*Postfix=*/QuoteStyle == SingleQuotes ? "'" : "\"",
291 UnbreakableTailLength, InPPDirective, Encoding, Style),
292 BracesNeeded(Tok.isNot(Kind: TT_StringInConcatenation)),
293 QuoteStyle(QuoteStyle) {
294 // Find the replacement text for inserting braces and quotes and line breaks.
295 // We don't create an allocated string concatenated from parts here because it
296 // has to outlive the BreakableStringliteral object. The brace replacements
297 // include a quote so that WhitespaceManager can tell it apart from whitespace
298 // replacements between the string and surrounding tokens.
299
300 // The option is not implemented in JavaScript.
301 bool SignOnNewLine =
302 !Style.isJavaScript() &&
303 Style.BreakBeforeBinaryOperators != FormatStyle::BOS_None;
304
305 if (Style.isVerilog()) {
306 // In Verilog, all strings are quoted by double quotes, joined by commas,
307 // and wrapped in braces. The comma is always before the newline.
308 assert(QuoteStyle == DoubleQuotes);
309 LeftBraceQuote = Style.Cpp11BracedListStyle ? "{\"" : "{ \"";
310 RightBraceQuote = Style.Cpp11BracedListStyle ? "\"}" : "\" }";
311 Postfix = "\",";
312 Prefix = "\"";
313 } else {
314 // The plus sign may be on either line. And also C# and JavaScript have
315 // several quoting styles.
316 if (QuoteStyle == SingleQuotes) {
317 LeftBraceQuote = Style.SpacesInParensOptions.Other ? "( '" : "('";
318 RightBraceQuote = Style.SpacesInParensOptions.Other ? "' )" : "')";
319 Postfix = SignOnNewLine ? "'" : "' +";
320 Prefix = SignOnNewLine ? "+ '" : "'";
321 } else {
322 if (QuoteStyle == AtDoubleQuotes) {
323 LeftBraceQuote = Style.SpacesInParensOptions.Other ? "( @" : "(@";
324 Prefix = SignOnNewLine ? "+ @\"" : "@\"";
325 } else {
326 LeftBraceQuote = Style.SpacesInParensOptions.Other ? "( \"" : "(\"";
327 Prefix = SignOnNewLine ? "+ \"" : "\"";
328 }
329 RightBraceQuote = Style.SpacesInParensOptions.Other ? "\" )" : "\")";
330 Postfix = SignOnNewLine ? "\"" : "\" +";
331 }
332 }
333
334 // Following lines are indented by the width of the brace and space if any.
335 ContinuationIndent = BracesNeeded ? LeftBraceQuote.size() - 1 : 0;
336 // The plus sign may need to be unindented depending on the style.
337 // FIXME: Add support for DontAlign.
338 if (!Style.isVerilog() && SignOnNewLine && !BracesNeeded && UnindentPlus &&
339 Style.AlignOperands == FormatStyle::OAS_AlignAfterOperator) {
340 ContinuationIndent -= 2;
341 }
342}
343
344unsigned BreakableStringLiteralUsingOperators::getRemainingLength(
345 unsigned LineIndex, unsigned Offset, unsigned StartColumn) const {
346 return UnbreakableTailLength + (BracesNeeded ? RightBraceQuote.size() : 1) +
347 encoding::columnWidthWithTabs(Text: Line.substr(Start: Offset), StartColumn,
348 TabWidth: Style.TabWidth, Encoding);
349}
350
351unsigned
352BreakableStringLiteralUsingOperators::getContentStartColumn(unsigned LineIndex,
353 bool Break) const {
354 return std::max(
355 a: 0,
356 b: static_cast<int>(StartColumn) +
357 (Break ? ContinuationIndent + static_cast<int>(Prefix.size())
358 : (BracesNeeded ? static_cast<int>(LeftBraceQuote.size()) - 1
359 : 0) +
360 (QuoteStyle == AtDoubleQuotes ? 2 : 1)));
361}
362
363void BreakableStringLiteralUsingOperators::insertBreak(
364 unsigned LineIndex, unsigned TailOffset, Split Split,
365 unsigned ContentIndent, WhitespaceManager &Whitespaces) const {
366 Whitespaces.replaceWhitespaceInToken(
367 Tok, /*Offset=*/(QuoteStyle == AtDoubleQuotes ? 2 : 1) + TailOffset +
368 Split.first,
369 /*ReplaceChars=*/Split.second, /*PreviousPostfix=*/Postfix,
370 /*CurrentPrefix=*/Prefix, InPPDirective, /*NewLines=*/Newlines: 1,
371 /*Spaces=*/
372 std::max(a: 0, b: static_cast<int>(StartColumn) + ContinuationIndent));
373}
374
375void BreakableStringLiteralUsingOperators::updateAfterBroken(
376 WhitespaceManager &Whitespaces) const {
377 // Add the braces required for breaking the token if they are needed.
378 if (!BracesNeeded)
379 return;
380
381 // To add a brace or parenthesis, we replace the quote (or the at sign) with a
382 // brace and another quote. This is because the rest of the program requires
383 // one replacement for each source range. If we replace the empty strings
384 // around the string, it may conflict with whitespace replacements between the
385 // string and adjacent tokens.
386 Whitespaces.replaceWhitespaceInToken(
387 Tok, /*Offset=*/0, /*ReplaceChars=*/1, /*PreviousPostfix=*/"",
388 /*CurrentPrefix=*/LeftBraceQuote, InPPDirective, /*NewLines=*/Newlines: 0,
389 /*Spaces=*/0);
390 Whitespaces.replaceWhitespaceInToken(
391 Tok, /*Offset=*/Tok.TokenText.size() - 1, /*ReplaceChars=*/1,
392 /*PreviousPostfix=*/RightBraceQuote,
393 /*CurrentPrefix=*/"", InPPDirective, /*NewLines=*/Newlines: 0, /*Spaces=*/0);
394}
395
396BreakableComment::BreakableComment(const FormatToken &Token,
397 unsigned StartColumn, bool InPPDirective,
398 encoding::Encoding Encoding,
399 const FormatStyle &Style)
400 : BreakableToken(Token, InPPDirective, Encoding, Style),
401 StartColumn(StartColumn) {}
402
403unsigned BreakableComment::getLineCount() const { return Lines.size(); }
404
405BreakableToken::Split
406BreakableComment::getSplit(unsigned LineIndex, unsigned TailOffset,
407 unsigned ColumnLimit, unsigned ContentStartColumn,
408 const llvm::Regex &CommentPragmasRegex) const {
409 // Don't break lines matching the comment pragmas regex.
410 if (!AlwaysReflow || CommentPragmasRegex.match(String: Content[LineIndex]))
411 return Split(StringRef::npos, 0);
412 return getCommentSplit(Text: Content[LineIndex].substr(Start: TailOffset),
413 ContentStartColumn, ColumnLimit, TabWidth: Style.TabWidth,
414 Encoding, Style);
415}
416
417void BreakableComment::compressWhitespace(
418 unsigned LineIndex, unsigned TailOffset, Split Split,
419 WhitespaceManager &Whitespaces) const {
420 StringRef Text = Content[LineIndex].substr(Start: TailOffset);
421 // Text is relative to the content line, but Whitespaces operates relative to
422 // the start of the corresponding token, so compute the start of the Split
423 // that needs to be compressed into a single space relative to the start of
424 // its token.
425 unsigned BreakOffsetInToken =
426 Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
427 unsigned CharsToRemove = Split.second;
428 Whitespaces.replaceWhitespaceInToken(
429 Tok: tokenAt(LineIndex), Offset: BreakOffsetInToken, ReplaceChars: CharsToRemove, PreviousPostfix: "", CurrentPrefix: "",
430 /*InPPDirective=*/false, /*Newlines=*/0, /*Spaces=*/1);
431}
432
433const FormatToken &BreakableComment::tokenAt(unsigned LineIndex) const {
434 return Tokens[LineIndex] ? *Tokens[LineIndex] : Tok;
435}
436
437static bool mayReflowContent(StringRef Content) {
438 Content = Content.trim(Chars: Blanks);
439 // Lines starting with '@' or '\' commonly have special meaning.
440 // Lines starting with '-', '-#', '+' or '*' are bulleted/numbered lists.
441 bool hasSpecialMeaningPrefix = false;
442 for (StringRef Prefix :
443 {"@", "\\", "TODO", "FIXME", "XXX", "-# ", "- ", "+ ", "* "}) {
444 if (Content.starts_with(Prefix)) {
445 hasSpecialMeaningPrefix = true;
446 break;
447 }
448 }
449
450 // Numbered lists may also start with a number followed by '.'
451 // To avoid issues if a line starts with a number which is actually the end
452 // of a previous line, we only consider numbers with up to 2 digits.
453 static const auto kNumberedListRegexp = llvm::Regex("^[1-9][0-9]?\\. ");
454 hasSpecialMeaningPrefix =
455 hasSpecialMeaningPrefix || kNumberedListRegexp.match(String: Content);
456
457 // Simple heuristic for what to reflow: content should contain at least two
458 // characters and either the first or second character must be
459 // non-punctuation.
460 return Content.size() >= 2 && !hasSpecialMeaningPrefix &&
461 !Content.ends_with(Suffix: "\\") &&
462 // Note that this is UTF-8 safe, since if isPunctuation(Content[0]) is
463 // true, then the first code point must be 1 byte long.
464 (!isPunctuation(c: Content[0]) || !isPunctuation(c: Content[1]));
465}
466
467BreakableBlockComment::BreakableBlockComment(
468 const FormatToken &Token, unsigned StartColumn,
469 unsigned OriginalStartColumn, bool FirstInLine, bool InPPDirective,
470 encoding::Encoding Encoding, const FormatStyle &Style, bool UseCRLF)
471 : BreakableComment(Token, StartColumn, InPPDirective, Encoding, Style),
472 DelimitersOnNewline(false),
473 UnbreakableTailLength(Token.UnbreakableTailLength) {
474 assert(Tok.is(TT_BlockComment) &&
475 "block comment section must start with a block comment");
476
477 StringRef TokenText(Tok.TokenText);
478 assert(TokenText.starts_with("/*") && TokenText.ends_with("*/"));
479 TokenText.substr(Start: 2, N: TokenText.size() - 4)
480 .split(A&: Lines, Separator: UseCRLF ? "\r\n" : "\n");
481
482 int IndentDelta = StartColumn - OriginalStartColumn;
483 Content.resize(N: Lines.size());
484 Content[0] = Lines[0];
485 ContentColumn.resize(N: Lines.size());
486 // Account for the initial '/*'.
487 ContentColumn[0] = StartColumn + 2;
488 Tokens.resize(N: Lines.size());
489 for (size_t i = 1; i < Lines.size(); ++i)
490 adjustWhitespace(LineIndex: i, IndentDelta);
491
492 // Align decorations with the column of the star on the first line,
493 // that is one column after the start "/*".
494 DecorationColumn = StartColumn + 1;
495
496 // Account for comment decoration patterns like this:
497 //
498 // /*
499 // ** blah blah blah
500 // */
501 if (Lines.size() >= 2 && Content[1].starts_with(Prefix: "**") &&
502 static_cast<unsigned>(ContentColumn[1]) == StartColumn) {
503 DecorationColumn = StartColumn;
504 }
505
506 Decoration = "* ";
507 if (Lines.size() == 1 && !FirstInLine) {
508 // Comments for which FirstInLine is false can start on arbitrary column,
509 // and available horizontal space can be too small to align consecutive
510 // lines with the first one.
511 // FIXME: We could, probably, align them to current indentation level, but
512 // now we just wrap them without stars.
513 Decoration = "";
514 }
515 for (size_t i = 1, e = Content.size(); i < e && !Decoration.empty(); ++i) {
516 const StringRef &Text = Content[i];
517 if (i + 1 == e) {
518 // If the last line is empty, the closing "*/" will have a star.
519 if (Text.empty())
520 break;
521 } else if (!Text.empty() && Decoration.starts_with(Prefix: Text)) {
522 continue;
523 }
524 while (!Text.starts_with(Prefix: Decoration))
525 Decoration = Decoration.drop_back(N: 1);
526 }
527
528 LastLineNeedsDecoration = true;
529 IndentAtLineBreak = ContentColumn[0] + 1;
530 for (size_t i = 1, e = Lines.size(); i < e; ++i) {
531 if (Content[i].empty()) {
532 if (i + 1 == e) {
533 // Empty last line means that we already have a star as a part of the
534 // trailing */. We also need to preserve whitespace, so that */ is
535 // correctly indented.
536 LastLineNeedsDecoration = false;
537 // Align the star in the last '*/' with the stars on the previous lines.
538 if (e >= 2 && !Decoration.empty())
539 ContentColumn[i] = DecorationColumn;
540 } else if (Decoration.empty()) {
541 // For all other lines, set the start column to 0 if they're empty, so
542 // we do not insert trailing whitespace anywhere.
543 ContentColumn[i] = 0;
544 }
545 continue;
546 }
547
548 // The first line already excludes the star.
549 // The last line excludes the star if LastLineNeedsDecoration is false.
550 // For all other lines, adjust the line to exclude the star and
551 // (optionally) the first whitespace.
552 unsigned DecorationSize = Decoration.starts_with(Prefix: Content[i])
553 ? Content[i].size()
554 : Decoration.size();
555 if (DecorationSize)
556 ContentColumn[i] = DecorationColumn + DecorationSize;
557 Content[i] = Content[i].substr(Start: DecorationSize);
558 if (!Decoration.starts_with(Prefix: Content[i])) {
559 IndentAtLineBreak =
560 std::min<int>(a: IndentAtLineBreak, b: std::max(a: 0, b: ContentColumn[i]));
561 }
562 }
563 IndentAtLineBreak = std::max<unsigned>(a: IndentAtLineBreak, b: Decoration.size());
564
565 // Detect a multiline jsdoc comment and set DelimitersOnNewline in that case.
566 if (Style.isJavaScript() || Style.isJava()) {
567 if ((Lines[0] == "*" || Lines[0].starts_with(Prefix: "* ")) && Lines.size() > 1) {
568 // This is a multiline jsdoc comment.
569 DelimitersOnNewline = true;
570 } else if (Lines[0].starts_with(Prefix: "* ") && Lines.size() == 1) {
571 // Detect a long single-line comment, like:
572 // /** long long long */
573 // Below, '2' is the width of '*/'.
574 unsigned EndColumn =
575 ContentColumn[0] +
576 encoding::columnWidthWithTabs(Text: Lines[0], StartColumn: ContentColumn[0],
577 TabWidth: Style.TabWidth, Encoding) +
578 2;
579 DelimitersOnNewline = EndColumn > Style.ColumnLimit;
580 }
581 }
582
583 LLVM_DEBUG({
584 llvm::dbgs() << "IndentAtLineBreak " << IndentAtLineBreak << "\n";
585 llvm::dbgs() << "DelimitersOnNewline " << DelimitersOnNewline << "\n";
586 for (size_t i = 0; i < Lines.size(); ++i) {
587 llvm::dbgs() << i << " |" << Content[i] << "| "
588 << "CC=" << ContentColumn[i] << "| "
589 << "IN=" << (Content[i].data() - Lines[i].data()) << "\n";
590 }
591 });
592}
593
594BreakableToken::Split BreakableBlockComment::getSplit(
595 unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit,
596 unsigned ContentStartColumn, const llvm::Regex &CommentPragmasRegex) const {
597 // Don't break lines matching the comment pragmas regex.
598 if (!AlwaysReflow || CommentPragmasRegex.match(String: Content[LineIndex]))
599 return Split(StringRef::npos, 0);
600 return getCommentSplit(Text: Content[LineIndex].substr(Start: TailOffset),
601 ContentStartColumn, ColumnLimit, TabWidth: Style.TabWidth,
602 Encoding, Style, DecorationEndsWithStar: Decoration.ends_with(Suffix: "*"));
603}
604
605void BreakableBlockComment::adjustWhitespace(unsigned LineIndex,
606 int IndentDelta) {
607 // When in a preprocessor directive, the trailing backslash in a block comment
608 // is not needed, but can serve a purpose of uniformity with necessary escaped
609 // newlines outside the comment. In this case we remove it here before
610 // trimming the trailing whitespace. The backslash will be re-added later when
611 // inserting a line break.
612 size_t EndOfPreviousLine = Lines[LineIndex - 1].size();
613 if (InPPDirective && Lines[LineIndex - 1].ends_with(Suffix: "\\"))
614 --EndOfPreviousLine;
615
616 // Calculate the end of the non-whitespace text in the previous line.
617 EndOfPreviousLine =
618 Lines[LineIndex - 1].find_last_not_of(Chars: Blanks, From: EndOfPreviousLine);
619 if (EndOfPreviousLine == StringRef::npos)
620 EndOfPreviousLine = 0;
621 else
622 ++EndOfPreviousLine;
623 // Calculate the start of the non-whitespace text in the current line.
624 size_t StartOfLine = Lines[LineIndex].find_first_not_of(Chars: Blanks);
625 if (StartOfLine == StringRef::npos)
626 StartOfLine = Lines[LineIndex].size();
627
628 StringRef Whitespace = Lines[LineIndex].substr(Start: 0, N: StartOfLine);
629 // Adjust Lines to only contain relevant text.
630 size_t PreviousContentOffset =
631 Content[LineIndex - 1].data() - Lines[LineIndex - 1].data();
632 Content[LineIndex - 1] = Lines[LineIndex - 1].substr(
633 Start: PreviousContentOffset, N: EndOfPreviousLine - PreviousContentOffset);
634 Content[LineIndex] = Lines[LineIndex].substr(Start: StartOfLine);
635
636 // Adjust the start column uniformly across all lines.
637 ContentColumn[LineIndex] =
638 encoding::columnWidthWithTabs(Text: Whitespace, StartColumn: 0, TabWidth: Style.TabWidth, Encoding) +
639 IndentDelta;
640}
641
642unsigned BreakableBlockComment::getRangeLength(unsigned LineIndex,
643 unsigned Offset,
644 StringRef::size_type Length,
645 unsigned StartColumn) const {
646 return encoding::columnWidthWithTabs(
647 Text: Content[LineIndex].substr(Start: Offset, N: Length), StartColumn, TabWidth: Style.TabWidth,
648 Encoding);
649}
650
651unsigned BreakableBlockComment::getRemainingLength(unsigned LineIndex,
652 unsigned Offset,
653 unsigned StartColumn) const {
654 unsigned LineLength =
655 UnbreakableTailLength +
656 getRangeLength(LineIndex, Offset, Length: StringRef::npos, StartColumn);
657 if (LineIndex + 1 == Lines.size()) {
658 LineLength += 2;
659 // We never need a decoration when breaking just the trailing "*/" postfix.
660 bool HasRemainingText = Offset < Content[LineIndex].size();
661 if (!HasRemainingText) {
662 bool HasDecoration = Lines[LineIndex].ltrim().starts_with(Prefix: Decoration);
663 if (HasDecoration)
664 LineLength -= Decoration.size();
665 }
666 }
667 return LineLength;
668}
669
670unsigned BreakableBlockComment::getContentStartColumn(unsigned LineIndex,
671 bool Break) const {
672 if (Break)
673 return IndentAtLineBreak;
674 return std::max(a: 0, b: ContentColumn[LineIndex]);
675}
676
677const llvm::StringSet<>
678 BreakableBlockComment::ContentIndentingJavadocAnnotations = {
679 "@param", "@return", "@returns", "@throws", "@type", "@template",
680 "@see", "@deprecated", "@define", "@exports", "@mods", "@private",
681};
682
683unsigned BreakableBlockComment::getContentIndent(unsigned LineIndex) const {
684 if (!Style.isJava() && !Style.isJavaScript())
685 return 0;
686 // The content at LineIndex 0 of a comment like:
687 // /** line 0 */
688 // is "* line 0", so we need to skip over the decoration in that case.
689 StringRef ContentWithNoDecoration = Content[LineIndex];
690 if (LineIndex == 0 && ContentWithNoDecoration.starts_with(Prefix: "*"))
691 ContentWithNoDecoration = ContentWithNoDecoration.substr(Start: 1).ltrim(Chars: Blanks);
692 StringRef FirstWord = ContentWithNoDecoration.substr(
693 Start: 0, N: ContentWithNoDecoration.find_first_of(Chars: Blanks));
694 if (ContentIndentingJavadocAnnotations.contains(key: FirstWord))
695 return Style.ContinuationIndentWidth;
696 return 0;
697}
698
699void BreakableBlockComment::insertBreak(unsigned LineIndex, unsigned TailOffset,
700 Split Split, unsigned ContentIndent,
701 WhitespaceManager &Whitespaces) const {
702 StringRef Text = Content[LineIndex].substr(Start: TailOffset);
703 StringRef Prefix = Decoration;
704 // We need this to account for the case when we have a decoration "* " for all
705 // the lines except for the last one, where the star in "*/" acts as a
706 // decoration.
707 unsigned LocalIndentAtLineBreak = IndentAtLineBreak;
708 if (LineIndex + 1 == Lines.size() &&
709 Text.size() == Split.first + Split.second) {
710 // For the last line we need to break before "*/", but not to add "* ".
711 Prefix = "";
712 if (LocalIndentAtLineBreak >= 2)
713 LocalIndentAtLineBreak -= 2;
714 }
715 // The split offset is from the beginning of the line. Convert it to an offset
716 // from the beginning of the token text.
717 unsigned BreakOffsetInToken =
718 Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
719 unsigned CharsToRemove = Split.second;
720 assert(LocalIndentAtLineBreak >= Prefix.size());
721 std::string PrefixWithTrailingIndent = std::string(Prefix);
722 PrefixWithTrailingIndent.append(n: ContentIndent, c: ' ');
723 Whitespaces.replaceWhitespaceInToken(
724 Tok: tokenAt(LineIndex), Offset: BreakOffsetInToken, ReplaceChars: CharsToRemove, PreviousPostfix: "",
725 CurrentPrefix: PrefixWithTrailingIndent, InPPDirective, /*Newlines=*/1,
726 /*Spaces=*/LocalIndentAtLineBreak + ContentIndent -
727 PrefixWithTrailingIndent.size());
728}
729
730BreakableToken::Split BreakableBlockComment::getReflowSplit(
731 unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
732 if (!mayReflow(LineIndex, CommentPragmasRegex))
733 return Split(StringRef::npos, 0);
734
735 // If we're reflowing into a line with content indent, only reflow the next
736 // line if its starting whitespace matches the content indent.
737 size_t Trimmed = Content[LineIndex].find_first_not_of(Chars: Blanks);
738 if (LineIndex) {
739 unsigned PreviousContentIndent = getContentIndent(LineIndex: LineIndex - 1);
740 if (PreviousContentIndent && Trimmed != StringRef::npos &&
741 Trimmed != PreviousContentIndent) {
742 return Split(StringRef::npos, 0);
743 }
744 }
745
746 return Split(0, Trimmed != StringRef::npos ? Trimmed : 0);
747}
748
749bool BreakableBlockComment::introducesBreakBeforeToken() const {
750 // A break is introduced when we want delimiters on newline.
751 return DelimitersOnNewline &&
752 Lines[0].substr(Start: 1).find_first_not_of(Chars: Blanks) != StringRef::npos;
753}
754
755void BreakableBlockComment::reflow(unsigned LineIndex,
756 WhitespaceManager &Whitespaces) const {
757 StringRef TrimmedContent = Content[LineIndex].ltrim(Chars: Blanks);
758 // Here we need to reflow.
759 assert(Tokens[LineIndex - 1] == Tokens[LineIndex] &&
760 "Reflowing whitespace within a token");
761 // This is the offset of the end of the last line relative to the start of
762 // the token text in the token.
763 unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() +
764 Content[LineIndex - 1].size() -
765 tokenAt(LineIndex).TokenText.data();
766 unsigned WhitespaceLength = TrimmedContent.data() -
767 tokenAt(LineIndex).TokenText.data() -
768 WhitespaceOffsetInToken;
769 Whitespaces.replaceWhitespaceInToken(
770 Tok: tokenAt(LineIndex), Offset: WhitespaceOffsetInToken,
771 /*ReplaceChars=*/WhitespaceLength, /*PreviousPostfix=*/"",
772 /*CurrentPrefix=*/ReflowPrefix, InPPDirective, /*Newlines=*/0,
773 /*Spaces=*/0);
774}
775
776void BreakableBlockComment::adaptStartOfLine(
777 unsigned LineIndex, WhitespaceManager &Whitespaces) const {
778 if (LineIndex == 0) {
779 if (DelimitersOnNewline) {
780 // Since we're breaking at index 1 below, the break position and the
781 // break length are the same.
782 // Note: this works because getCommentSplit is careful never to split at
783 // the beginning of a line.
784 size_t BreakLength = Lines[0].substr(Start: 1).find_first_not_of(Chars: Blanks);
785 if (BreakLength != StringRef::npos) {
786 insertBreak(LineIndex, TailOffset: 0, Split: Split(1, BreakLength), /*ContentIndent=*/0,
787 Whitespaces);
788 }
789 }
790 return;
791 }
792 // Here no reflow with the previous line will happen.
793 // Fix the decoration of the line at LineIndex.
794 StringRef Prefix = Decoration;
795 if (Content[LineIndex].empty()) {
796 if (LineIndex + 1 == Lines.size()) {
797 if (!LastLineNeedsDecoration) {
798 // If the last line was empty, we don't need a prefix, as the */ will
799 // line up with the decoration (if it exists).
800 Prefix = "";
801 }
802 } else if (!Decoration.empty()) {
803 // For other empty lines, if we do have a decoration, adapt it to not
804 // contain a trailing whitespace.
805 Prefix = Prefix.substr(Start: 0, N: 1);
806 }
807 } else if (ContentColumn[LineIndex] == 1) {
808 // This line starts immediately after the decorating *.
809 Prefix = Prefix.substr(Start: 0, N: 1);
810 }
811 // This is the offset of the end of the last line relative to the start of the
812 // token text in the token.
813 unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() +
814 Content[LineIndex - 1].size() -
815 tokenAt(LineIndex).TokenText.data();
816 unsigned WhitespaceLength = Content[LineIndex].data() -
817 tokenAt(LineIndex).TokenText.data() -
818 WhitespaceOffsetInToken;
819 Whitespaces.replaceWhitespaceInToken(
820 Tok: tokenAt(LineIndex), Offset: WhitespaceOffsetInToken, ReplaceChars: WhitespaceLength, PreviousPostfix: "", CurrentPrefix: Prefix,
821 InPPDirective, /*Newlines=*/1, Spaces: ContentColumn[LineIndex] - Prefix.size());
822}
823
824BreakableToken::Split
825BreakableBlockComment::getSplitAfterLastLine(unsigned TailOffset) const {
826 if (DelimitersOnNewline) {
827 // Replace the trailing whitespace of the last line with a newline.
828 // In case the last line is empty, the ending '*/' is already on its own
829 // line.
830 StringRef Line = Content.back().substr(Start: TailOffset);
831 StringRef TrimmedLine = Line.rtrim(Chars: Blanks);
832 if (!TrimmedLine.empty())
833 return Split(TrimmedLine.size(), Line.size() - TrimmedLine.size());
834 }
835 return Split(StringRef::npos, 0);
836}
837
838bool BreakableBlockComment::mayReflow(
839 unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
840 // Content[LineIndex] may exclude the indent after the '*' decoration. In that
841 // case, we compute the start of the comment pragma manually.
842 StringRef IndentContent = Content[LineIndex];
843 if (Lines[LineIndex].ltrim(Chars: Blanks).starts_with(Prefix: "*"))
844 IndentContent = Lines[LineIndex].ltrim(Chars: Blanks).substr(Start: 1);
845 return LineIndex > 0 && AlwaysReflow &&
846 !CommentPragmasRegex.match(String: IndentContent) &&
847 mayReflowContent(Content: Content[LineIndex]) && !Tok.Finalized &&
848 !switchesFormatting(Token: tokenAt(LineIndex));
849}
850
851BreakableLineCommentSection::BreakableLineCommentSection(
852 const FormatToken &Token, unsigned StartColumn, bool InPPDirective,
853 encoding::Encoding Encoding, const FormatStyle &Style)
854 : BreakableComment(Token, StartColumn, InPPDirective, Encoding, Style) {
855 assert(Tok.is(TT_LineComment) &&
856 "line comment section must start with a line comment");
857 FormatToken *LineTok = nullptr;
858 const int Minimum = Style.SpacesInLineCommentPrefix.Minimum;
859 // How many spaces we changed in the first line of the section, this will be
860 // applied in all following lines
861 int FirstLineSpaceChange = 0;
862 for (const FormatToken *CurrentTok = &Tok;
863 CurrentTok && CurrentTok->is(TT: TT_LineComment);
864 CurrentTok = CurrentTok->Next) {
865 LastLineTok = LineTok;
866 StringRef TokenText(CurrentTok->TokenText);
867 assert((TokenText.starts_with("//") || TokenText.starts_with("#")) &&
868 "unsupported line comment prefix, '//' and '#' are supported");
869 size_t FirstLineIndex = Lines.size();
870 TokenText.split(A&: Lines, Separator: "\n");
871 Content.resize(N: Lines.size());
872 ContentColumn.resize(N: Lines.size());
873 PrefixSpaceChange.resize(N: Lines.size());
874 Tokens.resize(N: Lines.size());
875 Prefix.resize(N: Lines.size());
876 OriginalPrefix.resize(N: Lines.size());
877 for (size_t i = FirstLineIndex, e = Lines.size(); i < e; ++i) {
878 Lines[i] = Lines[i].ltrim(Chars: Blanks);
879 StringRef IndentPrefix = getLineCommentIndentPrefix(Comment: Lines[i], Style);
880 OriginalPrefix[i] = IndentPrefix;
881 const int SpacesInPrefix = llvm::count(Range&: IndentPrefix, Element: ' ');
882
883 // This lambda also considers multibyte character that is not handled in
884 // functions like isPunctuation provided by CharInfo.
885 const auto NoSpaceBeforeFirstCommentChar = [&]() {
886 assert(Lines[i].size() > IndentPrefix.size());
887 const char FirstCommentChar = Lines[i][IndentPrefix.size()];
888 const unsigned FirstCharByteSize =
889 encoding::getCodePointNumBytes(FirstChar: FirstCommentChar, Encoding);
890 if (encoding::columnWidth(
891 Text: Lines[i].substr(Start: IndentPrefix.size(), N: FirstCharByteSize),
892 Encoding) != 1) {
893 return false;
894 }
895 // In C-like comments, add a space before #. For example this is useful
896 // to preserve the relative indentation when commenting out code with
897 // #includes.
898 //
899 // In languages using # as the comment leader such as proto, don't
900 // add a space to support patterns like:
901 // #########
902 // # section
903 // #########
904 if (FirstCommentChar == '#' && !TokenText.starts_with(Prefix: "#"))
905 return false;
906 return FirstCommentChar == '\\' || isPunctuation(c: FirstCommentChar) ||
907 isHorizontalWhitespace(c: FirstCommentChar);
908 };
909
910 // On the first line of the comment section we calculate how many spaces
911 // are to be added or removed, all lines after that just get only the
912 // change and we will not look at the maximum anymore. Additionally to the
913 // actual first line, we calculate that when the non space Prefix changes,
914 // e.g. from "///" to "//".
915 if (i == 0 || OriginalPrefix[i].rtrim(Chars: Blanks) !=
916 OriginalPrefix[i - 1].rtrim(Chars: Blanks)) {
917 if (SpacesInPrefix < Minimum && Lines[i].size() > IndentPrefix.size() &&
918 !NoSpaceBeforeFirstCommentChar()) {
919 FirstLineSpaceChange = Minimum - SpacesInPrefix;
920 } else if (static_cast<unsigned>(SpacesInPrefix) >
921 Style.SpacesInLineCommentPrefix.Maximum) {
922 FirstLineSpaceChange =
923 Style.SpacesInLineCommentPrefix.Maximum - SpacesInPrefix;
924 } else {
925 FirstLineSpaceChange = 0;
926 }
927 }
928
929 if (Lines[i].size() != IndentPrefix.size()) {
930 assert(Lines[i].size() > IndentPrefix.size());
931
932 PrefixSpaceChange[i] = SpacesInPrefix + FirstLineSpaceChange < Minimum
933 ? Minimum - SpacesInPrefix
934 : FirstLineSpaceChange;
935
936 const auto FirstNonSpace = Lines[i][IndentPrefix.size()];
937 const bool IsFormatComment = LineTok && switchesFormatting(Token: *LineTok);
938 const bool LineRequiresLeadingSpace =
939 !NoSpaceBeforeFirstCommentChar() ||
940 (FirstNonSpace == '}' && FirstLineSpaceChange != 0);
941 const bool AllowsSpaceChange =
942 !IsFormatComment &&
943 (SpacesInPrefix != 0 || LineRequiresLeadingSpace);
944
945 if (PrefixSpaceChange[i] > 0 && AllowsSpaceChange) {
946 Prefix[i] = IndentPrefix.str();
947 Prefix[i].append(n: PrefixSpaceChange[i], c: ' ');
948 } else if (PrefixSpaceChange[i] < 0 && AllowsSpaceChange) {
949 Prefix[i] = IndentPrefix
950 .drop_back(N: std::min<std::size_t>(
951 a: -PrefixSpaceChange[i], b: SpacesInPrefix))
952 .str();
953 } else {
954 Prefix[i] = IndentPrefix.str();
955 }
956 } else {
957 // If the IndentPrefix is the whole line, there is no content and we
958 // drop just all space
959 Prefix[i] = IndentPrefix.drop_back(N: SpacesInPrefix).str();
960 }
961
962 Tokens[i] = LineTok;
963 Content[i] = Lines[i].substr(Start: IndentPrefix.size());
964 ContentColumn[i] =
965 StartColumn + encoding::columnWidthWithTabs(Text: Prefix[i], StartColumn,
966 TabWidth: Style.TabWidth, Encoding);
967
968 // Calculate the end of the non-whitespace text in this line.
969 size_t EndOfLine = Content[i].find_last_not_of(Chars: Blanks);
970 if (EndOfLine == StringRef::npos)
971 EndOfLine = Content[i].size();
972 else
973 ++EndOfLine;
974 Content[i] = Content[i].substr(Start: 0, N: EndOfLine);
975 }
976 LineTok = CurrentTok->Next;
977 if (CurrentTok->Next && !CurrentTok->Next->ContinuesLineCommentSection) {
978 // A line comment section needs to broken by a line comment that is
979 // preceded by at least two newlines. Note that we put this break here
980 // instead of breaking at a previous stage during parsing, since that
981 // would split the contents of the enum into two unwrapped lines in this
982 // example, which is undesirable:
983 // enum A {
984 // a, // comment about a
985 //
986 // // comment about b
987 // b
988 // };
989 //
990 // FIXME: Consider putting separate line comment sections as children to
991 // the unwrapped line instead.
992 break;
993 }
994 }
995}
996
997unsigned
998BreakableLineCommentSection::getRangeLength(unsigned LineIndex, unsigned Offset,
999 StringRef::size_type Length,
1000 unsigned StartColumn) const {
1001 return encoding::columnWidthWithTabs(
1002 Text: Content[LineIndex].substr(Start: Offset, N: Length), StartColumn, TabWidth: Style.TabWidth,
1003 Encoding);
1004}
1005
1006unsigned
1007BreakableLineCommentSection::getContentStartColumn(unsigned LineIndex,
1008 bool /*Break*/) const {
1009 return ContentColumn[LineIndex];
1010}
1011
1012void BreakableLineCommentSection::insertBreak(
1013 unsigned LineIndex, unsigned TailOffset, Split Split,
1014 unsigned ContentIndent, WhitespaceManager &Whitespaces) const {
1015 StringRef Text = Content[LineIndex].substr(Start: TailOffset);
1016 // Compute the offset of the split relative to the beginning of the token
1017 // text.
1018 unsigned BreakOffsetInToken =
1019 Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
1020 unsigned CharsToRemove = Split.second;
1021 Whitespaces.replaceWhitespaceInToken(
1022 Tok: tokenAt(LineIndex), Offset: BreakOffsetInToken, ReplaceChars: CharsToRemove, PreviousPostfix: "",
1023 CurrentPrefix: Prefix[LineIndex], InPPDirective, /*Newlines=*/1,
1024 /*Spaces=*/ContentColumn[LineIndex] - Prefix[LineIndex].size());
1025}
1026
1027BreakableComment::Split BreakableLineCommentSection::getReflowSplit(
1028 unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
1029 if (!mayReflow(LineIndex, CommentPragmasRegex))
1030 return Split(StringRef::npos, 0);
1031
1032 size_t Trimmed = Content[LineIndex].find_first_not_of(Chars: Blanks);
1033
1034 // In a line comment section each line is a separate token; thus, after a
1035 // split we replace all whitespace before the current line comment token
1036 // (which does not need to be included in the split), plus the start of the
1037 // line up to where the content starts.
1038 return Split(0, Trimmed != StringRef::npos ? Trimmed : 0);
1039}
1040
1041void BreakableLineCommentSection::reflow(unsigned LineIndex,
1042 WhitespaceManager &Whitespaces) const {
1043 if (LineIndex > 0 && Tokens[LineIndex] != Tokens[LineIndex - 1]) {
1044 // Reflow happens between tokens. Replace the whitespace between the
1045 // tokens by the empty string.
1046 Whitespaces.replaceWhitespace(
1047 Tok&: *Tokens[LineIndex], /*Newlines=*/0, /*Spaces=*/0,
1048 /*StartOfTokenColumn=*/StartColumn, /*IsAligned=*/true,
1049 /*InPPDirective=*/false);
1050 } else if (LineIndex > 0) {
1051 // In case we're reflowing after the '\' in:
1052 //
1053 // // line comment \
1054 // // line 2
1055 //
1056 // the reflow happens inside the single comment token (it is a single line
1057 // comment with an unescaped newline).
1058 // Replace the whitespace between the '\' and '//' with the empty string.
1059 //
1060 // Offset points to after the '\' relative to start of the token.
1061 unsigned Offset = Lines[LineIndex - 1].data() +
1062 Lines[LineIndex - 1].size() -
1063 tokenAt(LineIndex: LineIndex - 1).TokenText.data();
1064 // WhitespaceLength is the number of chars between the '\' and the '//' on
1065 // the next line.
1066 unsigned WhitespaceLength =
1067 Lines[LineIndex].data() - tokenAt(LineIndex).TokenText.data() - Offset;
1068 Whitespaces.replaceWhitespaceInToken(Tok: *Tokens[LineIndex], Offset,
1069 /*ReplaceChars=*/WhitespaceLength,
1070 /*PreviousPostfix=*/"",
1071 /*CurrentPrefix=*/"",
1072 /*InPPDirective=*/false,
1073 /*Newlines=*/0,
1074 /*Spaces=*/0);
1075 }
1076 // Replace the indent and prefix of the token with the reflow prefix.
1077 unsigned Offset =
1078 Lines[LineIndex].data() - tokenAt(LineIndex).TokenText.data();
1079 unsigned WhitespaceLength =
1080 Content[LineIndex].data() - Lines[LineIndex].data();
1081 Whitespaces.replaceWhitespaceInToken(Tok: *Tokens[LineIndex], Offset,
1082 /*ReplaceChars=*/WhitespaceLength,
1083 /*PreviousPostfix=*/"",
1084 /*CurrentPrefix=*/ReflowPrefix,
1085 /*InPPDirective=*/false,
1086 /*Newlines=*/0,
1087 /*Spaces=*/0);
1088}
1089
1090void BreakableLineCommentSection::adaptStartOfLine(
1091 unsigned LineIndex, WhitespaceManager &Whitespaces) const {
1092 // If this is the first line of a token, we need to inform Whitespace Manager
1093 // about it: either adapt the whitespace range preceding it, or mark it as an
1094 // untouchable token.
1095 // This happens for instance here:
1096 // // line 1 \
1097 // // line 2
1098 if (LineIndex > 0 && Tokens[LineIndex] != Tokens[LineIndex - 1]) {
1099 // This is the first line for the current token, but no reflow with the
1100 // previous token is necessary. However, we still may need to adjust the
1101 // start column. Note that ContentColumn[LineIndex] is the expected
1102 // content column after a possible update to the prefix, hence the prefix
1103 // length change is included.
1104 unsigned LineColumn =
1105 ContentColumn[LineIndex] -
1106 (Content[LineIndex].data() - Lines[LineIndex].data()) +
1107 (OriginalPrefix[LineIndex].size() - Prefix[LineIndex].size());
1108
1109 // We always want to create a replacement instead of adding an untouchable
1110 // token, even if LineColumn is the same as the original column of the
1111 // token. This is because WhitespaceManager doesn't align trailing
1112 // comments if they are untouchable.
1113 Whitespaces.replaceWhitespace(Tok&: *Tokens[LineIndex],
1114 /*Newlines=*/1,
1115 /*Spaces=*/LineColumn,
1116 /*StartOfTokenColumn=*/LineColumn,
1117 /*IsAligned=*/true,
1118 /*InPPDirective=*/false);
1119 }
1120 if (OriginalPrefix[LineIndex] != Prefix[LineIndex]) {
1121 // Adjust the prefix if necessary.
1122 const auto SpacesToRemove = -std::min(a: PrefixSpaceChange[LineIndex], b: 0);
1123 const auto SpacesToAdd = std::max(a: PrefixSpaceChange[LineIndex], b: 0);
1124 Whitespaces.replaceWhitespaceInToken(
1125 Tok: tokenAt(LineIndex), Offset: OriginalPrefix[LineIndex].size() - SpacesToRemove,
1126 /*ReplaceChars=*/SpacesToRemove, PreviousPostfix: "", CurrentPrefix: "", /*InPPDirective=*/false,
1127 /*Newlines=*/0, /*Spaces=*/SpacesToAdd);
1128 }
1129}
1130
1131void BreakableLineCommentSection::updateNextToken(LineState &State) const {
1132 if (LastLineTok)
1133 State.NextToken = LastLineTok->Next;
1134}
1135
1136bool BreakableLineCommentSection::mayReflow(
1137 unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
1138 // Line comments have the indent as part of the prefix, so we need to
1139 // recompute the start of the line.
1140 StringRef IndentContent = Content[LineIndex];
1141 if (Lines[LineIndex].starts_with(Prefix: "//"))
1142 IndentContent = Lines[LineIndex].substr(Start: 2);
1143 // FIXME: Decide whether we want to reflow non-regular indents:
1144 // Currently, we only reflow when the OriginalPrefix[LineIndex] matches the
1145 // OriginalPrefix[LineIndex-1]. That means we don't reflow
1146 // // text that protrudes
1147 // // into text with different indent
1148 // We do reflow in that case in block comments.
1149 return LineIndex > 0 && AlwaysReflow &&
1150 !CommentPragmasRegex.match(String: IndentContent) &&
1151 mayReflowContent(Content: Content[LineIndex]) && !Tok.Finalized &&
1152 !switchesFormatting(Token: tokenAt(LineIndex)) &&
1153 OriginalPrefix[LineIndex] == OriginalPrefix[LineIndex - 1];
1154}
1155
1156} // namespace format
1157} // namespace clang
1158

source code of clang/lib/Format/BreakableToken.cpp