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