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