1 | //===--- BreakableToken.cpp - Format C++ code -----------------------------===// |
2 | // |
3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
4 | // See https://llvm.org/LICENSE.txt for license information. |
5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
6 | // |
7 | //===----------------------------------------------------------------------===// |
8 | /// |
9 | /// \file |
10 | /// Contains implementation of BreakableToken class and classes derived |
11 | /// from it. |
12 | /// |
13 | //===----------------------------------------------------------------------===// |
14 | |
15 | #include "BreakableToken.h" |
16 | #include "ContinuationIndenter.h" |
17 | #include "clang/Basic/CharInfo.h" |
18 | #include "clang/Format/Format.h" |
19 | #include "llvm/ADT/STLExtras.h" |
20 | #include "llvm/Support/Debug.h" |
21 | #include <algorithm> |
22 | |
23 | #define DEBUG_TYPE "format-token-breaker" |
24 | |
25 | namespace clang { |
26 | namespace format { |
27 | |
28 | static constexpr StringRef Blanks = " \t\v\f\r" ; |
29 | 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 | |
42 | static StringRef (StringRef , |
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( |
53 | llvm::is_sorted(KnownPrefixes, [](StringRef Lhs, StringRef Rhs) noexcept { |
54 | return Lhs.size() > Rhs.size(); |
55 | })); |
56 | |
57 | for (StringRef KnownPrefix : KnownPrefixes) { |
58 | if (Comment.starts_with(Prefix: KnownPrefix)) { |
59 | const auto PrefixLength = |
60 | Comment.find_first_not_of(C: ' ', From: KnownPrefix.size()); |
61 | return Comment.substr(Start: 0, N: PrefixLength); |
62 | } |
63 | } |
64 | return {}; |
65 | } |
66 | |
67 | static BreakableToken::Split |
68 | (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(FirstChar: Text[MaxSplitBytes], Encoding); |
85 | NumChars += encoding::columnWidthWithTabs( |
86 | Text: Text.substr(Start: MaxSplitBytes, N: BytesInChar), StartColumn: ContentStartColumn + NumChars, |
87 | 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.isJavaScript()) { |
95 | StringRef::size_type SpaceOffset = |
96 | Text.find_first_of(Chars: Blanks, From: 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(Chars: Blanks, From: 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(Chars: Blanks, From: SpaceOffset); |
116 | if (LastNonBlank != StringRef::npos && Text[LastNonBlank] == '\\') { |
117 | SpaceOffset = Text.find_last_of(Chars: Blanks, From: 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(String: Text.substr(Start: SpaceOffset).ltrim(Chars: Blanks))) { |
125 | SpaceOffset = Text.find_last_of(Chars: Blanks, From: SpaceOffset); |
126 | continue; |
127 | } |
128 | |
129 | // Avoid ever breaking before a @tag or a { in JavaScript. |
130 | if (Style.isJavaScript() && SpaceOffset + 1 < Text.size() && |
131 | (Text[SpaceOffset + 1] == '{' || Text[SpaceOffset + 1] == '@')) { |
132 | SpaceOffset = Text.find_last_of(Chars: Blanks, From: SpaceOffset); |
133 | continue; |
134 | } |
135 | |
136 | break; |
137 | } |
138 | |
139 | if (SpaceOffset == StringRef::npos || |
140 | // Don't break at leading whitespace. |
141 | Text.find_last_not_of(Chars: Blanks, From: SpaceOffset) == StringRef::npos) { |
142 | // Make sure that we don't break at leading whitespace that |
143 | // reaches past MaxSplit. |
144 | StringRef::size_type FirstNonWhitespace = Text.find_first_not_of(Chars: Blanks); |
145 | if (FirstNonWhitespace == StringRef::npos) { |
146 | // If the comment is only whitespace, we cannot split. |
147 | return BreakableToken::Split(StringRef::npos, 0); |
148 | } |
149 | SpaceOffset = Text.find_first_of( |
150 | Chars: Blanks, From: std::max<unsigned>(a: MaxSplitBytes, b: 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(Start: 0, N: SpaceOffset).rtrim(Chars: Blanks); |
160 | StringRef AfterCut = Text.substr(Start: 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(Chars: 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 |
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(FirstChar: Text[0], Encoding); |
190 | Chars += encoding::columnWidthWithTabs( |
191 | Text: Text.substr(Start: 0, N: Advance), StartColumn: UsedColumns + Chars, TabWidth, Encoding); |
192 | } |
193 | |
194 | if (Chars > MaxSplit || Text.size() <= Advance) |
195 | break; |
196 | |
197 | if (IsBlank(C: Text[0])) |
198 | SpaceOffset = SplitPoint; |
199 | if (Text[0] == '/') |
200 | SlashOffset = SplitPoint; |
201 | if (Advance == 1 && !isAlphanumeric(c: Text[0])) |
202 | WordStartOffset = SplitPoint; |
203 | |
204 | SplitPoint += Advance; |
205 | Text = Text.substr(Start: 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 | |
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(Start: 2).ltrim(); |
223 | return Content.starts_with(Prefix: "clang-format on" ) || |
224 | Content.starts_with(Prefix: "clang-format off" ); |
225 | } |
226 | |
227 | unsigned |
228 | 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 | |
244 | unsigned BreakableStringLiteral::getLineCount() const { return 1; } |
245 | |
246 | 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 |
255 | BreakableStringLiteral::getRemainingLength(unsigned LineIndex, unsigned Offset, |
256 | unsigned StartColumn) const { |
257 | return UnbreakableTailLength + Postfix.size() + |
258 | encoding::columnWidthWithTabs(Text: Line.substr(Start: Offset), StartColumn, |
259 | TabWidth: Style.TabWidth, Encoding); |
260 | } |
261 | |
262 | unsigned BreakableStringLiteral::getContentStartColumn(unsigned LineIndex, |
263 | bool Break) const { |
264 | return StartColumn + Prefix.size(); |
265 | } |
266 | |
267 | BreakableStringLiteral::BreakableStringLiteral( |
268 | const FormatToken &Tok, unsigned StartColumn, StringRef Prefix, |
269 | StringRef Postfix, unsigned UnbreakableTailLength, bool InPPDirective, |
270 | encoding::Encoding Encoding, const FormatStyle &Style) |
271 | : BreakableToken(Tok, InPPDirective, Encoding, Style), |
272 | StartColumn(StartColumn), Prefix(Prefix), Postfix(Postfix), |
273 | UnbreakableTailLength(UnbreakableTailLength) { |
274 | assert(Tok.TokenText.starts_with(Prefix) && Tok.TokenText.ends_with(Postfix)); |
275 | Line = Tok.TokenText.substr( |
276 | Start: Prefix.size(), N: Tok.TokenText.size() - Prefix.size() - Postfix.size()); |
277 | } |
278 | |
279 | BreakableToken::Split BreakableStringLiteral::getSplit( |
280 | unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit, |
281 | unsigned ContentStartColumn, const llvm::Regex &) const { |
282 | return getStringSplit(Text: Line.substr(Start: TailOffset), UsedColumns: ContentStartColumn, |
283 | ColumnLimit: ColumnLimit - Postfix.size(), TabWidth: Style.TabWidth, Encoding); |
284 | } |
285 | |
286 | void BreakableStringLiteral::insertBreak(unsigned LineIndex, |
287 | unsigned TailOffset, Split Split, |
288 | unsigned ContentIndent, |
289 | WhitespaceManager &Whitespaces) const { |
290 | Whitespaces.replaceWhitespaceInToken( |
291 | Tok, Offset: Prefix.size() + TailOffset + Split.first, ReplaceChars: Split.second, PreviousPostfix: Postfix, |
292 | CurrentPrefix: Prefix, InPPDirective, Newlines: 1, Spaces: StartColumn); |
293 | } |
294 | |
295 | BreakableStringLiteralUsingOperators::BreakableStringLiteralUsingOperators( |
296 | const FormatToken &Tok, QuoteStyleType QuoteStyle, bool UnindentPlus, |
297 | unsigned StartColumn, unsigned UnbreakableTailLength, bool InPPDirective, |
298 | encoding::Encoding Encoding, const FormatStyle &Style) |
299 | : BreakableStringLiteral( |
300 | Tok, StartColumn, /*Prefix=*/QuoteStyle == SingleQuotes ? "'" |
301 | : QuoteStyle == AtDoubleQuotes ? "@\"" |
302 | : "\"" , |
303 | /*Postfix=*/QuoteStyle == SingleQuotes ? "'" : "\"" , |
304 | UnbreakableTailLength, InPPDirective, Encoding, Style), |
305 | BracesNeeded(Tok.isNot(Kind: TT_StringInConcatenation)), |
306 | QuoteStyle(QuoteStyle) { |
307 | // Find the replacement text for inserting braces and quotes and line breaks. |
308 | // We don't create an allocated string concatenated from parts here because it |
309 | // has to outlive the BreakableStringliteral object. The brace replacements |
310 | // include a quote so that WhitespaceManager can tell it apart from whitespace |
311 | // replacements between the string and surrounding tokens. |
312 | |
313 | // The option is not implemented in JavaScript. |
314 | bool SignOnNewLine = |
315 | !Style.isJavaScript() && |
316 | Style.BreakBeforeBinaryOperators != FormatStyle::BOS_None; |
317 | |
318 | if (Style.isVerilog()) { |
319 | // In Verilog, all strings are quoted by double quotes, joined by commas, |
320 | // and wrapped in braces. The comma is always before the newline. |
321 | assert(QuoteStyle == DoubleQuotes); |
322 | LeftBraceQuote = Style.Cpp11BracedListStyle ? "{\"" : "{ \"" ; |
323 | RightBraceQuote = Style.Cpp11BracedListStyle ? "\"}" : "\" }" ; |
324 | Postfix = "\"," ; |
325 | Prefix = "\"" ; |
326 | } else { |
327 | // The plus sign may be on either line. And also C# and JavaScript have |
328 | // several quoting styles. |
329 | if (QuoteStyle == SingleQuotes) { |
330 | LeftBraceQuote = Style.SpacesInParensOptions.Other ? "( '" : "('" ; |
331 | RightBraceQuote = Style.SpacesInParensOptions.Other ? "' )" : "')" ; |
332 | Postfix = SignOnNewLine ? "'" : "' +" ; |
333 | Prefix = SignOnNewLine ? "+ '" : "'" ; |
334 | } else { |
335 | if (QuoteStyle == AtDoubleQuotes) { |
336 | LeftBraceQuote = Style.SpacesInParensOptions.Other ? "( @" : "(@" ; |
337 | Prefix = SignOnNewLine ? "+ @\"" : "@\"" ; |
338 | } else { |
339 | LeftBraceQuote = Style.SpacesInParensOptions.Other ? "( \"" : "(\"" ; |
340 | Prefix = SignOnNewLine ? "+ \"" : "\"" ; |
341 | } |
342 | RightBraceQuote = Style.SpacesInParensOptions.Other ? "\" )" : "\")" ; |
343 | Postfix = SignOnNewLine ? "\"" : "\" +" ; |
344 | } |
345 | } |
346 | |
347 | // Following lines are indented by the width of the brace and space if any. |
348 | ContinuationIndent = BracesNeeded ? LeftBraceQuote.size() - 1 : 0; |
349 | // The plus sign may need to be unindented depending on the style. |
350 | // FIXME: Add support for DontAlign. |
351 | if (!Style.isVerilog() && SignOnNewLine && !BracesNeeded && UnindentPlus && |
352 | Style.AlignOperands == FormatStyle::OAS_AlignAfterOperator) { |
353 | ContinuationIndent -= 2; |
354 | } |
355 | } |
356 | |
357 | unsigned BreakableStringLiteralUsingOperators::getRemainingLength( |
358 | unsigned LineIndex, unsigned Offset, unsigned StartColumn) const { |
359 | return UnbreakableTailLength + (BracesNeeded ? RightBraceQuote.size() : 1) + |
360 | encoding::columnWidthWithTabs(Text: Line.substr(Start: Offset), StartColumn, |
361 | TabWidth: Style.TabWidth, Encoding); |
362 | } |
363 | |
364 | unsigned |
365 | BreakableStringLiteralUsingOperators::getContentStartColumn(unsigned LineIndex, |
366 | bool Break) const { |
367 | return std::max( |
368 | a: 0, |
369 | b: static_cast<int>(StartColumn) + |
370 | (Break ? ContinuationIndent + static_cast<int>(Prefix.size()) |
371 | : (BracesNeeded ? static_cast<int>(LeftBraceQuote.size()) - 1 |
372 | : 0) + |
373 | (QuoteStyle == AtDoubleQuotes ? 2 : 1))); |
374 | } |
375 | |
376 | void BreakableStringLiteralUsingOperators::insertBreak( |
377 | unsigned LineIndex, unsigned TailOffset, Split Split, |
378 | unsigned ContentIndent, WhitespaceManager &Whitespaces) const { |
379 | Whitespaces.replaceWhitespaceInToken( |
380 | Tok, /*Offset=*/(QuoteStyle == AtDoubleQuotes ? 2 : 1) + TailOffset + |
381 | Split.first, |
382 | /*ReplaceChars=*/Split.second, /*PreviousPostfix=*/Postfix, |
383 | /*CurrentPrefix=*/Prefix, InPPDirective, /*NewLines=*/Newlines: 1, |
384 | /*Spaces=*/ |
385 | std::max(a: 0, b: static_cast<int>(StartColumn) + ContinuationIndent)); |
386 | } |
387 | |
388 | void BreakableStringLiteralUsingOperators::updateAfterBroken( |
389 | WhitespaceManager &Whitespaces) const { |
390 | // Add the braces required for breaking the token if they are needed. |
391 | if (!BracesNeeded) |
392 | return; |
393 | |
394 | // To add a brace or parenthesis, we replace the quote (or the at sign) with a |
395 | // brace and another quote. This is because the rest of the program requires |
396 | // one replacement for each source range. If we replace the empty strings |
397 | // around the string, it may conflict with whitespace replacements between the |
398 | // string and adjacent tokens. |
399 | Whitespaces.replaceWhitespaceInToken( |
400 | Tok, /*Offset=*/0, /*ReplaceChars=*/1, /*PreviousPostfix=*/"" , |
401 | /*CurrentPrefix=*/LeftBraceQuote, InPPDirective, /*NewLines=*/Newlines: 0, |
402 | /*Spaces=*/0); |
403 | Whitespaces.replaceWhitespaceInToken( |
404 | Tok, /*Offset=*/Tok.TokenText.size() - 1, /*ReplaceChars=*/1, |
405 | /*PreviousPostfix=*/RightBraceQuote, |
406 | /*CurrentPrefix=*/"" , InPPDirective, /*NewLines=*/Newlines: 0, /*Spaces=*/0); |
407 | } |
408 | |
409 | BreakableComment::(const FormatToken &Token, |
410 | unsigned StartColumn, bool InPPDirective, |
411 | encoding::Encoding Encoding, |
412 | const FormatStyle &Style) |
413 | : BreakableToken(Token, InPPDirective, Encoding, Style), |
414 | StartColumn(StartColumn) {} |
415 | |
416 | unsigned BreakableComment::() const { return Lines.size(); } |
417 | |
418 | BreakableToken::Split |
419 | BreakableComment::(unsigned LineIndex, unsigned TailOffset, |
420 | unsigned ColumnLimit, unsigned ContentStartColumn, |
421 | const llvm::Regex &) const { |
422 | // Don't break lines matching the comment pragmas regex. |
423 | if (CommentPragmasRegex.match(String: Content[LineIndex])) |
424 | return Split(StringRef::npos, 0); |
425 | return getCommentSplit(Text: Content[LineIndex].substr(Start: TailOffset), |
426 | ContentStartColumn, ColumnLimit, TabWidth: Style.TabWidth, |
427 | Encoding, Style); |
428 | } |
429 | |
430 | void BreakableComment::( |
431 | unsigned LineIndex, unsigned TailOffset, Split Split, |
432 | WhitespaceManager &Whitespaces) const { |
433 | StringRef Text = Content[LineIndex].substr(Start: TailOffset); |
434 | // Text is relative to the content line, but Whitespaces operates relative to |
435 | // the start of the corresponding token, so compute the start of the Split |
436 | // that needs to be compressed into a single space relative to the start of |
437 | // its token. |
438 | unsigned BreakOffsetInToken = |
439 | Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first; |
440 | unsigned CharsToRemove = Split.second; |
441 | Whitespaces.replaceWhitespaceInToken( |
442 | Tok: tokenAt(LineIndex), Offset: BreakOffsetInToken, ReplaceChars: CharsToRemove, PreviousPostfix: "" , CurrentPrefix: "" , |
443 | /*InPPDirective=*/false, /*Newlines=*/0, /*Spaces=*/1); |
444 | } |
445 | |
446 | const FormatToken &BreakableComment::(unsigned LineIndex) const { |
447 | return Tokens[LineIndex] ? *Tokens[LineIndex] : Tok; |
448 | } |
449 | |
450 | static bool mayReflowContent(StringRef Content) { |
451 | Content = Content.trim(Chars: Blanks); |
452 | // Lines starting with '@' or '\' commonly have special meaning. |
453 | // Lines starting with '-', '-#', '+' or '*' are bulleted/numbered lists. |
454 | bool hasSpecialMeaningPrefix = false; |
455 | for (StringRef Prefix : |
456 | {"@" , "\\" , "TODO" , "FIXME" , "XXX" , "-# " , "- " , "+ " , "* " }) { |
457 | if (Content.starts_with(Prefix)) { |
458 | hasSpecialMeaningPrefix = true; |
459 | break; |
460 | } |
461 | } |
462 | |
463 | // Numbered lists may also start with a number followed by '.' |
464 | // To avoid issues if a line starts with a number which is actually the end |
465 | // of a previous line, we only consider numbers with up to 2 digits. |
466 | static const auto kNumberedListRegexp = llvm::Regex("^[1-9][0-9]?\\. " ); |
467 | hasSpecialMeaningPrefix = |
468 | hasSpecialMeaningPrefix || kNumberedListRegexp.match(String: Content); |
469 | |
470 | // Simple heuristic for what to reflow: content should contain at least two |
471 | // characters and either the first or second character must be |
472 | // non-punctuation. |
473 | return Content.size() >= 2 && !hasSpecialMeaningPrefix && |
474 | !Content.ends_with(Suffix: "\\" ) && |
475 | // Note that this is UTF-8 safe, since if isPunctuation(Content[0]) is |
476 | // true, then the first code point must be 1 byte long. |
477 | (!isPunctuation(c: Content[0]) || !isPunctuation(c: Content[1])); |
478 | } |
479 | |
480 | BreakableBlockComment::( |
481 | const FormatToken &Token, unsigned StartColumn, |
482 | unsigned OriginalStartColumn, bool FirstInLine, bool InPPDirective, |
483 | encoding::Encoding Encoding, const FormatStyle &Style, bool UseCRLF) |
484 | : BreakableComment(Token, StartColumn, InPPDirective, Encoding, Style), |
485 | DelimitersOnNewline(false), |
486 | UnbreakableTailLength(Token.UnbreakableTailLength) { |
487 | assert(Tok.is(TT_BlockComment) && |
488 | "block comment section must start with a block comment" ); |
489 | |
490 | StringRef TokenText(Tok.TokenText); |
491 | assert(TokenText.starts_with("/*" ) && TokenText.ends_with("*/" )); |
492 | TokenText.substr(Start: 2, N: TokenText.size() - 4) |
493 | .split(A&: Lines, Separator: UseCRLF ? "\r\n" : "\n" ); |
494 | |
495 | int IndentDelta = StartColumn - OriginalStartColumn; |
496 | Content.resize(N: Lines.size()); |
497 | Content[0] = Lines[0]; |
498 | ContentColumn.resize(N: Lines.size()); |
499 | // Account for the initial '/*'. |
500 | ContentColumn[0] = StartColumn + 2; |
501 | Tokens.resize(N: Lines.size()); |
502 | for (size_t i = 1; i < Lines.size(); ++i) |
503 | adjustWhitespace(LineIndex: i, IndentDelta); |
504 | |
505 | // Align decorations with the column of the star on the first line, |
506 | // that is one column after the start "/*". |
507 | DecorationColumn = StartColumn + 1; |
508 | |
509 | // Account for comment decoration patterns like this: |
510 | // |
511 | // /* |
512 | // ** blah blah blah |
513 | // */ |
514 | if (Lines.size() >= 2 && Content[1].starts_with(Prefix: "**" ) && |
515 | static_cast<unsigned>(ContentColumn[1]) == StartColumn) { |
516 | DecorationColumn = StartColumn; |
517 | } |
518 | |
519 | Decoration = "* " ; |
520 | if (Lines.size() == 1 && !FirstInLine) { |
521 | // Comments for which FirstInLine is false can start on arbitrary column, |
522 | // and available horizontal space can be too small to align consecutive |
523 | // lines with the first one. |
524 | // FIXME: We could, probably, align them to current indentation level, but |
525 | // now we just wrap them without stars. |
526 | Decoration = "" ; |
527 | } |
528 | for (size_t i = 1, e = Content.size(); i < e && !Decoration.empty(); ++i) { |
529 | const StringRef &Text = Content[i]; |
530 | if (i + 1 == e) { |
531 | // If the last line is empty, the closing "*/" will have a star. |
532 | if (Text.empty()) |
533 | break; |
534 | } else if (!Text.empty() && Decoration.starts_with(Prefix: Text)) { |
535 | continue; |
536 | } |
537 | while (!Text.starts_with(Prefix: Decoration)) |
538 | Decoration = Decoration.drop_back(N: 1); |
539 | } |
540 | |
541 | LastLineNeedsDecoration = true; |
542 | IndentAtLineBreak = ContentColumn[0] + 1; |
543 | for (size_t i = 1, e = Lines.size(); i < e; ++i) { |
544 | if (Content[i].empty()) { |
545 | if (i + 1 == e) { |
546 | // Empty last line means that we already have a star as a part of the |
547 | // trailing */. We also need to preserve whitespace, so that */ is |
548 | // correctly indented. |
549 | LastLineNeedsDecoration = false; |
550 | // Align the star in the last '*/' with the stars on the previous lines. |
551 | if (e >= 2 && !Decoration.empty()) |
552 | ContentColumn[i] = DecorationColumn; |
553 | } else if (Decoration.empty()) { |
554 | // For all other lines, set the start column to 0 if they're empty, so |
555 | // we do not insert trailing whitespace anywhere. |
556 | ContentColumn[i] = 0; |
557 | } |
558 | continue; |
559 | } |
560 | |
561 | // The first line already excludes the star. |
562 | // The last line excludes the star if LastLineNeedsDecoration is false. |
563 | // For all other lines, adjust the line to exclude the star and |
564 | // (optionally) the first whitespace. |
565 | unsigned DecorationSize = Decoration.starts_with(Prefix: Content[i]) |
566 | ? Content[i].size() |
567 | : Decoration.size(); |
568 | if (DecorationSize) |
569 | ContentColumn[i] = DecorationColumn + DecorationSize; |
570 | Content[i] = Content[i].substr(Start: DecorationSize); |
571 | if (!Decoration.starts_with(Prefix: Content[i])) { |
572 | IndentAtLineBreak = |
573 | std::min<int>(a: IndentAtLineBreak, b: std::max(a: 0, b: ContentColumn[i])); |
574 | } |
575 | } |
576 | IndentAtLineBreak = std::max<unsigned>(a: IndentAtLineBreak, b: Decoration.size()); |
577 | |
578 | // Detect a multiline jsdoc comment and set DelimitersOnNewline in that case. |
579 | if (Style.isJavaScript() || Style.Language == FormatStyle::LK_Java) { |
580 | if ((Lines[0] == "*" || Lines[0].starts_with(Prefix: "* " )) && Lines.size() > 1) { |
581 | // This is a multiline jsdoc comment. |
582 | DelimitersOnNewline = true; |
583 | } else if (Lines[0].starts_with(Prefix: "* " ) && Lines.size() == 1) { |
584 | // Detect a long single-line comment, like: |
585 | // /** long long long */ |
586 | // Below, '2' is the width of '*/'. |
587 | unsigned EndColumn = |
588 | ContentColumn[0] + |
589 | encoding::columnWidthWithTabs(Text: Lines[0], StartColumn: ContentColumn[0], |
590 | TabWidth: Style.TabWidth, Encoding) + |
591 | 2; |
592 | DelimitersOnNewline = EndColumn > Style.ColumnLimit; |
593 | } |
594 | } |
595 | |
596 | LLVM_DEBUG({ |
597 | llvm::dbgs() << "IndentAtLineBreak " << IndentAtLineBreak << "\n" ; |
598 | llvm::dbgs() << "DelimitersOnNewline " << DelimitersOnNewline << "\n" ; |
599 | for (size_t i = 0; i < Lines.size(); ++i) { |
600 | llvm::dbgs() << i << " |" << Content[i] << "| " |
601 | << "CC=" << ContentColumn[i] << "| " |
602 | << "IN=" << (Content[i].data() - Lines[i].data()) << "\n" ; |
603 | } |
604 | }); |
605 | } |
606 | |
607 | BreakableToken::Split BreakableBlockComment::( |
608 | unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit, |
609 | unsigned ContentStartColumn, const llvm::Regex &) const { |
610 | // Don't break lines matching the comment pragmas regex. |
611 | if (CommentPragmasRegex.match(String: Content[LineIndex])) |
612 | return Split(StringRef::npos, 0); |
613 | return getCommentSplit(Text: Content[LineIndex].substr(Start: TailOffset), |
614 | ContentStartColumn, ColumnLimit, TabWidth: Style.TabWidth, |
615 | Encoding, Style, DecorationEndsWithStar: Decoration.ends_with(Suffix: "*" )); |
616 | } |
617 | |
618 | void BreakableBlockComment::(unsigned LineIndex, |
619 | int IndentDelta) { |
620 | // When in a preprocessor directive, the trailing backslash in a block comment |
621 | // is not needed, but can serve a purpose of uniformity with necessary escaped |
622 | // newlines outside the comment. In this case we remove it here before |
623 | // trimming the trailing whitespace. The backslash will be re-added later when |
624 | // inserting a line break. |
625 | size_t EndOfPreviousLine = Lines[LineIndex - 1].size(); |
626 | if (InPPDirective && Lines[LineIndex - 1].ends_with(Suffix: "\\" )) |
627 | --EndOfPreviousLine; |
628 | |
629 | // Calculate the end of the non-whitespace text in the previous line. |
630 | EndOfPreviousLine = |
631 | Lines[LineIndex - 1].find_last_not_of(Chars: Blanks, From: EndOfPreviousLine); |
632 | if (EndOfPreviousLine == StringRef::npos) |
633 | EndOfPreviousLine = 0; |
634 | else |
635 | ++EndOfPreviousLine; |
636 | // Calculate the start of the non-whitespace text in the current line. |
637 | size_t StartOfLine = Lines[LineIndex].find_first_not_of(Chars: Blanks); |
638 | if (StartOfLine == StringRef::npos) |
639 | StartOfLine = Lines[LineIndex].size(); |
640 | |
641 | StringRef Whitespace = Lines[LineIndex].substr(Start: 0, N: StartOfLine); |
642 | // Adjust Lines to only contain relevant text. |
643 | size_t PreviousContentOffset = |
644 | Content[LineIndex - 1].data() - Lines[LineIndex - 1].data(); |
645 | Content[LineIndex - 1] = Lines[LineIndex - 1].substr( |
646 | Start: PreviousContentOffset, N: EndOfPreviousLine - PreviousContentOffset); |
647 | Content[LineIndex] = Lines[LineIndex].substr(Start: StartOfLine); |
648 | |
649 | // Adjust the start column uniformly across all lines. |
650 | ContentColumn[LineIndex] = |
651 | encoding::columnWidthWithTabs(Text: Whitespace, StartColumn: 0, TabWidth: Style.TabWidth, Encoding) + |
652 | IndentDelta; |
653 | } |
654 | |
655 | unsigned BreakableBlockComment::(unsigned LineIndex, |
656 | unsigned Offset, |
657 | StringRef::size_type Length, |
658 | unsigned StartColumn) const { |
659 | return encoding::columnWidthWithTabs( |
660 | Text: Content[LineIndex].substr(Start: Offset, N: Length), StartColumn, TabWidth: Style.TabWidth, |
661 | Encoding); |
662 | } |
663 | |
664 | unsigned BreakableBlockComment::getRemainingLength(unsigned LineIndex, |
665 | unsigned Offset, |
666 | unsigned StartColumn) const { |
667 | unsigned LineLength = |
668 | UnbreakableTailLength + |
669 | getRangeLength(LineIndex, Offset, Length: StringRef::npos, StartColumn); |
670 | if (LineIndex + 1 == Lines.size()) { |
671 | LineLength += 2; |
672 | // We never need a decoration when breaking just the trailing "*/" postfix. |
673 | bool HasRemainingText = Offset < Content[LineIndex].size(); |
674 | if (!HasRemainingText) { |
675 | bool HasDecoration = Lines[LineIndex].ltrim().starts_with(Prefix: Decoration); |
676 | if (HasDecoration) |
677 | LineLength -= Decoration.size(); |
678 | } |
679 | } |
680 | return LineLength; |
681 | } |
682 | |
683 | unsigned BreakableBlockComment::getContentStartColumn(unsigned LineIndex, |
684 | bool Break) const { |
685 | if (Break) |
686 | return IndentAtLineBreak; |
687 | return std::max(a: 0, b: ContentColumn[LineIndex]); |
688 | } |
689 | |
690 | const llvm::StringSet<> |
691 | BreakableBlockComment:: = { |
692 | "@param" , "@return" , "@returns" , "@throws" , "@type" , "@template" , |
693 | "@see" , "@deprecated" , "@define" , "@exports" , "@mods" , "@private" , |
694 | }; |
695 | |
696 | unsigned BreakableBlockComment::(unsigned LineIndex) const { |
697 | if (Style.Language != FormatStyle::LK_Java && !Style.isJavaScript()) |
698 | return 0; |
699 | // The content at LineIndex 0 of a comment like: |
700 | // /** line 0 */ |
701 | // is "* line 0", so we need to skip over the decoration in that case. |
702 | StringRef ContentWithNoDecoration = Content[LineIndex]; |
703 | if (LineIndex == 0 && ContentWithNoDecoration.starts_with(Prefix: "*" )) |
704 | ContentWithNoDecoration = ContentWithNoDecoration.substr(Start: 1).ltrim(Chars: Blanks); |
705 | StringRef FirstWord = ContentWithNoDecoration.substr( |
706 | Start: 0, N: ContentWithNoDecoration.find_first_of(Chars: Blanks)); |
707 | if (ContentIndentingJavadocAnnotations.contains(key: FirstWord)) |
708 | return Style.ContinuationIndentWidth; |
709 | return 0; |
710 | } |
711 | |
712 | void BreakableBlockComment::(unsigned LineIndex, unsigned TailOffset, |
713 | Split Split, unsigned ContentIndent, |
714 | WhitespaceManager &Whitespaces) const { |
715 | StringRef Text = Content[LineIndex].substr(Start: TailOffset); |
716 | StringRef Prefix = Decoration; |
717 | // We need this to account for the case when we have a decoration "* " for all |
718 | // the lines except for the last one, where the star in "*/" acts as a |
719 | // decoration. |
720 | unsigned LocalIndentAtLineBreak = IndentAtLineBreak; |
721 | if (LineIndex + 1 == Lines.size() && |
722 | Text.size() == Split.first + Split.second) { |
723 | // For the last line we need to break before "*/", but not to add "* ". |
724 | Prefix = "" ; |
725 | if (LocalIndentAtLineBreak >= 2) |
726 | LocalIndentAtLineBreak -= 2; |
727 | } |
728 | // The split offset is from the beginning of the line. Convert it to an offset |
729 | // from the beginning of the token text. |
730 | unsigned BreakOffsetInToken = |
731 | Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first; |
732 | unsigned CharsToRemove = Split.second; |
733 | assert(LocalIndentAtLineBreak >= Prefix.size()); |
734 | std::string PrefixWithTrailingIndent = std::string(Prefix); |
735 | PrefixWithTrailingIndent.append(n: ContentIndent, c: ' '); |
736 | Whitespaces.replaceWhitespaceInToken( |
737 | Tok: tokenAt(LineIndex), Offset: BreakOffsetInToken, ReplaceChars: CharsToRemove, PreviousPostfix: "" , |
738 | CurrentPrefix: PrefixWithTrailingIndent, InPPDirective, /*Newlines=*/1, |
739 | /*Spaces=*/LocalIndentAtLineBreak + ContentIndent - |
740 | PrefixWithTrailingIndent.size()); |
741 | } |
742 | |
743 | BreakableToken::Split BreakableBlockComment::( |
744 | unsigned LineIndex, const llvm::Regex &) const { |
745 | if (!mayReflow(LineIndex, CommentPragmasRegex)) |
746 | return Split(StringRef::npos, 0); |
747 | |
748 | // If we're reflowing into a line with content indent, only reflow the next |
749 | // line if its starting whitespace matches the content indent. |
750 | size_t Trimmed = Content[LineIndex].find_first_not_of(Chars: Blanks); |
751 | if (LineIndex) { |
752 | unsigned PreviousContentIndent = getContentIndent(LineIndex: LineIndex - 1); |
753 | if (PreviousContentIndent && Trimmed != StringRef::npos && |
754 | Trimmed != PreviousContentIndent) { |
755 | return Split(StringRef::npos, 0); |
756 | } |
757 | } |
758 | |
759 | return Split(0, Trimmed != StringRef::npos ? Trimmed : 0); |
760 | } |
761 | |
762 | bool BreakableBlockComment::() const { |
763 | // A break is introduced when we want delimiters on newline. |
764 | return DelimitersOnNewline && |
765 | Lines[0].substr(Start: 1).find_first_not_of(Chars: Blanks) != StringRef::npos; |
766 | } |
767 | |
768 | void BreakableBlockComment::(unsigned LineIndex, |
769 | WhitespaceManager &Whitespaces) const { |
770 | StringRef TrimmedContent = Content[LineIndex].ltrim(Chars: Blanks); |
771 | // Here we need to reflow. |
772 | assert(Tokens[LineIndex - 1] == Tokens[LineIndex] && |
773 | "Reflowing whitespace within a token" ); |
774 | // This is the offset of the end of the last line relative to the start of |
775 | // the token text in the token. |
776 | unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() + |
777 | Content[LineIndex - 1].size() - |
778 | tokenAt(LineIndex).TokenText.data(); |
779 | unsigned WhitespaceLength = TrimmedContent.data() - |
780 | tokenAt(LineIndex).TokenText.data() - |
781 | WhitespaceOffsetInToken; |
782 | Whitespaces.replaceWhitespaceInToken( |
783 | Tok: tokenAt(LineIndex), Offset: WhitespaceOffsetInToken, |
784 | /*ReplaceChars=*/WhitespaceLength, /*PreviousPostfix=*/"" , |
785 | /*CurrentPrefix=*/ReflowPrefix, InPPDirective, /*Newlines=*/0, |
786 | /*Spaces=*/0); |
787 | } |
788 | |
789 | void BreakableBlockComment::( |
790 | unsigned LineIndex, WhitespaceManager &Whitespaces) const { |
791 | if (LineIndex == 0) { |
792 | if (DelimitersOnNewline) { |
793 | // Since we're breaking at index 1 below, the break position and the |
794 | // break length are the same. |
795 | // Note: this works because getCommentSplit is careful never to split at |
796 | // the beginning of a line. |
797 | size_t BreakLength = Lines[0].substr(Start: 1).find_first_not_of(Chars: Blanks); |
798 | if (BreakLength != StringRef::npos) { |
799 | insertBreak(LineIndex, TailOffset: 0, Split: Split(1, BreakLength), /*ContentIndent=*/0, |
800 | Whitespaces); |
801 | } |
802 | } |
803 | return; |
804 | } |
805 | // Here no reflow with the previous line will happen. |
806 | // Fix the decoration of the line at LineIndex. |
807 | StringRef Prefix = Decoration; |
808 | if (Content[LineIndex].empty()) { |
809 | if (LineIndex + 1 == Lines.size()) { |
810 | if (!LastLineNeedsDecoration) { |
811 | // If the last line was empty, we don't need a prefix, as the */ will |
812 | // line up with the decoration (if it exists). |
813 | Prefix = "" ; |
814 | } |
815 | } else if (!Decoration.empty()) { |
816 | // For other empty lines, if we do have a decoration, adapt it to not |
817 | // contain a trailing whitespace. |
818 | Prefix = Prefix.substr(Start: 0, N: 1); |
819 | } |
820 | } else if (ContentColumn[LineIndex] == 1) { |
821 | // This line starts immediately after the decorating *. |
822 | Prefix = Prefix.substr(Start: 0, N: 1); |
823 | } |
824 | // This is the offset of the end of the last line relative to the start of the |
825 | // token text in the token. |
826 | unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() + |
827 | Content[LineIndex - 1].size() - |
828 | tokenAt(LineIndex).TokenText.data(); |
829 | unsigned WhitespaceLength = Content[LineIndex].data() - |
830 | tokenAt(LineIndex).TokenText.data() - |
831 | WhitespaceOffsetInToken; |
832 | Whitespaces.replaceWhitespaceInToken( |
833 | Tok: tokenAt(LineIndex), Offset: WhitespaceOffsetInToken, ReplaceChars: WhitespaceLength, PreviousPostfix: "" , CurrentPrefix: Prefix, |
834 | InPPDirective, /*Newlines=*/1, Spaces: ContentColumn[LineIndex] - Prefix.size()); |
835 | } |
836 | |
837 | BreakableToken::Split |
838 | BreakableBlockComment::(unsigned TailOffset) const { |
839 | if (DelimitersOnNewline) { |
840 | // Replace the trailing whitespace of the last line with a newline. |
841 | // In case the last line is empty, the ending '*/' is already on its own |
842 | // line. |
843 | StringRef Line = Content.back().substr(Start: TailOffset); |
844 | StringRef TrimmedLine = Line.rtrim(Chars: Blanks); |
845 | if (!TrimmedLine.empty()) |
846 | return Split(TrimmedLine.size(), Line.size() - TrimmedLine.size()); |
847 | } |
848 | return Split(StringRef::npos, 0); |
849 | } |
850 | |
851 | bool BreakableBlockComment::( |
852 | unsigned LineIndex, const llvm::Regex &) const { |
853 | // Content[LineIndex] may exclude the indent after the '*' decoration. In that |
854 | // case, we compute the start of the comment pragma manually. |
855 | StringRef IndentContent = Content[LineIndex]; |
856 | if (Lines[LineIndex].ltrim(Chars: Blanks).starts_with(Prefix: "*" )) |
857 | IndentContent = Lines[LineIndex].ltrim(Chars: Blanks).substr(Start: 1); |
858 | return LineIndex > 0 && !CommentPragmasRegex.match(String: IndentContent) && |
859 | mayReflowContent(Content: Content[LineIndex]) && !Tok.Finalized && |
860 | !switchesFormatting(Token: tokenAt(LineIndex)); |
861 | } |
862 | |
863 | BreakableLineCommentSection::( |
864 | const FormatToken &Token, unsigned StartColumn, bool InPPDirective, |
865 | encoding::Encoding Encoding, const FormatStyle &Style) |
866 | : BreakableComment(Token, StartColumn, InPPDirective, Encoding, Style) { |
867 | assert(Tok.is(TT_LineComment) && |
868 | "line comment section must start with a line comment" ); |
869 | FormatToken *LineTok = nullptr; |
870 | const int Minimum = Style.SpacesInLineCommentPrefix.Minimum; |
871 | // How many spaces we changed in the first line of the section, this will be |
872 | // applied in all following lines |
873 | int FirstLineSpaceChange = 0; |
874 | for (const FormatToken *CurrentTok = &Tok; |
875 | CurrentTok && CurrentTok->is(TT: TT_LineComment); |
876 | CurrentTok = CurrentTok->Next) { |
877 | LastLineTok = LineTok; |
878 | StringRef TokenText(CurrentTok->TokenText); |
879 | assert((TokenText.starts_with("//" ) || TokenText.starts_with("#" )) && |
880 | "unsupported line comment prefix, '//' and '#' are supported" ); |
881 | size_t FirstLineIndex = Lines.size(); |
882 | TokenText.split(A&: Lines, Separator: "\n" ); |
883 | Content.resize(N: Lines.size()); |
884 | ContentColumn.resize(N: Lines.size()); |
885 | PrefixSpaceChange.resize(N: Lines.size()); |
886 | Tokens.resize(N: Lines.size()); |
887 | Prefix.resize(N: Lines.size()); |
888 | OriginalPrefix.resize(N: Lines.size()); |
889 | for (size_t i = FirstLineIndex, e = Lines.size(); i < e; ++i) { |
890 | Lines[i] = Lines[i].ltrim(Chars: Blanks); |
891 | StringRef IndentPrefix = getLineCommentIndentPrefix(Comment: Lines[i], Style); |
892 | OriginalPrefix[i] = IndentPrefix; |
893 | const int SpacesInPrefix = llvm::count(Range&: IndentPrefix, Element: ' '); |
894 | |
895 | // This lambda also considers multibyte character that is not handled in |
896 | // functions like isPunctuation provided by CharInfo. |
897 | const auto = [&]() { |
898 | assert(Lines[i].size() > IndentPrefix.size()); |
899 | const char = Lines[i][IndentPrefix.size()]; |
900 | const unsigned FirstCharByteSize = |
901 | encoding::getCodePointNumBytes(FirstChar: FirstCommentChar, Encoding); |
902 | if (encoding::columnWidth( |
903 | Text: Lines[i].substr(Start: IndentPrefix.size(), N: FirstCharByteSize), |
904 | Encoding) != 1) { |
905 | return false; |
906 | } |
907 | // In C-like comments, add a space before #. For example this is useful |
908 | // to preserve the relative indentation when commenting out code with |
909 | // #includes. |
910 | // |
911 | // In languages using # as the comment leader such as proto, don't |
912 | // add a space to support patterns like: |
913 | // ######### |
914 | // # section |
915 | // ######### |
916 | if (FirstCommentChar == '#' && !TokenText.starts_with(Prefix: "#" )) |
917 | return false; |
918 | return FirstCommentChar == '\\' || isPunctuation(c: FirstCommentChar) || |
919 | isHorizontalWhitespace(c: FirstCommentChar); |
920 | }; |
921 | |
922 | // On the first line of the comment section we calculate how many spaces |
923 | // are to be added or removed, all lines after that just get only the |
924 | // change and we will not look at the maximum anymore. Additionally to the |
925 | // actual first line, we calculate that when the non space Prefix changes, |
926 | // e.g. from "///" to "//". |
927 | if (i == 0 || OriginalPrefix[i].rtrim(Chars: Blanks) != |
928 | OriginalPrefix[i - 1].rtrim(Chars: Blanks)) { |
929 | if (SpacesInPrefix < Minimum && Lines[i].size() > IndentPrefix.size() && |
930 | !NoSpaceBeforeFirstCommentChar()) { |
931 | FirstLineSpaceChange = Minimum - SpacesInPrefix; |
932 | } else if (static_cast<unsigned>(SpacesInPrefix) > |
933 | Style.SpacesInLineCommentPrefix.Maximum) { |
934 | FirstLineSpaceChange = |
935 | Style.SpacesInLineCommentPrefix.Maximum - SpacesInPrefix; |
936 | } else { |
937 | FirstLineSpaceChange = 0; |
938 | } |
939 | } |
940 | |
941 | if (Lines[i].size() != IndentPrefix.size()) { |
942 | PrefixSpaceChange[i] = FirstLineSpaceChange; |
943 | |
944 | if (SpacesInPrefix + PrefixSpaceChange[i] < Minimum) { |
945 | PrefixSpaceChange[i] += |
946 | Minimum - (SpacesInPrefix + PrefixSpaceChange[i]); |
947 | } |
948 | |
949 | assert(Lines[i].size() > IndentPrefix.size()); |
950 | const auto FirstNonSpace = Lines[i][IndentPrefix.size()]; |
951 | const bool = LineTok && switchesFormatting(Token: *LineTok); |
952 | const bool LineRequiresLeadingSpace = |
953 | !NoSpaceBeforeFirstCommentChar() || |
954 | (FirstNonSpace == '}' && FirstLineSpaceChange != 0); |
955 | const bool AllowsSpaceChange = |
956 | !IsFormatComment && |
957 | (SpacesInPrefix != 0 || LineRequiresLeadingSpace); |
958 | |
959 | if (PrefixSpaceChange[i] > 0 && AllowsSpaceChange) { |
960 | Prefix[i] = IndentPrefix.str(); |
961 | Prefix[i].append(n: PrefixSpaceChange[i], c: ' '); |
962 | } else if (PrefixSpaceChange[i] < 0 && AllowsSpaceChange) { |
963 | Prefix[i] = IndentPrefix |
964 | .drop_back(N: std::min<std::size_t>( |
965 | a: -PrefixSpaceChange[i], b: SpacesInPrefix)) |
966 | .str(); |
967 | } else { |
968 | Prefix[i] = IndentPrefix.str(); |
969 | } |
970 | } else { |
971 | // If the IndentPrefix is the whole line, there is no content and we |
972 | // drop just all space |
973 | Prefix[i] = IndentPrefix.drop_back(N: SpacesInPrefix).str(); |
974 | } |
975 | |
976 | Tokens[i] = LineTok; |
977 | Content[i] = Lines[i].substr(Start: IndentPrefix.size()); |
978 | ContentColumn[i] = |
979 | StartColumn + encoding::columnWidthWithTabs(Text: Prefix[i], StartColumn, |
980 | TabWidth: Style.TabWidth, Encoding); |
981 | |
982 | // Calculate the end of the non-whitespace text in this line. |
983 | size_t EndOfLine = Content[i].find_last_not_of(Chars: Blanks); |
984 | if (EndOfLine == StringRef::npos) |
985 | EndOfLine = Content[i].size(); |
986 | else |
987 | ++EndOfLine; |
988 | Content[i] = Content[i].substr(Start: 0, N: EndOfLine); |
989 | } |
990 | LineTok = CurrentTok->Next; |
991 | if (CurrentTok->Next && !CurrentTok->Next->ContinuesLineCommentSection) { |
992 | // A line comment section needs to broken by a line comment that is |
993 | // preceded by at least two newlines. Note that we put this break here |
994 | // instead of breaking at a previous stage during parsing, since that |
995 | // would split the contents of the enum into two unwrapped lines in this |
996 | // example, which is undesirable: |
997 | // enum A { |
998 | // a, // comment about a |
999 | // |
1000 | // // comment about b |
1001 | // b |
1002 | // }; |
1003 | // |
1004 | // FIXME: Consider putting separate line comment sections as children to |
1005 | // the unwrapped line instead. |
1006 | break; |
1007 | } |
1008 | } |
1009 | } |
1010 | |
1011 | unsigned |
1012 | BreakableLineCommentSection::(unsigned LineIndex, unsigned Offset, |
1013 | StringRef::size_type Length, |
1014 | unsigned StartColumn) const { |
1015 | return encoding::columnWidthWithTabs( |
1016 | Text: Content[LineIndex].substr(Start: Offset, N: Length), StartColumn, TabWidth: Style.TabWidth, |
1017 | Encoding); |
1018 | } |
1019 | |
1020 | unsigned |
1021 | BreakableLineCommentSection::getContentStartColumn(unsigned LineIndex, |
1022 | bool /*Break*/) const { |
1023 | return ContentColumn[LineIndex]; |
1024 | } |
1025 | |
1026 | void BreakableLineCommentSection::( |
1027 | unsigned LineIndex, unsigned TailOffset, Split Split, |
1028 | unsigned ContentIndent, WhitespaceManager &Whitespaces) const { |
1029 | StringRef Text = Content[LineIndex].substr(Start: TailOffset); |
1030 | // Compute the offset of the split relative to the beginning of the token |
1031 | // text. |
1032 | unsigned BreakOffsetInToken = |
1033 | Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first; |
1034 | unsigned CharsToRemove = Split.second; |
1035 | Whitespaces.replaceWhitespaceInToken( |
1036 | Tok: tokenAt(LineIndex), Offset: BreakOffsetInToken, ReplaceChars: CharsToRemove, PreviousPostfix: "" , |
1037 | CurrentPrefix: Prefix[LineIndex], InPPDirective, /*Newlines=*/1, |
1038 | /*Spaces=*/ContentColumn[LineIndex] - Prefix[LineIndex].size()); |
1039 | } |
1040 | |
1041 | BreakableComment::Split BreakableLineCommentSection::( |
1042 | unsigned LineIndex, const llvm::Regex &) const { |
1043 | if (!mayReflow(LineIndex, CommentPragmasRegex)) |
1044 | return Split(StringRef::npos, 0); |
1045 | |
1046 | size_t Trimmed = Content[LineIndex].find_first_not_of(Chars: Blanks); |
1047 | |
1048 | // In a line comment section each line is a separate token; thus, after a |
1049 | // split we replace all whitespace before the current line comment token |
1050 | // (which does not need to be included in the split), plus the start of the |
1051 | // line up to where the content starts. |
1052 | return Split(0, Trimmed != StringRef::npos ? Trimmed : 0); |
1053 | } |
1054 | |
1055 | void BreakableLineCommentSection::(unsigned LineIndex, |
1056 | WhitespaceManager &Whitespaces) const { |
1057 | if (LineIndex > 0 && Tokens[LineIndex] != Tokens[LineIndex - 1]) { |
1058 | // Reflow happens between tokens. Replace the whitespace between the |
1059 | // tokens by the empty string. |
1060 | Whitespaces.replaceWhitespace( |
1061 | Tok&: *Tokens[LineIndex], /*Newlines=*/0, /*Spaces=*/0, |
1062 | /*StartOfTokenColumn=*/StartColumn, /*IsAligned=*/true, |
1063 | /*InPPDirective=*/false); |
1064 | } else if (LineIndex > 0) { |
1065 | // In case we're reflowing after the '\' in: |
1066 | // |
1067 | // // line comment \ |
1068 | // // line 2 |
1069 | // |
1070 | // the reflow happens inside the single comment token (it is a single line |
1071 | // comment with an unescaped newline). |
1072 | // Replace the whitespace between the '\' and '//' with the empty string. |
1073 | // |
1074 | // Offset points to after the '\' relative to start of the token. |
1075 | unsigned Offset = Lines[LineIndex - 1].data() + |
1076 | Lines[LineIndex - 1].size() - |
1077 | tokenAt(LineIndex: LineIndex - 1).TokenText.data(); |
1078 | // WhitespaceLength is the number of chars between the '\' and the '//' on |
1079 | // the next line. |
1080 | unsigned WhitespaceLength = |
1081 | Lines[LineIndex].data() - tokenAt(LineIndex).TokenText.data() - Offset; |
1082 | Whitespaces.replaceWhitespaceInToken(Tok: *Tokens[LineIndex], Offset, |
1083 | /*ReplaceChars=*/WhitespaceLength, |
1084 | /*PreviousPostfix=*/"" , |
1085 | /*CurrentPrefix=*/"" , |
1086 | /*InPPDirective=*/false, |
1087 | /*Newlines=*/0, |
1088 | /*Spaces=*/0); |
1089 | } |
1090 | // Replace the indent and prefix of the token with the reflow prefix. |
1091 | unsigned Offset = |
1092 | Lines[LineIndex].data() - tokenAt(LineIndex).TokenText.data(); |
1093 | unsigned WhitespaceLength = |
1094 | Content[LineIndex].data() - Lines[LineIndex].data(); |
1095 | Whitespaces.replaceWhitespaceInToken(Tok: *Tokens[LineIndex], Offset, |
1096 | /*ReplaceChars=*/WhitespaceLength, |
1097 | /*PreviousPostfix=*/"" , |
1098 | /*CurrentPrefix=*/ReflowPrefix, |
1099 | /*InPPDirective=*/false, |
1100 | /*Newlines=*/0, |
1101 | /*Spaces=*/0); |
1102 | } |
1103 | |
1104 | void BreakableLineCommentSection::( |
1105 | unsigned LineIndex, WhitespaceManager &Whitespaces) const { |
1106 | // If this is the first line of a token, we need to inform Whitespace Manager |
1107 | // about it: either adapt the whitespace range preceding it, or mark it as an |
1108 | // untouchable token. |
1109 | // This happens for instance here: |
1110 | // // line 1 \ |
1111 | // // line 2 |
1112 | if (LineIndex > 0 && Tokens[LineIndex] != Tokens[LineIndex - 1]) { |
1113 | // This is the first line for the current token, but no reflow with the |
1114 | // previous token is necessary. However, we still may need to adjust the |
1115 | // start column. Note that ContentColumn[LineIndex] is the expected |
1116 | // content column after a possible update to the prefix, hence the prefix |
1117 | // length change is included. |
1118 | unsigned LineColumn = |
1119 | ContentColumn[LineIndex] - |
1120 | (Content[LineIndex].data() - Lines[LineIndex].data()) + |
1121 | (OriginalPrefix[LineIndex].size() - Prefix[LineIndex].size()); |
1122 | |
1123 | // We always want to create a replacement instead of adding an untouchable |
1124 | // token, even if LineColumn is the same as the original column of the |
1125 | // token. This is because WhitespaceManager doesn't align trailing |
1126 | // comments if they are untouchable. |
1127 | Whitespaces.replaceWhitespace(Tok&: *Tokens[LineIndex], |
1128 | /*Newlines=*/1, |
1129 | /*Spaces=*/LineColumn, |
1130 | /*StartOfTokenColumn=*/LineColumn, |
1131 | /*IsAligned=*/true, |
1132 | /*InPPDirective=*/false); |
1133 | } |
1134 | if (OriginalPrefix[LineIndex] != Prefix[LineIndex]) { |
1135 | // Adjust the prefix if necessary. |
1136 | const auto SpacesToRemove = -std::min(a: PrefixSpaceChange[LineIndex], b: 0); |
1137 | const auto SpacesToAdd = std::max(a: PrefixSpaceChange[LineIndex], b: 0); |
1138 | Whitespaces.replaceWhitespaceInToken( |
1139 | Tok: tokenAt(LineIndex), Offset: OriginalPrefix[LineIndex].size() - SpacesToRemove, |
1140 | /*ReplaceChars=*/SpacesToRemove, PreviousPostfix: "" , CurrentPrefix: "" , /*InPPDirective=*/false, |
1141 | /*Newlines=*/0, /*Spaces=*/SpacesToAdd); |
1142 | } |
1143 | } |
1144 | |
1145 | void BreakableLineCommentSection::(LineState &State) const { |
1146 | if (LastLineTok) |
1147 | State.NextToken = LastLineTok->Next; |
1148 | } |
1149 | |
1150 | bool BreakableLineCommentSection::( |
1151 | unsigned LineIndex, const llvm::Regex &) const { |
1152 | // Line comments have the indent as part of the prefix, so we need to |
1153 | // recompute the start of the line. |
1154 | StringRef IndentContent = Content[LineIndex]; |
1155 | if (Lines[LineIndex].starts_with(Prefix: "//" )) |
1156 | IndentContent = Lines[LineIndex].substr(Start: 2); |
1157 | // FIXME: Decide whether we want to reflow non-regular indents: |
1158 | // Currently, we only reflow when the OriginalPrefix[LineIndex] matches the |
1159 | // OriginalPrefix[LineIndex-1]. That means we don't reflow |
1160 | // // text that protrudes |
1161 | // // into text with different indent |
1162 | // We do reflow in that case in block comments. |
1163 | return LineIndex > 0 && !CommentPragmasRegex.match(String: IndentContent) && |
1164 | mayReflowContent(Content: Content[LineIndex]) && !Tok.Finalized && |
1165 | !switchesFormatting(Token: tokenAt(LineIndex)) && |
1166 | OriginalPrefix[LineIndex] == OriginalPrefix[LineIndex - 1]; |
1167 | } |
1168 | |
1169 | } // namespace format |
1170 | } // namespace clang |
1171 | |