|
150
|
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 bool DecorationEndsWithStar = false) {
|
|
|
70 LLVM_DEBUG(llvm::dbgs() << "Comment split: \"" << Text
|
|
|
71 << "\", Column limit: " << ColumnLimit
|
|
|
72 << ", Content start: " << ContentStartColumn << "\n");
|
|
|
73 if (ColumnLimit <= ContentStartColumn + 1)
|
|
|
74 return BreakableToken::Split(StringRef::npos, 0);
|
|
|
75
|
|
|
76 unsigned MaxSplit = ColumnLimit - ContentStartColumn + 1;
|
|
|
77 unsigned MaxSplitBytes = 0;
|
|
|
78
|
|
|
79 for (unsigned NumChars = 0;
|
|
|
80 NumChars < MaxSplit && MaxSplitBytes < Text.size();) {
|
|
|
81 unsigned BytesInChar =
|
|
|
82 encoding::getCodePointNumBytes(Text[MaxSplitBytes], Encoding);
|
|
|
83 NumChars +=
|
|
|
84 encoding::columnWidthWithTabs(Text.substr(MaxSplitBytes, BytesInChar),
|
|
|
85 ContentStartColumn, TabWidth, Encoding);
|
|
|
86 MaxSplitBytes += BytesInChar;
|
|
|
87 }
|
|
|
88
|
|
|
89 StringRef::size_type SpaceOffset = Text.find_last_of(Blanks, MaxSplitBytes);
|
|
|
90
|
|
|
91 static const auto kNumberedListRegexp = llvm::Regex("^[1-9][0-9]?\\.");
|
|
|
92 while (SpaceOffset != StringRef::npos) {
|
|
|
93 // Do not split before a number followed by a dot: this would be interpreted
|
|
|
94 // as a numbered list, which would prevent re-flowing in subsequent passes.
|
|
|
95 if (kNumberedListRegexp.match(Text.substr(SpaceOffset).ltrim(Blanks)))
|
|
|
96 SpaceOffset = Text.find_last_of(Blanks, SpaceOffset);
|
|
|
97 // In JavaScript, some @tags can be followed by {, and machinery that parses
|
|
|
98 // these comments will fail to understand the comment if followed by a line
|
|
|
99 // break. So avoid ever breaking before a {.
|
|
|
100 else if (Style.Language == FormatStyle::LK_JavaScript &&
|
|
|
101 SpaceOffset + 1 < Text.size() && Text[SpaceOffset + 1] == '{')
|
|
|
102 SpaceOffset = Text.find_last_of(Blanks, SpaceOffset);
|
|
|
103 else
|
|
|
104 break;
|
|
|
105 }
|
|
|
106
|
|
|
107 if (SpaceOffset == StringRef::npos ||
|
|
|
108 // Don't break at leading whitespace.
|
|
|
109 Text.find_last_not_of(Blanks, SpaceOffset) == StringRef::npos) {
|
|
|
110 // Make sure that we don't break at leading whitespace that
|
|
|
111 // reaches past MaxSplit.
|
|
|
112 StringRef::size_type FirstNonWhitespace = Text.find_first_not_of(Blanks);
|
|
|
113 if (FirstNonWhitespace == StringRef::npos)
|
|
|
114 // If the comment is only whitespace, we cannot split.
|
|
|
115 return BreakableToken::Split(StringRef::npos, 0);
|
|
|
116 SpaceOffset = Text.find_first_of(
|
|
|
117 Blanks, std::max<unsigned>(MaxSplitBytes, FirstNonWhitespace));
|
|
|
118 }
|
|
|
119 if (SpaceOffset != StringRef::npos && SpaceOffset != 0) {
|
|
|
120 // adaptStartOfLine will break after lines starting with /** if the comment
|
|
|
121 // is broken anywhere. Avoid emitting this break twice here.
|
|
|
122 // Example: in /** longtextcomesherethatbreaks */ (with ColumnLimit 20) will
|
|
|
123 // insert a break after /**, so this code must not insert the same break.
|
|
|
124 if (SpaceOffset == 1 && Text[SpaceOffset - 1] == '*')
|
|
|
125 return BreakableToken::Split(StringRef::npos, 0);
|
|
|
126 StringRef BeforeCut = Text.substr(0, SpaceOffset).rtrim(Blanks);
|
|
|
127 StringRef AfterCut = Text.substr(SpaceOffset);
|
|
|
128 // Don't trim the leading blanks if it would create a */ after the break.
|
|
|
129 if (!DecorationEndsWithStar || AfterCut.size() <= 1 || AfterCut[1] != '/')
|
|
|
130 AfterCut = AfterCut.ltrim(Blanks);
|
|
|
131 return BreakableToken::Split(BeforeCut.size(),
|
|
|
132 AfterCut.begin() - BeforeCut.end());
|
|
|
133 }
|
|
|
134 return BreakableToken::Split(StringRef::npos, 0);
|
|
|
135 }
|
|
|
136
|
|
|
137 static BreakableToken::Split
|
|
|
138 getStringSplit(StringRef Text, unsigned UsedColumns, unsigned ColumnLimit,
|
|
|
139 unsigned TabWidth, encoding::Encoding Encoding) {
|
|
|
140 // FIXME: Reduce unit test case.
|
|
|
141 if (Text.empty())
|
|
|
142 return BreakableToken::Split(StringRef::npos, 0);
|
|
|
143 if (ColumnLimit <= UsedColumns)
|
|
|
144 return BreakableToken::Split(StringRef::npos, 0);
|
|
|
145 unsigned MaxSplit = ColumnLimit - UsedColumns;
|
|
|
146 StringRef::size_type SpaceOffset = 0;
|
|
|
147 StringRef::size_type SlashOffset = 0;
|
|
|
148 StringRef::size_type WordStartOffset = 0;
|
|
|
149 StringRef::size_type SplitPoint = 0;
|
|
|
150 for (unsigned Chars = 0;;) {
|
|
|
151 unsigned Advance;
|
|
|
152 if (Text[0] == '\\') {
|
|
|
153 Advance = encoding::getEscapeSequenceLength(Text);
|
|
|
154 Chars += Advance;
|
|
|
155 } else {
|
|
|
156 Advance = encoding::getCodePointNumBytes(Text[0], Encoding);
|
|
|
157 Chars += encoding::columnWidthWithTabs(
|
|
|
158 Text.substr(0, Advance), UsedColumns + Chars, TabWidth, Encoding);
|
|
|
159 }
|
|
|
160
|
|
|
161 if (Chars > MaxSplit || Text.size() <= Advance)
|
|
|
162 break;
|
|
|
163
|
|
|
164 if (IsBlank(Text[0]))
|
|
|
165 SpaceOffset = SplitPoint;
|
|
|
166 if (Text[0] == '/')
|
|
|
167 SlashOffset = SplitPoint;
|
|
|
168 if (Advance == 1 && !isAlphanumeric(Text[0]))
|
|
|
169 WordStartOffset = SplitPoint;
|
|
|
170
|
|
|
171 SplitPoint += Advance;
|
|
|
172 Text = Text.substr(Advance);
|
|
|
173 }
|
|
|
174
|
|
|
175 if (SpaceOffset != 0)
|
|
|
176 return BreakableToken::Split(SpaceOffset + 1, 0);
|
|
|
177 if (SlashOffset != 0)
|
|
|
178 return BreakableToken::Split(SlashOffset + 1, 0);
|
|
|
179 if (WordStartOffset != 0)
|
|
|
180 return BreakableToken::Split(WordStartOffset + 1, 0);
|
|
|
181 if (SplitPoint != 0)
|
|
|
182 return BreakableToken::Split(SplitPoint, 0);
|
|
|
183 return BreakableToken::Split(StringRef::npos, 0);
|
|
|
184 }
|
|
|
185
|
|
|
186 bool switchesFormatting(const FormatToken &Token) {
|
|
|
187 assert((Token.is(TT_BlockComment) || Token.is(TT_LineComment)) &&
|
|
|
188 "formatting regions are switched by comment tokens");
|
|
|
189 StringRef Content = Token.TokenText.substr(2).ltrim();
|
|
|
190 return Content.startswith("clang-format on") ||
|
|
|
191 Content.startswith("clang-format off");
|
|
|
192 }
|
|
|
193
|
|
|
194 unsigned
|
|
|
195 BreakableToken::getLengthAfterCompression(unsigned RemainingTokenColumns,
|
|
|
196 Split Split) const {
|
|
|
197 // Example: consider the content
|
|
|
198 // lala lala
|
|
|
199 // - RemainingTokenColumns is the original number of columns, 10;
|
|
|
200 // - Split is (4, 2), denoting the two spaces between the two words;
|
|
|
201 //
|
|
|
202 // We compute the number of columns when the split is compressed into a single
|
|
|
203 // space, like:
|
|
|
204 // lala lala
|
|
|
205 //
|
|
|
206 // FIXME: Correctly measure the length of whitespace in Split.second so it
|
|
|
207 // works with tabs.
|
|
|
208 return RemainingTokenColumns + 1 - Split.second;
|
|
|
209 }
|
|
|
210
|
|
|
211 unsigned BreakableStringLiteral::getLineCount() const { return 1; }
|
|
|
212
|
|
|
213 unsigned BreakableStringLiteral::getRangeLength(unsigned LineIndex,
|
|
|
214 unsigned Offset,
|
|
|
215 StringRef::size_type Length,
|
|
|
216 unsigned StartColumn) const {
|
|
|
217 llvm_unreachable("Getting the length of a part of the string literal "
|
|
|
218 "indicates that the code tries to reflow it.");
|
|
|
219 }
|
|
|
220
|
|
|
221 unsigned
|
|
|
222 BreakableStringLiteral::getRemainingLength(unsigned LineIndex, unsigned Offset,
|
|
|
223 unsigned StartColumn) const {
|
|
|
224 return UnbreakableTailLength + Postfix.size() +
|
|
|
225 encoding::columnWidthWithTabs(Line.substr(Offset, StringRef::npos),
|
|
|
226 StartColumn, Style.TabWidth, Encoding);
|
|
|
227 }
|
|
|
228
|
|
|
229 unsigned BreakableStringLiteral::getContentStartColumn(unsigned LineIndex,
|
|
|
230 bool Break) const {
|
|
|
231 return StartColumn + Prefix.size();
|
|
|
232 }
|
|
|
233
|
|
|
234 BreakableStringLiteral::BreakableStringLiteral(
|
|
|
235 const FormatToken &Tok, unsigned StartColumn, StringRef Prefix,
|
|
|
236 StringRef Postfix, unsigned UnbreakableTailLength, bool InPPDirective,
|
|
|
237 encoding::Encoding Encoding, const FormatStyle &Style)
|
|
|
238 : BreakableToken(Tok, InPPDirective, Encoding, Style),
|
|
|
239 StartColumn(StartColumn), Prefix(Prefix), Postfix(Postfix),
|
|
|
240 UnbreakableTailLength(UnbreakableTailLength) {
|
|
|
241 assert(Tok.TokenText.startswith(Prefix) && Tok.TokenText.endswith(Postfix));
|
|
|
242 Line = Tok.TokenText.substr(
|
|
|
243 Prefix.size(), Tok.TokenText.size() - Prefix.size() - Postfix.size());
|
|
|
244 }
|
|
|
245
|
|
|
246 BreakableToken::Split BreakableStringLiteral::getSplit(
|
|
|
247 unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit,
|
|
|
248 unsigned ContentStartColumn, const llvm::Regex &CommentPragmasRegex) const {
|
|
|
249 return getStringSplit(Line.substr(TailOffset), ContentStartColumn,
|
|
|
250 ColumnLimit - Postfix.size(), Style.TabWidth, Encoding);
|
|
|
251 }
|
|
|
252
|
|
|
253 void BreakableStringLiteral::insertBreak(unsigned LineIndex,
|
|
|
254 unsigned TailOffset, Split Split,
|
|
|
255 unsigned ContentIndent,
|
|
|
256 WhitespaceManager &Whitespaces) const {
|
|
|
257 Whitespaces.replaceWhitespaceInToken(
|
|
|
258 Tok, Prefix.size() + TailOffset + Split.first, Split.second, Postfix,
|
|
|
259 Prefix, InPPDirective, 1, StartColumn);
|
|
|
260 }
|
|
|
261
|
|
|
262 BreakableComment::BreakableComment(const FormatToken &Token,
|
|
|
263 unsigned StartColumn, bool InPPDirective,
|
|
|
264 encoding::Encoding Encoding,
|
|
|
265 const FormatStyle &Style)
|
|
|
266 : BreakableToken(Token, InPPDirective, Encoding, Style),
|
|
|
267 StartColumn(StartColumn) {}
|
|
|
268
|
|
|
269 unsigned BreakableComment::getLineCount() const { return Lines.size(); }
|
|
|
270
|
|
|
271 BreakableToken::Split
|
|
|
272 BreakableComment::getSplit(unsigned LineIndex, unsigned TailOffset,
|
|
|
273 unsigned ColumnLimit, unsigned ContentStartColumn,
|
|
|
274 const llvm::Regex &CommentPragmasRegex) const {
|
|
|
275 // Don't break lines matching the comment pragmas regex.
|
|
|
276 if (CommentPragmasRegex.match(Content[LineIndex]))
|
|
|
277 return Split(StringRef::npos, 0);
|
|
|
278 return getCommentSplit(Content[LineIndex].substr(TailOffset),
|
|
|
279 ContentStartColumn, ColumnLimit, Style.TabWidth,
|
|
|
280 Encoding, Style);
|
|
|
281 }
|
|
|
282
|
|
|
283 void BreakableComment::compressWhitespace(
|
|
|
284 unsigned LineIndex, unsigned TailOffset, Split Split,
|
|
|
285 WhitespaceManager &Whitespaces) const {
|
|
|
286 StringRef Text = Content[LineIndex].substr(TailOffset);
|
|
|
287 // Text is relative to the content line, but Whitespaces operates relative to
|
|
|
288 // the start of the corresponding token, so compute the start of the Split
|
|
|
289 // that needs to be compressed into a single space relative to the start of
|
|
|
290 // its token.
|
|
|
291 unsigned BreakOffsetInToken =
|
|
|
292 Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
|
|
|
293 unsigned CharsToRemove = Split.second;
|
|
|
294 Whitespaces.replaceWhitespaceInToken(
|
|
|
295 tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "", "",
|
|
|
296 /*InPPDirective=*/false, /*Newlines=*/0, /*Spaces=*/1);
|
|
|
297 }
|
|
|
298
|
|
|
299 const FormatToken &BreakableComment::tokenAt(unsigned LineIndex) const {
|
|
|
300 return Tokens[LineIndex] ? *Tokens[LineIndex] : Tok;
|
|
|
301 }
|
|
|
302
|
|
|
303 static bool mayReflowContent(StringRef Content) {
|
|
|
304 Content = Content.trim(Blanks);
|
|
|
305 // Lines starting with '@' commonly have special meaning.
|
|
|
306 // Lines starting with '-', '-#', '+' or '*' are bulleted/numbered lists.
|
|
|
307 bool hasSpecialMeaningPrefix = false;
|
|
|
308 for (StringRef Prefix :
|
|
|
309 {"@", "TODO", "FIXME", "XXX", "-# ", "- ", "+ ", "* "}) {
|
|
|
310 if (Content.startswith(Prefix)) {
|
|
|
311 hasSpecialMeaningPrefix = true;
|
|
|
312 break;
|
|
|
313 }
|
|
|
314 }
|
|
|
315
|
|
|
316 // Numbered lists may also start with a number followed by '.'
|
|
|
317 // To avoid issues if a line starts with a number which is actually the end
|
|
|
318 // of a previous line, we only consider numbers with up to 2 digits.
|
|
|
319 static const auto kNumberedListRegexp = llvm::Regex("^[1-9][0-9]?\\. ");
|
|
|
320 hasSpecialMeaningPrefix =
|
|
|
321 hasSpecialMeaningPrefix || kNumberedListRegexp.match(Content);
|
|
|
322
|
|
|
323 // Simple heuristic for what to reflow: content should contain at least two
|
|
|
324 // characters and either the first or second character must be
|
|
|
325 // non-punctuation.
|
|
|
326 return Content.size() >= 2 && !hasSpecialMeaningPrefix &&
|
|
|
327 !Content.endswith("\\") &&
|
|
|
328 // Note that this is UTF-8 safe, since if isPunctuation(Content[0]) is
|
|
|
329 // true, then the first code point must be 1 byte long.
|
|
|
330 (!isPunctuation(Content[0]) || !isPunctuation(Content[1]));
|
|
|
331 }
|
|
|
332
|
|
|
333 BreakableBlockComment::BreakableBlockComment(
|
|
|
334 const FormatToken &Token, unsigned StartColumn,
|
|
|
335 unsigned OriginalStartColumn, bool FirstInLine, bool InPPDirective,
|
|
|
336 encoding::Encoding Encoding, const FormatStyle &Style, bool UseCRLF)
|
|
|
337 : BreakableComment(Token, StartColumn, InPPDirective, Encoding, Style),
|
|
|
338 DelimitersOnNewline(false),
|
|
|
339 UnbreakableTailLength(Token.UnbreakableTailLength) {
|
|
|
340 assert(Tok.is(TT_BlockComment) &&
|
|
|
341 "block comment section must start with a block comment");
|
|
|
342
|
|
|
343 StringRef TokenText(Tok.TokenText);
|
|
|
344 assert(TokenText.startswith("/*") && TokenText.endswith("*/"));
|
|
|
345 TokenText.substr(2, TokenText.size() - 4)
|
|
|
346 .split(Lines, UseCRLF ? "\r\n" : "\n");
|
|
|
347
|
|
|
348 int IndentDelta = StartColumn - OriginalStartColumn;
|
|
|
349 Content.resize(Lines.size());
|
|
|
350 Content[0] = Lines[0];
|
|
|
351 ContentColumn.resize(Lines.size());
|
|
|
352 // Account for the initial '/*'.
|
|
|
353 ContentColumn[0] = StartColumn + 2;
|
|
|
354 Tokens.resize(Lines.size());
|
|
|
355 for (size_t i = 1; i < Lines.size(); ++i)
|
|
|
356 adjustWhitespace(i, IndentDelta);
|
|
|
357
|
|
|
358 // Align decorations with the column of the star on the first line,
|
|
|
359 // that is one column after the start "/*".
|
|
|
360 DecorationColumn = StartColumn + 1;
|
|
|
361
|
|
|
362 // Account for comment decoration patterns like this:
|
|
|
363 //
|
|
|
364 // /*
|
|
|
365 // ** blah blah blah
|
|
|
366 // */
|
|
|
367 if (Lines.size() >= 2 && Content[1].startswith("**") &&
|
|
|
368 static_cast<unsigned>(ContentColumn[1]) == StartColumn) {
|
|
|
369 DecorationColumn = StartColumn;
|
|
|
370 }
|
|
|
371
|
|
|
372 Decoration = "* ";
|
|
|
373 if (Lines.size() == 1 && !FirstInLine) {
|
|
|
374 // Comments for which FirstInLine is false can start on arbitrary column,
|
|
|
375 // and available horizontal space can be too small to align consecutive
|
|
|
376 // lines with the first one.
|
|
|
377 // FIXME: We could, probably, align them to current indentation level, but
|
|
|
378 // now we just wrap them without stars.
|
|
|
379 Decoration = "";
|
|
|
380 }
|
|
|
381 for (size_t i = 1, e = Lines.size(); i < e && !Decoration.empty(); ++i) {
|
|
|
382 // If the last line is empty, the closing "*/" will have a star.
|
|
|
383 if (i + 1 == e && Content[i].empty())
|
|
|
384 break;
|
|
|
385 if (!Content[i].empty() && i + 1 != e && Decoration.startswith(Content[i]))
|
|
|
386 continue;
|
|
|
387 while (!Content[i].startswith(Decoration))
|
|
|
388 Decoration = Decoration.substr(0, Decoration.size() - 1);
|
|
|
389 }
|
|
|
390
|
|
|
391 LastLineNeedsDecoration = true;
|
|
|
392 IndentAtLineBreak = ContentColumn[0] + 1;
|
|
|
393 for (size_t i = 1, e = Lines.size(); i < e; ++i) {
|
|
|
394 if (Content[i].empty()) {
|
|
|
395 if (i + 1 == e) {
|
|
|
396 // Empty last line means that we already have a star as a part of the
|
|
|
397 // trailing */. We also need to preserve whitespace, so that */ is
|
|
|
398 // correctly indented.
|
|
|
399 LastLineNeedsDecoration = false;
|
|
|
400 // Align the star in the last '*/' with the stars on the previous lines.
|
|
|
401 if (e >= 2 && !Decoration.empty()) {
|
|
|
402 ContentColumn[i] = DecorationColumn;
|
|
|
403 }
|
|
|
404 } else if (Decoration.empty()) {
|
|
|
405 // For all other lines, set the start column to 0 if they're empty, so
|
|
|
406 // we do not insert trailing whitespace anywhere.
|
|
|
407 ContentColumn[i] = 0;
|
|
|
408 }
|
|
|
409 continue;
|
|
|
410 }
|
|
|
411
|
|
|
412 // The first line already excludes the star.
|
|
|
413 // The last line excludes the star if LastLineNeedsDecoration is false.
|
|
|
414 // For all other lines, adjust the line to exclude the star and
|
|
|
415 // (optionally) the first whitespace.
|
|
|
416 unsigned DecorationSize = Decoration.startswith(Content[i])
|
|
|
417 ? Content[i].size()
|
|
|
418 : Decoration.size();
|
|
|
419 if (DecorationSize) {
|
|
|
420 ContentColumn[i] = DecorationColumn + DecorationSize;
|
|
|
421 }
|
|
|
422 Content[i] = Content[i].substr(DecorationSize);
|
|
|
423 if (!Decoration.startswith(Content[i]))
|
|
|
424 IndentAtLineBreak =
|
|
|
425 std::min<int>(IndentAtLineBreak, std::max(0, ContentColumn[i]));
|
|
|
426 }
|
|
|
427 IndentAtLineBreak = std::max<unsigned>(IndentAtLineBreak, Decoration.size());
|
|
|
428
|
|
|
429 // Detect a multiline jsdoc comment and set DelimitersOnNewline in that case.
|
|
|
430 if (Style.Language == FormatStyle::LK_JavaScript ||
|
|
|
431 Style.Language == FormatStyle::LK_Java) {
|
|
|
432 if ((Lines[0] == "*" || Lines[0].startswith("* ")) && Lines.size() > 1) {
|
|
|
433 // This is a multiline jsdoc comment.
|
|
|
434 DelimitersOnNewline = true;
|
|
|
435 } else if (Lines[0].startswith("* ") && Lines.size() == 1) {
|
|
|
436 // Detect a long single-line comment, like:
|
|
|
437 // /** long long long */
|
|
|
438 // Below, '2' is the width of '*/'.
|
|
|
439 unsigned EndColumn =
|
|
|
440 ContentColumn[0] +
|
|
|
441 encoding::columnWidthWithTabs(Lines[0], ContentColumn[0],
|
|
|
442 Style.TabWidth, Encoding) +
|
|
|
443 2;
|
|
|
444 DelimitersOnNewline = EndColumn > Style.ColumnLimit;
|
|
|
445 }
|
|
|
446 }
|
|
|
447
|
|
|
448 LLVM_DEBUG({
|
|
|
449 llvm::dbgs() << "IndentAtLineBreak " << IndentAtLineBreak << "\n";
|
|
|
450 llvm::dbgs() << "DelimitersOnNewline " << DelimitersOnNewline << "\n";
|
|
|
451 for (size_t i = 0; i < Lines.size(); ++i) {
|
|
|
452 llvm::dbgs() << i << " |" << Content[i] << "| "
|
|
|
453 << "CC=" << ContentColumn[i] << "| "
|
|
|
454 << "IN=" << (Content[i].data() - Lines[i].data()) << "\n";
|
|
|
455 }
|
|
|
456 });
|
|
|
457 }
|
|
|
458
|
|
|
459 BreakableToken::Split BreakableBlockComment::getSplit(
|
|
|
460 unsigned LineIndex, unsigned TailOffset, unsigned ColumnLimit,
|
|
|
461 unsigned ContentStartColumn, const llvm::Regex &CommentPragmasRegex) const {
|
|
|
462 // Don't break lines matching the comment pragmas regex.
|
|
|
463 if (CommentPragmasRegex.match(Content[LineIndex]))
|
|
|
464 return Split(StringRef::npos, 0);
|
|
|
465 return getCommentSplit(Content[LineIndex].substr(TailOffset),
|
|
|
466 ContentStartColumn, ColumnLimit, Style.TabWidth,
|
|
|
467 Encoding, Style, Decoration.endswith("*"));
|
|
|
468 }
|
|
|
469
|
|
|
470 void BreakableBlockComment::adjustWhitespace(unsigned LineIndex,
|
|
|
471 int IndentDelta) {
|
|
|
472 // When in a preprocessor directive, the trailing backslash in a block comment
|
|
|
473 // is not needed, but can serve a purpose of uniformity with necessary escaped
|
|
|
474 // newlines outside the comment. In this case we remove it here before
|
|
|
475 // trimming the trailing whitespace. The backslash will be re-added later when
|
|
|
476 // inserting a line break.
|
|
|
477 size_t EndOfPreviousLine = Lines[LineIndex - 1].size();
|
|
|
478 if (InPPDirective && Lines[LineIndex - 1].endswith("\\"))
|
|
|
479 --EndOfPreviousLine;
|
|
|
480
|
|
|
481 // Calculate the end of the non-whitespace text in the previous line.
|
|
|
482 EndOfPreviousLine =
|
|
|
483 Lines[LineIndex - 1].find_last_not_of(Blanks, EndOfPreviousLine);
|
|
|
484 if (EndOfPreviousLine == StringRef::npos)
|
|
|
485 EndOfPreviousLine = 0;
|
|
|
486 else
|
|
|
487 ++EndOfPreviousLine;
|
|
|
488 // Calculate the start of the non-whitespace text in the current line.
|
|
|
489 size_t StartOfLine = Lines[LineIndex].find_first_not_of(Blanks);
|
|
|
490 if (StartOfLine == StringRef::npos)
|
|
|
491 StartOfLine = Lines[LineIndex].size();
|
|
|
492
|
|
|
493 StringRef Whitespace = Lines[LineIndex].substr(0, StartOfLine);
|
|
|
494 // Adjust Lines to only contain relevant text.
|
|
|
495 size_t PreviousContentOffset =
|
|
|
496 Content[LineIndex - 1].data() - Lines[LineIndex - 1].data();
|
|
|
497 Content[LineIndex - 1] = Lines[LineIndex - 1].substr(
|
|
|
498 PreviousContentOffset, EndOfPreviousLine - PreviousContentOffset);
|
|
|
499 Content[LineIndex] = Lines[LineIndex].substr(StartOfLine);
|
|
|
500
|
|
|
501 // Adjust the start column uniformly across all lines.
|
|
|
502 ContentColumn[LineIndex] =
|
|
|
503 encoding::columnWidthWithTabs(Whitespace, 0, Style.TabWidth, Encoding) +
|
|
|
504 IndentDelta;
|
|
|
505 }
|
|
|
506
|
|
|
507 unsigned BreakableBlockComment::getRangeLength(unsigned LineIndex,
|
|
|
508 unsigned Offset,
|
|
|
509 StringRef::size_type Length,
|
|
|
510 unsigned StartColumn) const {
|
|
|
511 unsigned LineLength =
|
|
|
512 encoding::columnWidthWithTabs(Content[LineIndex].substr(Offset, Length),
|
|
|
513 StartColumn, Style.TabWidth, Encoding);
|
|
|
514 // FIXME: This should go into getRemainingLength instead, but we currently
|
|
|
515 // break tests when putting it there. Investigate how to fix those tests.
|
|
|
516 // The last line gets a "*/" postfix.
|
|
|
517 if (LineIndex + 1 == Lines.size()) {
|
|
|
518 LineLength += 2;
|
|
|
519 // We never need a decoration when breaking just the trailing "*/" postfix.
|
|
|
520 // Note that checking that Length == 0 is not enough, since Length could
|
|
|
521 // also be StringRef::npos.
|
|
|
522 if (Content[LineIndex].substr(Offset, StringRef::npos).empty()) {
|
|
|
523 LineLength -= Decoration.size();
|
|
|
524 }
|
|
|
525 }
|
|
|
526 return LineLength;
|
|
|
527 }
|
|
|
528
|
|
|
529 unsigned BreakableBlockComment::getRemainingLength(unsigned LineIndex,
|
|
|
530 unsigned Offset,
|
|
|
531 unsigned StartColumn) const {
|
|
|
532 return UnbreakableTailLength +
|
|
|
533 getRangeLength(LineIndex, Offset, StringRef::npos, StartColumn);
|
|
|
534 }
|
|
|
535
|
|
|
536 unsigned BreakableBlockComment::getContentStartColumn(unsigned LineIndex,
|
|
|
537 bool Break) const {
|
|
|
538 if (Break)
|
|
|
539 return IndentAtLineBreak;
|
|
|
540 return std::max(0, ContentColumn[LineIndex]);
|
|
|
541 }
|
|
|
542
|
|
|
543 const llvm::StringSet<>
|
|
|
544 BreakableBlockComment::ContentIndentingJavadocAnnotations = {
|
|
|
545 "@param", "@return", "@returns", "@throws", "@type", "@template",
|
|
|
546 "@see", "@deprecated", "@define", "@exports", "@mods", "@private",
|
|
|
547 };
|
|
|
548
|
|
|
549 unsigned BreakableBlockComment::getContentIndent(unsigned LineIndex) const {
|
|
|
550 if (Style.Language != FormatStyle::LK_Java &&
|
|
|
551 Style.Language != FormatStyle::LK_JavaScript)
|
|
|
552 return 0;
|
|
|
553 // The content at LineIndex 0 of a comment like:
|
|
|
554 // /** line 0 */
|
|
|
555 // is "* line 0", so we need to skip over the decoration in that case.
|
|
|
556 StringRef ContentWithNoDecoration = Content[LineIndex];
|
|
|
557 if (LineIndex == 0 && ContentWithNoDecoration.startswith("*")) {
|
|
|
558 ContentWithNoDecoration = ContentWithNoDecoration.substr(1).ltrim(Blanks);
|
|
|
559 }
|
|
|
560 StringRef FirstWord = ContentWithNoDecoration.substr(
|
|
|
561 0, ContentWithNoDecoration.find_first_of(Blanks));
|
|
|
562 if (ContentIndentingJavadocAnnotations.find(FirstWord) !=
|
|
|
563 ContentIndentingJavadocAnnotations.end())
|
|
|
564 return Style.ContinuationIndentWidth;
|
|
|
565 return 0;
|
|
|
566 }
|
|
|
567
|
|
|
568 void BreakableBlockComment::insertBreak(unsigned LineIndex, unsigned TailOffset,
|
|
|
569 Split Split, unsigned ContentIndent,
|
|
|
570 WhitespaceManager &Whitespaces) const {
|
|
|
571 StringRef Text = Content[LineIndex].substr(TailOffset);
|
|
|
572 StringRef Prefix = Decoration;
|
|
|
573 // We need this to account for the case when we have a decoration "* " for all
|
|
|
574 // the lines except for the last one, where the star in "*/" acts as a
|
|
|
575 // decoration.
|
|
|
576 unsigned LocalIndentAtLineBreak = IndentAtLineBreak;
|
|
|
577 if (LineIndex + 1 == Lines.size() &&
|
|
|
578 Text.size() == Split.first + Split.second) {
|
|
|
579 // For the last line we need to break before "*/", but not to add "* ".
|
|
|
580 Prefix = "";
|
|
|
581 if (LocalIndentAtLineBreak >= 2)
|
|
|
582 LocalIndentAtLineBreak -= 2;
|
|
|
583 }
|
|
|
584 // The split offset is from the beginning of the line. Convert it to an offset
|
|
|
585 // from the beginning of the token text.
|
|
|
586 unsigned BreakOffsetInToken =
|
|
|
587 Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
|
|
|
588 unsigned CharsToRemove = Split.second;
|
|
|
589 assert(LocalIndentAtLineBreak >= Prefix.size());
|
|
|
590 std::string PrefixWithTrailingIndent = std::string(Prefix);
|
|
|
591 for (unsigned I = 0; I < ContentIndent; ++I)
|
|
|
592 PrefixWithTrailingIndent += " ";
|
|
|
593 Whitespaces.replaceWhitespaceInToken(
|
|
|
594 tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "",
|
|
|
595 PrefixWithTrailingIndent, InPPDirective, /*Newlines=*/1,
|
|
|
596 /*Spaces=*/LocalIndentAtLineBreak + ContentIndent -
|
|
|
597 PrefixWithTrailingIndent.size());
|
|
|
598 }
|
|
|
599
|
|
|
600 BreakableToken::Split BreakableBlockComment::getReflowSplit(
|
|
|
601 unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
|
|
|
602 if (!mayReflow(LineIndex, CommentPragmasRegex))
|
|
|
603 return Split(StringRef::npos, 0);
|
|
|
604
|
|
|
605 // If we're reflowing into a line with content indent, only reflow the next
|
|
|
606 // line if its starting whitespace matches the content indent.
|
|
|
607 size_t Trimmed = Content[LineIndex].find_first_not_of(Blanks);
|
|
|
608 if (LineIndex) {
|
|
|
609 unsigned PreviousContentIndent = getContentIndent(LineIndex - 1);
|
|
|
610 if (PreviousContentIndent && Trimmed != StringRef::npos &&
|
|
|
611 Trimmed != PreviousContentIndent)
|
|
|
612 return Split(StringRef::npos, 0);
|
|
|
613 }
|
|
|
614
|
|
|
615 return Split(0, Trimmed != StringRef::npos ? Trimmed : 0);
|
|
|
616 }
|
|
|
617
|
|
|
618 bool BreakableBlockComment::introducesBreakBeforeToken() const {
|
|
|
619 // A break is introduced when we want delimiters on newline.
|
|
|
620 return DelimitersOnNewline &&
|
|
|
621 Lines[0].substr(1).find_first_not_of(Blanks) != StringRef::npos;
|
|
|
622 }
|
|
|
623
|
|
|
624 void BreakableBlockComment::reflow(unsigned LineIndex,
|
|
|
625 WhitespaceManager &Whitespaces) const {
|
|
|
626 StringRef TrimmedContent = Content[LineIndex].ltrim(Blanks);
|
|
|
627 // Here we need to reflow.
|
|
|
628 assert(Tokens[LineIndex - 1] == Tokens[LineIndex] &&
|
|
|
629 "Reflowing whitespace within a token");
|
|
|
630 // This is the offset of the end of the last line relative to the start of
|
|
|
631 // the token text in the token.
|
|
|
632 unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() +
|
|
|
633 Content[LineIndex - 1].size() -
|
|
|
634 tokenAt(LineIndex).TokenText.data();
|
|
|
635 unsigned WhitespaceLength = TrimmedContent.data() -
|
|
|
636 tokenAt(LineIndex).TokenText.data() -
|
|
|
637 WhitespaceOffsetInToken;
|
|
|
638 Whitespaces.replaceWhitespaceInToken(
|
|
|
639 tokenAt(LineIndex), WhitespaceOffsetInToken,
|
|
|
640 /*ReplaceChars=*/WhitespaceLength, /*PreviousPostfix=*/"",
|
|
|
641 /*CurrentPrefix=*/ReflowPrefix, InPPDirective, /*Newlines=*/0,
|
|
|
642 /*Spaces=*/0);
|
|
|
643 }
|
|
|
644
|
|
|
645 void BreakableBlockComment::adaptStartOfLine(
|
|
|
646 unsigned LineIndex, WhitespaceManager &Whitespaces) const {
|
|
|
647 if (LineIndex == 0) {
|
|
|
648 if (DelimitersOnNewline) {
|
|
|
649 // Since we're breaking at index 1 below, the break position and the
|
|
|
650 // break length are the same.
|
|
|
651 // Note: this works because getCommentSplit is careful never to split at
|
|
|
652 // the beginning of a line.
|
|
|
653 size_t BreakLength = Lines[0].substr(1).find_first_not_of(Blanks);
|
|
|
654 if (BreakLength != StringRef::npos)
|
|
|
655 insertBreak(LineIndex, 0, Split(1, BreakLength), /*ContentIndent=*/0,
|
|
|
656 Whitespaces);
|
|
|
657 }
|
|
|
658 return;
|
|
|
659 }
|
|
|
660 // Here no reflow with the previous line will happen.
|
|
|
661 // Fix the decoration of the line at LineIndex.
|
|
|
662 StringRef Prefix = Decoration;
|
|
|
663 if (Content[LineIndex].empty()) {
|
|
|
664 if (LineIndex + 1 == Lines.size()) {
|
|
|
665 if (!LastLineNeedsDecoration) {
|
|
|
666 // If the last line was empty, we don't need a prefix, as the */ will
|
|
|
667 // line up with the decoration (if it exists).
|
|
|
668 Prefix = "";
|
|
|
669 }
|
|
|
670 } else if (!Decoration.empty()) {
|
|
|
671 // For other empty lines, if we do have a decoration, adapt it to not
|
|
|
672 // contain a trailing whitespace.
|
|
|
673 Prefix = Prefix.substr(0, 1);
|
|
|
674 }
|
|
|
675 } else {
|
|
|
676 if (ContentColumn[LineIndex] == 1) {
|
|
|
677 // This line starts immediately after the decorating *.
|
|
|
678 Prefix = Prefix.substr(0, 1);
|
|
|
679 }
|
|
|
680 }
|
|
|
681 // This is the offset of the end of the last line relative to the start of the
|
|
|
682 // token text in the token.
|
|
|
683 unsigned WhitespaceOffsetInToken = Content[LineIndex - 1].data() +
|
|
|
684 Content[LineIndex - 1].size() -
|
|
|
685 tokenAt(LineIndex).TokenText.data();
|
|
|
686 unsigned WhitespaceLength = Content[LineIndex].data() -
|
|
|
687 tokenAt(LineIndex).TokenText.data() -
|
|
|
688 WhitespaceOffsetInToken;
|
|
|
689 Whitespaces.replaceWhitespaceInToken(
|
|
|
690 tokenAt(LineIndex), WhitespaceOffsetInToken, WhitespaceLength, "", Prefix,
|
|
|
691 InPPDirective, /*Newlines=*/1, ContentColumn[LineIndex] - Prefix.size());
|
|
|
692 }
|
|
|
693
|
|
|
694 BreakableToken::Split
|
|
|
695 BreakableBlockComment::getSplitAfterLastLine(unsigned TailOffset) const {
|
|
|
696 if (DelimitersOnNewline) {
|
|
|
697 // Replace the trailing whitespace of the last line with a newline.
|
|
|
698 // In case the last line is empty, the ending '*/' is already on its own
|
|
|
699 // line.
|
|
|
700 StringRef Line = Content.back().substr(TailOffset);
|
|
|
701 StringRef TrimmedLine = Line.rtrim(Blanks);
|
|
|
702 if (!TrimmedLine.empty())
|
|
|
703 return Split(TrimmedLine.size(), Line.size() - TrimmedLine.size());
|
|
|
704 }
|
|
|
705 return Split(StringRef::npos, 0);
|
|
|
706 }
|
|
|
707
|
|
|
708 bool BreakableBlockComment::mayReflow(
|
|
|
709 unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
|
|
|
710 // Content[LineIndex] may exclude the indent after the '*' decoration. In that
|
|
|
711 // case, we compute the start of the comment pragma manually.
|
|
|
712 StringRef IndentContent = Content[LineIndex];
|
|
|
713 if (Lines[LineIndex].ltrim(Blanks).startswith("*")) {
|
|
|
714 IndentContent = Lines[LineIndex].ltrim(Blanks).substr(1);
|
|
|
715 }
|
|
|
716 return LineIndex > 0 && !CommentPragmasRegex.match(IndentContent) &&
|
|
|
717 mayReflowContent(Content[LineIndex]) && !Tok.Finalized &&
|
|
|
718 !switchesFormatting(tokenAt(LineIndex));
|
|
|
719 }
|
|
|
720
|
|
|
721 BreakableLineCommentSection::BreakableLineCommentSection(
|
|
|
722 const FormatToken &Token, unsigned StartColumn,
|
|
|
723 unsigned OriginalStartColumn, bool FirstInLine, bool InPPDirective,
|
|
|
724 encoding::Encoding Encoding, const FormatStyle &Style)
|
|
|
725 : BreakableComment(Token, StartColumn, InPPDirective, Encoding, Style) {
|
|
|
726 assert(Tok.is(TT_LineComment) &&
|
|
|
727 "line comment section must start with a line comment");
|
|
|
728 FormatToken *LineTok = nullptr;
|
|
|
729 for (const FormatToken *CurrentTok = &Tok;
|
|
|
730 CurrentTok && CurrentTok->is(TT_LineComment);
|
|
|
731 CurrentTok = CurrentTok->Next) {
|
|
|
732 LastLineTok = LineTok;
|
|
|
733 StringRef TokenText(CurrentTok->TokenText);
|
|
|
734 assert((TokenText.startswith("//") || TokenText.startswith("#")) &&
|
|
|
735 "unsupported line comment prefix, '//' and '#' are supported");
|
|
|
736 size_t FirstLineIndex = Lines.size();
|
|
|
737 TokenText.split(Lines, "\n");
|
|
|
738 Content.resize(Lines.size());
|
|
|
739 ContentColumn.resize(Lines.size());
|
|
|
740 OriginalContentColumn.resize(Lines.size());
|
|
|
741 Tokens.resize(Lines.size());
|
|
|
742 Prefix.resize(Lines.size());
|
|
|
743 OriginalPrefix.resize(Lines.size());
|
|
|
744 for (size_t i = FirstLineIndex, e = Lines.size(); i < e; ++i) {
|
|
|
745 Lines[i] = Lines[i].ltrim(Blanks);
|
|
|
746 // We need to trim the blanks in case this is not the first line in a
|
|
|
747 // multiline comment. Then the indent is included in Lines[i].
|
|
|
748 StringRef IndentPrefix =
|
|
|
749 getLineCommentIndentPrefix(Lines[i].ltrim(Blanks), Style);
|
|
|
750 assert((TokenText.startswith("//") || TokenText.startswith("#")) &&
|
|
|
751 "unsupported line comment prefix, '//' and '#' are supported");
|
|
|
752 OriginalPrefix[i] = Prefix[i] = IndentPrefix;
|
|
|
753 if (Lines[i].size() > Prefix[i].size() &&
|
|
|
754 isAlphanumeric(Lines[i][Prefix[i].size()])) {
|
|
|
755 if (Prefix[i] == "//")
|
|
|
756 Prefix[i] = "// ";
|
|
|
757 else if (Prefix[i] == "///")
|
|
|
758 Prefix[i] = "/// ";
|
|
|
759 else if (Prefix[i] == "//!")
|
|
|
760 Prefix[i] = "//! ";
|
|
|
761 else if (Prefix[i] == "///<")
|
|
|
762 Prefix[i] = "///< ";
|
|
|
763 else if (Prefix[i] == "//!<")
|
|
|
764 Prefix[i] = "//!< ";
|
|
|
765 else if (Prefix[i] == "#" &&
|
|
|
766 Style.Language == FormatStyle::LK_TextProto)
|
|
|
767 Prefix[i] = "# ";
|
|
|
768 }
|
|
|
769
|
|
|
770 Tokens[i] = LineTok;
|
|
|
771 Content[i] = Lines[i].substr(IndentPrefix.size());
|
|
|
772 OriginalContentColumn[i] =
|
|
|
773 StartColumn + encoding::columnWidthWithTabs(OriginalPrefix[i],
|
|
|
774 StartColumn,
|
|
|
775 Style.TabWidth, Encoding);
|
|
|
776 ContentColumn[i] =
|
|
|
777 StartColumn + encoding::columnWidthWithTabs(Prefix[i], StartColumn,
|
|
|
778 Style.TabWidth, Encoding);
|
|
|
779
|
|
|
780 // Calculate the end of the non-whitespace text in this line.
|
|
|
781 size_t EndOfLine = Content[i].find_last_not_of(Blanks);
|
|
|
782 if (EndOfLine == StringRef::npos)
|
|
|
783 EndOfLine = Content[i].size();
|
|
|
784 else
|
|
|
785 ++EndOfLine;
|
|
|
786 Content[i] = Content[i].substr(0, EndOfLine);
|
|
|
787 }
|
|
|
788 LineTok = CurrentTok->Next;
|
|
|
789 if (CurrentTok->Next && !CurrentTok->Next->ContinuesLineCommentSection) {
|
|
|
790 // A line comment section needs to broken by a line comment that is
|
|
|
791 // preceded by at least two newlines. Note that we put this break here
|
|
|
792 // instead of breaking at a previous stage during parsing, since that
|
|
|
793 // would split the contents of the enum into two unwrapped lines in this
|
|
|
794 // example, which is undesirable:
|
|
|
795 // enum A {
|
|
|
796 // a, // comment about a
|
|
|
797 //
|
|
|
798 // // comment about b
|
|
|
799 // b
|
|
|
800 // };
|
|
|
801 //
|
|
|
802 // FIXME: Consider putting separate line comment sections as children to
|
|
|
803 // the unwrapped line instead.
|
|
|
804 break;
|
|
|
805 }
|
|
|
806 }
|
|
|
807 }
|
|
|
808
|
|
|
809 unsigned
|
|
|
810 BreakableLineCommentSection::getRangeLength(unsigned LineIndex, unsigned Offset,
|
|
|
811 StringRef::size_type Length,
|
|
|
812 unsigned StartColumn) const {
|
|
|
813 return encoding::columnWidthWithTabs(
|
|
|
814 Content[LineIndex].substr(Offset, Length), StartColumn, Style.TabWidth,
|
|
|
815 Encoding);
|
|
|
816 }
|
|
|
817
|
|
|
818 unsigned BreakableLineCommentSection::getContentStartColumn(unsigned LineIndex,
|
|
|
819 bool Break) const {
|
|
|
820 if (Break)
|
|
|
821 return OriginalContentColumn[LineIndex];
|
|
|
822 return ContentColumn[LineIndex];
|
|
|
823 }
|
|
|
824
|
|
|
825 void BreakableLineCommentSection::insertBreak(
|
|
|
826 unsigned LineIndex, unsigned TailOffset, Split Split,
|
|
|
827 unsigned ContentIndent, WhitespaceManager &Whitespaces) const {
|
|
|
828 StringRef Text = Content[LineIndex].substr(TailOffset);
|
|
|
829 // Compute the offset of the split relative to the beginning of the token
|
|
|
830 // text.
|
|
|
831 unsigned BreakOffsetInToken =
|
|
|
832 Text.data() - tokenAt(LineIndex).TokenText.data() + Split.first;
|
|
|
833 unsigned CharsToRemove = Split.second;
|
|
|
834 // Compute the size of the new indent, including the size of the new prefix of
|
|
|
835 // the newly broken line.
|
|
|
836 unsigned IndentAtLineBreak = OriginalContentColumn[LineIndex] +
|
|
|
837 Prefix[LineIndex].size() -
|
|
|
838 OriginalPrefix[LineIndex].size();
|
|
|
839 assert(IndentAtLineBreak >= Prefix[LineIndex].size());
|
|
|
840 Whitespaces.replaceWhitespaceInToken(
|
|
|
841 tokenAt(LineIndex), BreakOffsetInToken, CharsToRemove, "",
|
|
|
842 Prefix[LineIndex], InPPDirective, /*Newlines=*/1,
|
|
|
843 /*Spaces=*/IndentAtLineBreak - Prefix[LineIndex].size());
|
|
|
844 }
|
|
|
845
|
|
|
846 BreakableComment::Split BreakableLineCommentSection::getReflowSplit(
|
|
|
847 unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
|
|
|
848 if (!mayReflow(LineIndex, CommentPragmasRegex))
|
|
|
849 return Split(StringRef::npos, 0);
|
|
|
850
|
|
|
851 size_t Trimmed = Content[LineIndex].find_first_not_of(Blanks);
|
|
|
852
|
|
|
853 // In a line comment section each line is a separate token; thus, after a
|
|
|
854 // split we replace all whitespace before the current line comment token
|
|
|
855 // (which does not need to be included in the split), plus the start of the
|
|
|
856 // line up to where the content starts.
|
|
|
857 return Split(0, Trimmed != StringRef::npos ? Trimmed : 0);
|
|
|
858 }
|
|
|
859
|
|
|
860 void BreakableLineCommentSection::reflow(unsigned LineIndex,
|
|
|
861 WhitespaceManager &Whitespaces) const {
|
|
|
862 if (LineIndex > 0 && Tokens[LineIndex] != Tokens[LineIndex - 1]) {
|
|
|
863 // Reflow happens between tokens. Replace the whitespace between the
|
|
|
864 // tokens by the empty string.
|
|
|
865 Whitespaces.replaceWhitespace(
|
|
|
866 *Tokens[LineIndex], /*Newlines=*/0, /*Spaces=*/0,
|
|
|
867 /*StartOfTokenColumn=*/StartColumn, /*InPPDirective=*/false);
|
|
|
868 } else if (LineIndex > 0) {
|
|
|
869 // In case we're reflowing after the '\' in:
|
|
|
870 //
|
|
|
871 // // line comment \
|
|
|
872 // // line 2
|
|
|
873 //
|
|
|
874 // the reflow happens inside the single comment token (it is a single line
|
|
|
875 // comment with an unescaped newline).
|
|
|
876 // Replace the whitespace between the '\' and '//' with the empty string.
|
|
|
877 //
|
|
|
878 // Offset points to after the '\' relative to start of the token.
|
|
|
879 unsigned Offset = Lines[LineIndex - 1].data() +
|
|
|
880 Lines[LineIndex - 1].size() -
|
|
|
881 tokenAt(LineIndex - 1).TokenText.data();
|
|
|
882 // WhitespaceLength is the number of chars between the '\' and the '//' on
|
|
|
883 // the next line.
|
|
|
884 unsigned WhitespaceLength =
|
|
|
885 Lines[LineIndex].data() - tokenAt(LineIndex).TokenText.data() - Offset;
|
|
|
886 Whitespaces.replaceWhitespaceInToken(*Tokens[LineIndex], Offset,
|
|
|
887 /*ReplaceChars=*/WhitespaceLength,
|
|
|
888 /*PreviousPostfix=*/"",
|
|
|
889 /*CurrentPrefix=*/"",
|
|
|
890 /*InPPDirective=*/false,
|
|
|
891 /*Newlines=*/0,
|
|
|
892 /*Spaces=*/0);
|
|
|
893 }
|
|
|
894 // Replace the indent and prefix of the token with the reflow prefix.
|
|
|
895 unsigned Offset =
|
|
|
896 Lines[LineIndex].data() - tokenAt(LineIndex).TokenText.data();
|
|
|
897 unsigned WhitespaceLength =
|
|
|
898 Content[LineIndex].data() - Lines[LineIndex].data();
|
|
|
899 Whitespaces.replaceWhitespaceInToken(*Tokens[LineIndex], Offset,
|
|
|
900 /*ReplaceChars=*/WhitespaceLength,
|
|
|
901 /*PreviousPostfix=*/"",
|
|
|
902 /*CurrentPrefix=*/ReflowPrefix,
|
|
|
903 /*InPPDirective=*/false,
|
|
|
904 /*Newlines=*/0,
|
|
|
905 /*Spaces=*/0);
|
|
|
906 }
|
|
|
907
|
|
|
908 void BreakableLineCommentSection::adaptStartOfLine(
|
|
|
909 unsigned LineIndex, WhitespaceManager &Whitespaces) const {
|
|
|
910 // If this is the first line of a token, we need to inform Whitespace Manager
|
|
|
911 // about it: either adapt the whitespace range preceding it, or mark it as an
|
|
|
912 // untouchable token.
|
|
|
913 // This happens for instance here:
|
|
|
914 // // line 1 \
|
|
|
915 // // line 2
|
|
|
916 if (LineIndex > 0 && Tokens[LineIndex] != Tokens[LineIndex - 1]) {
|
|
|
917 // This is the first line for the current token, but no reflow with the
|
|
|
918 // previous token is necessary. However, we still may need to adjust the
|
|
|
919 // start column. Note that ContentColumn[LineIndex] is the expected
|
|
|
920 // content column after a possible update to the prefix, hence the prefix
|
|
|
921 // length change is included.
|
|
|
922 unsigned LineColumn =
|
|
|
923 ContentColumn[LineIndex] -
|
|
|
924 (Content[LineIndex].data() - Lines[LineIndex].data()) +
|
|
|
925 (OriginalPrefix[LineIndex].size() - Prefix[LineIndex].size());
|
|
|
926
|
|
|
927 // We always want to create a replacement instead of adding an untouchable
|
|
|
928 // token, even if LineColumn is the same as the original column of the
|
|
|
929 // token. This is because WhitespaceManager doesn't align trailing
|
|
|
930 // comments if they are untouchable.
|
|
|
931 Whitespaces.replaceWhitespace(*Tokens[LineIndex],
|
|
|
932 /*Newlines=*/1,
|
|
|
933 /*Spaces=*/LineColumn,
|
|
|
934 /*StartOfTokenColumn=*/LineColumn,
|
|
|
935 /*InPPDirective=*/false);
|
|
|
936 }
|
|
|
937 if (OriginalPrefix[LineIndex] != Prefix[LineIndex]) {
|
|
|
938 // Adjust the prefix if necessary.
|
|
|
939
|
|
|
940 // Take care of the space possibly introduced after a decoration.
|
|
|
941 assert(Prefix[LineIndex] == (OriginalPrefix[LineIndex] + " ").str() &&
|
|
|
942 "Expecting a line comment prefix to differ from original by at most "
|
|
|
943 "a space");
|
|
|
944 Whitespaces.replaceWhitespaceInToken(
|
|
|
945 tokenAt(LineIndex), OriginalPrefix[LineIndex].size(), 0, "", "",
|
|
|
946 /*InPPDirective=*/false, /*Newlines=*/0, /*Spaces=*/1);
|
|
|
947 }
|
|
|
948 }
|
|
|
949
|
|
|
950 void BreakableLineCommentSection::updateNextToken(LineState &State) const {
|
|
|
951 if (LastLineTok) {
|
|
|
952 State.NextToken = LastLineTok->Next;
|
|
|
953 }
|
|
|
954 }
|
|
|
955
|
|
|
956 bool BreakableLineCommentSection::mayReflow(
|
|
|
957 unsigned LineIndex, const llvm::Regex &CommentPragmasRegex) const {
|
|
|
958 // Line comments have the indent as part of the prefix, so we need to
|
|
|
959 // recompute the start of the line.
|
|
|
960 StringRef IndentContent = Content[LineIndex];
|
|
|
961 if (Lines[LineIndex].startswith("//")) {
|
|
|
962 IndentContent = Lines[LineIndex].substr(2);
|
|
|
963 }
|
|
|
964 // FIXME: Decide whether we want to reflow non-regular indents:
|
|
|
965 // Currently, we only reflow when the OriginalPrefix[LineIndex] matches the
|
|
|
966 // OriginalPrefix[LineIndex-1]. That means we don't reflow
|
|
|
967 // // text that protrudes
|
|
|
968 // // into text with different indent
|
|
|
969 // We do reflow in that case in block comments.
|
|
|
970 return LineIndex > 0 && !CommentPragmasRegex.match(IndentContent) &&
|
|
|
971 mayReflowContent(Content[LineIndex]) && !Tok.Finalized &&
|
|
|
972 !switchesFormatting(tokenAt(LineIndex)) &&
|
|
|
973 OriginalPrefix[LineIndex] == OriginalPrefix[LineIndex - 1];
|
|
|
974 }
|
|
|
975
|
|
|
976 } // namespace format
|
|
|
977 } // namespace clang
|
