1 //===--- WhitespaceManager.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 /// This file implements WhitespaceManager class.
11 ///
12 //===----------------------------------------------------------------------===//
13
14 #include "WhitespaceManager.h"
15 #include "llvm/ADT/STLExtras.h"
16 #include "llvm/ADT/SmallVector.h"
17 #include <algorithm>
18
19 namespace clang {
20 namespace format {
21
operator ()(const Change & C1,const Change & C2) const22 bool WhitespaceManager::Change::IsBeforeInFile::operator()(
23 const Change &C1, const Change &C2) const {
24 return SourceMgr.isBeforeInTranslationUnit(
25 C1.OriginalWhitespaceRange.getBegin(),
26 C2.OriginalWhitespaceRange.getBegin());
27 }
28
Change(const FormatToken & Tok,bool CreateReplacement,SourceRange OriginalWhitespaceRange,int Spaces,unsigned StartOfTokenColumn,unsigned NewlinesBefore,StringRef PreviousLinePostfix,StringRef CurrentLinePrefix,bool IsAligned,bool ContinuesPPDirective,bool IsInsideToken)29 WhitespaceManager::Change::Change(const FormatToken &Tok,
30 bool CreateReplacement,
31 SourceRange OriginalWhitespaceRange,
32 int Spaces, unsigned StartOfTokenColumn,
33 unsigned NewlinesBefore,
34 StringRef PreviousLinePostfix,
35 StringRef CurrentLinePrefix, bool IsAligned,
36 bool ContinuesPPDirective, bool IsInsideToken)
37 : Tok(&Tok), CreateReplacement(CreateReplacement),
38 OriginalWhitespaceRange(OriginalWhitespaceRange),
39 StartOfTokenColumn(StartOfTokenColumn), NewlinesBefore(NewlinesBefore),
40 PreviousLinePostfix(PreviousLinePostfix),
41 CurrentLinePrefix(CurrentLinePrefix), IsAligned(IsAligned),
42 ContinuesPPDirective(ContinuesPPDirective), Spaces(Spaces),
43 IsInsideToken(IsInsideToken), IsTrailingComment(false), TokenLength(0),
44 PreviousEndOfTokenColumn(0), EscapedNewlineColumn(0),
45 StartOfBlockComment(nullptr), IndentationOffset(0), ConditionalsLevel(0) {
46 }
47
replaceWhitespace(FormatToken & Tok,unsigned Newlines,unsigned Spaces,unsigned StartOfTokenColumn,bool IsAligned,bool InPPDirective)48 void WhitespaceManager::replaceWhitespace(FormatToken &Tok, unsigned Newlines,
49 unsigned Spaces,
50 unsigned StartOfTokenColumn,
51 bool IsAligned, bool InPPDirective) {
52 if (Tok.Finalized)
53 return;
54 Tok.setDecision((Newlines > 0) ? FD_Break : FD_Continue);
55 Changes.push_back(Change(Tok, /*CreateReplacement=*/true, Tok.WhitespaceRange,
56 Spaces, StartOfTokenColumn, Newlines, "", "",
57 IsAligned, InPPDirective && !Tok.IsFirst,
58 /*IsInsideToken=*/false));
59 }
60
addUntouchableToken(const FormatToken & Tok,bool InPPDirective)61 void WhitespaceManager::addUntouchableToken(const FormatToken &Tok,
62 bool InPPDirective) {
63 if (Tok.Finalized)
64 return;
65 Changes.push_back(Change(Tok, /*CreateReplacement=*/false,
66 Tok.WhitespaceRange, /*Spaces=*/0,
67 Tok.OriginalColumn, Tok.NewlinesBefore, "", "",
68 /*IsAligned=*/false, InPPDirective && !Tok.IsFirst,
69 /*IsInsideToken=*/false));
70 }
71
72 llvm::Error
addReplacement(const tooling::Replacement & Replacement)73 WhitespaceManager::addReplacement(const tooling::Replacement &Replacement) {
74 return Replaces.add(Replacement);
75 }
76
inputUsesCRLF(StringRef Text,bool DefaultToCRLF)77 bool WhitespaceManager::inputUsesCRLF(StringRef Text, bool DefaultToCRLF) {
78 size_t LF = Text.count('\n');
79 size_t CR = Text.count('\r') * 2;
80 return LF == CR ? DefaultToCRLF : CR > LF;
81 }
82
replaceWhitespaceInToken(const FormatToken & Tok,unsigned Offset,unsigned ReplaceChars,StringRef PreviousPostfix,StringRef CurrentPrefix,bool InPPDirective,unsigned Newlines,int Spaces)83 void WhitespaceManager::replaceWhitespaceInToken(
84 const FormatToken &Tok, unsigned Offset, unsigned ReplaceChars,
85 StringRef PreviousPostfix, StringRef CurrentPrefix, bool InPPDirective,
86 unsigned Newlines, int Spaces) {
87 if (Tok.Finalized)
88 return;
89 SourceLocation Start = Tok.getStartOfNonWhitespace().getLocWithOffset(Offset);
90 Changes.push_back(
91 Change(Tok, /*CreateReplacement=*/true,
92 SourceRange(Start, Start.getLocWithOffset(ReplaceChars)), Spaces,
93 std::max(0, Spaces), Newlines, PreviousPostfix, CurrentPrefix,
94 /*IsAligned=*/true, InPPDirective && !Tok.IsFirst,
95 /*IsInsideToken=*/true));
96 }
97
generateReplacements()98 const tooling::Replacements &WhitespaceManager::generateReplacements() {
99 if (Changes.empty())
100 return Replaces;
101
102 llvm::sort(Changes, Change::IsBeforeInFile(SourceMgr));
103 calculateLineBreakInformation();
104 alignConsecutiveMacros();
105 alignConsecutiveDeclarations();
106 alignConsecutiveBitFields();
107 alignConsecutiveAssignments();
108 alignChainedConditionals();
109 alignTrailingComments();
110 alignEscapedNewlines();
111 alignArrayInitializers();
112 generateChanges();
113
114 return Replaces;
115 }
116
calculateLineBreakInformation()117 void WhitespaceManager::calculateLineBreakInformation() {
118 Changes[0].PreviousEndOfTokenColumn = 0;
119 Change *LastOutsideTokenChange = &Changes[0];
120 for (unsigned i = 1, e = Changes.size(); i != e; ++i) {
121 SourceLocation OriginalWhitespaceStart =
122 Changes[i].OriginalWhitespaceRange.getBegin();
123 SourceLocation PreviousOriginalWhitespaceEnd =
124 Changes[i - 1].OriginalWhitespaceRange.getEnd();
125 unsigned OriginalWhitespaceStartOffset =
126 SourceMgr.getFileOffset(OriginalWhitespaceStart);
127 unsigned PreviousOriginalWhitespaceEndOffset =
128 SourceMgr.getFileOffset(PreviousOriginalWhitespaceEnd);
129 assert(PreviousOriginalWhitespaceEndOffset <=
130 OriginalWhitespaceStartOffset);
131 const char *const PreviousOriginalWhitespaceEndData =
132 SourceMgr.getCharacterData(PreviousOriginalWhitespaceEnd);
133 StringRef Text(PreviousOriginalWhitespaceEndData,
134 SourceMgr.getCharacterData(OriginalWhitespaceStart) -
135 PreviousOriginalWhitespaceEndData);
136 // Usually consecutive changes would occur in consecutive tokens. This is
137 // not the case however when analyzing some preprocessor runs of the
138 // annotated lines. For example, in this code:
139 //
140 // #if A // line 1
141 // int i = 1;
142 // #else B // line 2
143 // int i = 2;
144 // #endif // line 3
145 //
146 // one of the runs will produce the sequence of lines marked with line 1, 2
147 // and 3. So the two consecutive whitespace changes just before '// line 2'
148 // and before '#endif // line 3' span multiple lines and tokens:
149 //
150 // #else B{change X}[// line 2
151 // int i = 2;
152 // ]{change Y}#endif // line 3
153 //
154 // For this reason, if the text between consecutive changes spans multiple
155 // newlines, the token length must be adjusted to the end of the original
156 // line of the token.
157 auto NewlinePos = Text.find_first_of('\n');
158 if (NewlinePos == StringRef::npos) {
159 Changes[i - 1].TokenLength = OriginalWhitespaceStartOffset -
160 PreviousOriginalWhitespaceEndOffset +
161 Changes[i].PreviousLinePostfix.size() +
162 Changes[i - 1].CurrentLinePrefix.size();
163 } else {
164 Changes[i - 1].TokenLength =
165 NewlinePos + Changes[i - 1].CurrentLinePrefix.size();
166 }
167
168 // If there are multiple changes in this token, sum up all the changes until
169 // the end of the line.
170 if (Changes[i - 1].IsInsideToken && Changes[i - 1].NewlinesBefore == 0) {
171 LastOutsideTokenChange->TokenLength +=
172 Changes[i - 1].TokenLength + Changes[i - 1].Spaces;
173 } else {
174 LastOutsideTokenChange = &Changes[i - 1];
175 }
176
177 Changes[i].PreviousEndOfTokenColumn =
178 Changes[i - 1].StartOfTokenColumn + Changes[i - 1].TokenLength;
179
180 Changes[i - 1].IsTrailingComment =
181 (Changes[i].NewlinesBefore > 0 || Changes[i].Tok->is(tok::eof) ||
182 (Changes[i].IsInsideToken && Changes[i].Tok->is(tok::comment))) &&
183 Changes[i - 1].Tok->is(tok::comment) &&
184 // FIXME: This is a dirty hack. The problem is that
185 // BreakableLineCommentSection does comment reflow changes and here is
186 // the aligning of trailing comments. Consider the case where we reflow
187 // the second line up in this example:
188 //
189 // // line 1
190 // // line 2
191 //
192 // That amounts to 2 changes by BreakableLineCommentSection:
193 // - the first, delimited by (), for the whitespace between the tokens,
194 // - and second, delimited by [], for the whitespace at the beginning
195 // of the second token:
196 //
197 // // line 1(
198 // )[// ]line 2
199 //
200 // So in the end we have two changes like this:
201 //
202 // // line1()[ ]line 2
203 //
204 // Note that the OriginalWhitespaceStart of the second change is the
205 // same as the PreviousOriginalWhitespaceEnd of the first change.
206 // In this case, the below check ensures that the second change doesn't
207 // get treated as a trailing comment change here, since this might
208 // trigger additional whitespace to be wrongly inserted before "line 2"
209 // by the comment aligner here.
210 //
211 // For a proper solution we need a mechanism to say to WhitespaceManager
212 // that a particular change breaks the current sequence of trailing
213 // comments.
214 OriginalWhitespaceStart != PreviousOriginalWhitespaceEnd;
215 }
216 // FIXME: The last token is currently not always an eof token; in those
217 // cases, setting TokenLength of the last token to 0 is wrong.
218 Changes.back().TokenLength = 0;
219 Changes.back().IsTrailingComment = Changes.back().Tok->is(tok::comment);
220
221 const WhitespaceManager::Change *LastBlockComment = nullptr;
222 for (auto &Change : Changes) {
223 // Reset the IsTrailingComment flag for changes inside of trailing comments
224 // so they don't get realigned later. Comment line breaks however still need
225 // to be aligned.
226 if (Change.IsInsideToken && Change.NewlinesBefore == 0)
227 Change.IsTrailingComment = false;
228 Change.StartOfBlockComment = nullptr;
229 Change.IndentationOffset = 0;
230 if (Change.Tok->is(tok::comment)) {
231 if (Change.Tok->is(TT_LineComment) || !Change.IsInsideToken) {
232 LastBlockComment = &Change;
233 } else if ((Change.StartOfBlockComment = LastBlockComment)) {
234 Change.IndentationOffset =
235 Change.StartOfTokenColumn -
236 Change.StartOfBlockComment->StartOfTokenColumn;
237 }
238 } else {
239 LastBlockComment = nullptr;
240 }
241 }
242
243 // Compute conditional nesting level
244 // Level is increased for each conditional, unless this conditional continues
245 // a chain of conditional, i.e. starts immediately after the colon of another
246 // conditional.
247 SmallVector<bool, 16> ScopeStack;
248 int ConditionalsLevel = 0;
249 for (auto &Change : Changes) {
250 for (unsigned i = 0, e = Change.Tok->FakeLParens.size(); i != e; ++i) {
251 bool isNestedConditional =
252 Change.Tok->FakeLParens[e - 1 - i] == prec::Conditional &&
253 !(i == 0 && Change.Tok->Previous &&
254 Change.Tok->Previous->is(TT_ConditionalExpr) &&
255 Change.Tok->Previous->is(tok::colon));
256 if (isNestedConditional)
257 ++ConditionalsLevel;
258 ScopeStack.push_back(isNestedConditional);
259 }
260
261 Change.ConditionalsLevel = ConditionalsLevel;
262
263 for (unsigned i = Change.Tok->FakeRParens; i > 0 && ScopeStack.size(); --i)
264 if (ScopeStack.pop_back_val())
265 --ConditionalsLevel;
266 }
267 }
268
269 // Align a single sequence of tokens, see AlignTokens below.
270 // Column - The token for which Matches returns true is moved to this column.
271 // RightJustify - Whether it is the token's right end or left end that gets
272 // moved to that column.
273 template <typename F>
274 static void
AlignTokenSequence(const FormatStyle & Style,unsigned Start,unsigned End,unsigned Column,bool RightJustify,F && Matches,SmallVector<WhitespaceManager::Change,16> & Changes)275 AlignTokenSequence(const FormatStyle &Style, unsigned Start, unsigned End,
276 unsigned Column, bool RightJustify, F &&Matches,
277 SmallVector<WhitespaceManager::Change, 16> &Changes) {
278 bool FoundMatchOnLine = false;
279 int Shift = 0;
280
281 // ScopeStack keeps track of the current scope depth. It contains indices of
282 // the first token on each scope.
283 // We only run the "Matches" function on tokens from the outer-most scope.
284 // However, we do need to pay special attention to one class of tokens
285 // that are not in the outer-most scope, and that is function parameters
286 // which are split across multiple lines, as illustrated by this example:
287 // double a(int x);
288 // int b(int y,
289 // double z);
290 // In the above example, we need to take special care to ensure that
291 // 'double z' is indented along with it's owning function 'b'.
292 // The same holds for calling a function:
293 // double a = foo(x);
294 // int b = bar(foo(y),
295 // foor(z));
296 // Similar for broken string literals:
297 // double x = 3.14;
298 // auto s = "Hello"
299 // "World";
300 // Special handling is required for 'nested' ternary operators.
301 SmallVector<unsigned, 16> ScopeStack;
302
303 for (unsigned i = Start; i != End; ++i) {
304 if (ScopeStack.size() != 0 &&
305 Changes[i].indentAndNestingLevel() <
306 Changes[ScopeStack.back()].indentAndNestingLevel()) {
307 ScopeStack.pop_back();
308 }
309
310 // Compare current token to previous non-comment token to ensure whether
311 // it is in a deeper scope or not.
312 unsigned PreviousNonComment = i - 1;
313 while (PreviousNonComment > Start &&
314 Changes[PreviousNonComment].Tok->is(tok::comment)) {
315 --PreviousNonComment;
316 }
317 if (i != Start && Changes[i].indentAndNestingLevel() >
318 Changes[PreviousNonComment].indentAndNestingLevel()) {
319 ScopeStack.push_back(i);
320 }
321
322 bool InsideNestedScope = ScopeStack.size() != 0;
323 bool ContinuedStringLiteral = i > Start &&
324 Changes[i].Tok->is(tok::string_literal) &&
325 Changes[i - 1].Tok->is(tok::string_literal);
326 bool SkipMatchCheck = InsideNestedScope || ContinuedStringLiteral;
327
328 if (Changes[i].NewlinesBefore > 0 && !SkipMatchCheck) {
329 Shift = 0;
330 FoundMatchOnLine = false;
331 }
332
333 // If this is the first matching token to be aligned, remember by how many
334 // spaces it has to be shifted, so the rest of the changes on the line are
335 // shifted by the same amount
336 if (!FoundMatchOnLine && !SkipMatchCheck && Matches(Changes[i])) {
337 FoundMatchOnLine = true;
338 Shift = Column - (RightJustify ? Changes[i].TokenLength : 0) -
339 Changes[i].StartOfTokenColumn;
340 Changes[i].Spaces += Shift;
341 // FIXME: This is a workaround that should be removed when we fix
342 // http://llvm.org/PR53699. An assertion later below verifies this.
343 if (Changes[i].NewlinesBefore == 0) {
344 Changes[i].Spaces =
345 std::max(Changes[i].Spaces,
346 static_cast<int>(Changes[i].Tok->SpacesRequiredBefore));
347 }
348 }
349
350 // This is for function parameters that are split across multiple lines,
351 // as mentioned in the ScopeStack comment.
352 if (InsideNestedScope && Changes[i].NewlinesBefore > 0) {
353 unsigned ScopeStart = ScopeStack.back();
354 auto ShouldShiftBeAdded = [&] {
355 // Function declaration
356 if (Changes[ScopeStart - 1].Tok->is(TT_FunctionDeclarationName))
357 return true;
358
359 // Lambda.
360 if (Changes[ScopeStart - 1].Tok->is(TT_LambdaLBrace))
361 return false;
362
363 // Continued function declaration
364 if (ScopeStart > Start + 1 &&
365 Changes[ScopeStart - 2].Tok->is(TT_FunctionDeclarationName)) {
366 return true;
367 }
368
369 // Continued function call
370 if (ScopeStart > Start + 1 &&
371 Changes[ScopeStart - 2].Tok->is(tok::identifier) &&
372 Changes[ScopeStart - 1].Tok->is(tok::l_paren) &&
373 Changes[ScopeStart].Tok->isNot(TT_LambdaLSquare)) {
374 if (Changes[i].Tok->MatchingParen &&
375 Changes[i].Tok->MatchingParen->is(TT_LambdaLBrace)) {
376 return false;
377 }
378 if (Changes[ScopeStart].NewlinesBefore > 0)
379 return false;
380 if (Changes[i].Tok->is(tok::l_brace) &&
381 Changes[i].Tok->is(BK_BracedInit)) {
382 return true;
383 }
384 return Style.BinPackArguments;
385 }
386
387 // Ternary operator
388 if (Changes[i].Tok->is(TT_ConditionalExpr))
389 return true;
390
391 // Period Initializer .XXX = 1.
392 if (Changes[i].Tok->is(TT_DesignatedInitializerPeriod))
393 return true;
394
395 // Continued ternary operator
396 if (Changes[i].Tok->Previous &&
397 Changes[i].Tok->Previous->is(TT_ConditionalExpr)) {
398 return true;
399 }
400
401 // Continued direct-list-initialization using braced list.
402 if (ScopeStart > Start + 1 &&
403 Changes[ScopeStart - 2].Tok->is(tok::identifier) &&
404 Changes[ScopeStart - 1].Tok->is(tok::l_brace) &&
405 Changes[i].Tok->is(tok::l_brace) &&
406 Changes[i].Tok->is(BK_BracedInit)) {
407 return true;
408 }
409
410 // Continued braced list.
411 if (ScopeStart > Start + 1 &&
412 Changes[ScopeStart - 2].Tok->isNot(tok::identifier) &&
413 Changes[ScopeStart - 1].Tok->is(tok::l_brace) &&
414 Changes[i].Tok->isNot(tok::r_brace)) {
415 for (unsigned OuterScopeStart : llvm::reverse(ScopeStack)) {
416 // Lambda.
417 if (OuterScopeStart > Start &&
418 Changes[OuterScopeStart - 1].Tok->is(TT_LambdaLBrace)) {
419 return false;
420 }
421 }
422 if (Changes[ScopeStart].NewlinesBefore > 0)
423 return false;
424 return true;
425 }
426
427 return false;
428 };
429
430 if (ShouldShiftBeAdded())
431 Changes[i].Spaces += Shift;
432 }
433
434 if (ContinuedStringLiteral)
435 Changes[i].Spaces += Shift;
436
437 // We should not remove required spaces unless we break the line before.
438 assert(Shift >= 0 || Changes[i].NewlinesBefore > 0 ||
439 Changes[i].Spaces >=
440 static_cast<int>(Changes[i].Tok->SpacesRequiredBefore) ||
441 Changes[i].Tok->is(tok::eof));
442
443 Changes[i].StartOfTokenColumn += Shift;
444 if (i + 1 != Changes.size())
445 Changes[i + 1].PreviousEndOfTokenColumn += Shift;
446
447 // If PointerAlignment is PAS_Right, keep *s or &s next to the token
448 if (Style.PointerAlignment == FormatStyle::PAS_Right &&
449 Changes[i].Spaces != 0) {
450 for (int Previous = i - 1;
451 Previous >= 0 &&
452 Changes[Previous].Tok->getType() == TT_PointerOrReference;
453 --Previous) {
454 Changes[Previous + 1].Spaces -= Shift;
455 Changes[Previous].Spaces += Shift;
456 Changes[Previous].StartOfTokenColumn += Shift;
457 }
458 }
459 }
460 }
461
462 // Walk through a subset of the changes, starting at StartAt, and find
463 // sequences of matching tokens to align. To do so, keep track of the lines and
464 // whether or not a matching token was found on a line. If a matching token is
465 // found, extend the current sequence. If the current line cannot be part of a
466 // sequence, e.g. because there is an empty line before it or it contains only
467 // non-matching tokens, finalize the previous sequence.
468 // The value returned is the token on which we stopped, either because we
469 // exhausted all items inside Changes, or because we hit a scope level higher
470 // than our initial scope.
471 // This function is recursive. Each invocation processes only the scope level
472 // equal to the initial level, which is the level of Changes[StartAt].
473 // If we encounter a scope level greater than the initial level, then we call
474 // ourselves recursively, thereby avoiding the pollution of the current state
475 // with the alignment requirements of the nested sub-level. This recursive
476 // behavior is necessary for aligning function prototypes that have one or more
477 // arguments.
478 // If this function encounters a scope level less than the initial level,
479 // it returns the current position.
480 // There is a non-obvious subtlety in the recursive behavior: Even though we
481 // defer processing of nested levels to recursive invocations of this
482 // function, when it comes time to align a sequence of tokens, we run the
483 // alignment on the entire sequence, including the nested levels.
484 // When doing so, most of the nested tokens are skipped, because their
485 // alignment was already handled by the recursive invocations of this function.
486 // However, the special exception is that we do NOT skip function parameters
487 // that are split across multiple lines. See the test case in FormatTest.cpp
488 // that mentions "split function parameter alignment" for an example of this.
489 // When the parameter RightJustify is true, the operator will be
490 // right-justified. It is used to align compound assignments like `+=` and `=`.
491 // When RightJustify and ACS.PadOperators are true, operators in each block to
492 // be aligned will be padded on the left to the same length before aligning.
493 template <typename F>
AlignTokens(const FormatStyle & Style,F && Matches,SmallVector<WhitespaceManager::Change,16> & Changes,unsigned StartAt,const FormatStyle::AlignConsecutiveStyle & ACS={},bool RightJustify=false)494 static unsigned AlignTokens(const FormatStyle &Style, F &&Matches,
495 SmallVector<WhitespaceManager::Change, 16> &Changes,
496 unsigned StartAt,
497 const FormatStyle::AlignConsecutiveStyle &ACS = {},
498 bool RightJustify = false) {
499 // We arrange each line in 3 parts. The operator to be aligned (the anchor),
500 // and text to its left and right. In the aligned text the width of each part
501 // will be the maximum of that over the block that has been aligned. Maximum
502 // widths of each part so far. When RightJustify is true and ACS.PadOperators
503 // is false, the part from start of line to the right end of the anchor.
504 // Otherwise, only the part to the left of the anchor. Including the space
505 // that exists on its left from the start. Not including the padding added on
506 // the left to right-justify the anchor.
507 unsigned WidthLeft = 0;
508 // The operator to be aligned when RightJustify is true and ACS.PadOperators
509 // is false. 0 otherwise.
510 unsigned WidthAnchor = 0;
511 // Width to the right of the anchor. Plus width of the anchor when
512 // RightJustify is false.
513 unsigned WidthRight = 0;
514
515 // Line number of the start and the end of the current token sequence.
516 unsigned StartOfSequence = 0;
517 unsigned EndOfSequence = 0;
518
519 // Measure the scope level (i.e. depth of (), [], {}) of the first token, and
520 // abort when we hit any token in a higher scope than the starting one.
521 auto IndentAndNestingLevel = StartAt < Changes.size()
522 ? Changes[StartAt].indentAndNestingLevel()
523 : std::tuple<unsigned, unsigned, unsigned>();
524
525 // Keep track if the first token has a non-zero indent and nesting level.
526 // This can happen when aligning the contents of "#else" preprocessor blocks,
527 // which is done separately.
528 bool HasInitialIndentAndNesting =
529 StartAt == 0 &&
530 IndentAndNestingLevel > std::tuple<unsigned, unsigned, unsigned>();
531
532 // Keep track of the number of commas before the matching tokens, we will only
533 // align a sequence of matching tokens if they are preceded by the same number
534 // of commas.
535 unsigned CommasBeforeLastMatch = 0;
536 unsigned CommasBeforeMatch = 0;
537
538 // Whether a matching token has been found on the current line.
539 bool FoundMatchOnLine = false;
540
541 // Whether the current line consists purely of comments.
542 bool LineIsComment = true;
543
544 // Aligns a sequence of matching tokens, on the MinColumn column.
545 //
546 // Sequences start from the first matching token to align, and end at the
547 // first token of the first line that doesn't need to be aligned.
548 //
549 // We need to adjust the StartOfTokenColumn of each Change that is on a line
550 // containing any matching token to be aligned and located after such token.
__anondea854270202null551 auto AlignCurrentSequence = [&] {
552 if (StartOfSequence > 0 && StartOfSequence < EndOfSequence) {
553 AlignTokenSequence(Style, StartOfSequence, EndOfSequence,
554 WidthLeft + WidthAnchor, RightJustify, Matches,
555 Changes);
556 }
557 WidthLeft = 0;
558 WidthAnchor = 0;
559 WidthRight = 0;
560 StartOfSequence = 0;
561 EndOfSequence = 0;
562 };
563
564 unsigned i = StartAt;
565 for (unsigned e = Changes.size(); i != e; ++i) {
566 if (Changes[i].indentAndNestingLevel() < IndentAndNestingLevel) {
567 if (!HasInitialIndentAndNesting)
568 break;
569 // The contents of preprocessor blocks are aligned separately.
570 // If the initial preprocessor block is indented or nested (e.g. it's in
571 // a function), do not align and exit after finishing this scope block.
572 // Instead, align, and then lower the baseline indent and nesting level
573 // in order to continue aligning subsequent blocks.
574 EndOfSequence = i;
575 AlignCurrentSequence();
576 IndentAndNestingLevel =
577 Changes[i].indentAndNestingLevel(); // new baseline
578 }
579
580 if (Changes[i].NewlinesBefore != 0) {
581 CommasBeforeMatch = 0;
582 EndOfSequence = i;
583
584 // Whether to break the alignment sequence because of an empty line.
585 bool EmptyLineBreak =
586 (Changes[i].NewlinesBefore > 1) && !ACS.AcrossEmptyLines;
587
588 // Whether to break the alignment sequence because of a line without a
589 // match.
590 bool NoMatchBreak =
591 !FoundMatchOnLine && !(LineIsComment && ACS.AcrossComments);
592
593 if (EmptyLineBreak || NoMatchBreak)
594 AlignCurrentSequence();
595
596 // A new line starts, re-initialize line status tracking bools.
597 // Keep the match state if a string literal is continued on this line.
598 if (i == 0 || !Changes[i].Tok->is(tok::string_literal) ||
599 !Changes[i - 1].Tok->is(tok::string_literal)) {
600 FoundMatchOnLine = false;
601 }
602 LineIsComment = true;
603 }
604
605 if (!Changes[i].Tok->is(tok::comment))
606 LineIsComment = false;
607
608 if (Changes[i].Tok->is(tok::comma)) {
609 ++CommasBeforeMatch;
610 } else if (Changes[i].indentAndNestingLevel() > IndentAndNestingLevel) {
611 // Call AlignTokens recursively, skipping over this scope block.
612 unsigned StoppedAt =
613 AlignTokens(Style, Matches, Changes, i, ACS, RightJustify);
614 i = StoppedAt - 1;
615 continue;
616 }
617
618 if (!Matches(Changes[i]))
619 continue;
620
621 // If there is more than one matching token per line, or if the number of
622 // preceding commas, do not match anymore, end the sequence.
623 if (FoundMatchOnLine || CommasBeforeMatch != CommasBeforeLastMatch)
624 AlignCurrentSequence();
625
626 CommasBeforeLastMatch = CommasBeforeMatch;
627 FoundMatchOnLine = true;
628
629 if (StartOfSequence == 0)
630 StartOfSequence = i;
631
632 unsigned ChangeWidthLeft = Changes[i].StartOfTokenColumn;
633 unsigned ChangeWidthAnchor = 0;
634 unsigned ChangeWidthRight = 0;
635 if (RightJustify)
636 if (ACS.PadOperators)
637 ChangeWidthAnchor = Changes[i].TokenLength;
638 else
639 ChangeWidthLeft += Changes[i].TokenLength;
640 else
641 ChangeWidthRight = Changes[i].TokenLength;
642 for (unsigned j = i + 1; j != e && Changes[j].NewlinesBefore == 0; ++j) {
643 ChangeWidthRight += Changes[j].Spaces;
644 // Changes are generally 1:1 with the tokens, but a change could also be
645 // inside of a token, in which case it's counted more than once: once for
646 // the whitespace surrounding the token (!IsInsideToken) and once for
647 // each whitespace change within it (IsInsideToken).
648 // Therefore, changes inside of a token should only count the space.
649 if (!Changes[j].IsInsideToken)
650 ChangeWidthRight += Changes[j].TokenLength;
651 }
652
653 // If we are restricted by the maximum column width, end the sequence.
654 unsigned NewLeft = std::max(ChangeWidthLeft, WidthLeft);
655 unsigned NewAnchor = std::max(ChangeWidthAnchor, WidthAnchor);
656 unsigned NewRight = std::max(ChangeWidthRight, WidthRight);
657 // `ColumnLimit == 0` means there is no column limit.
658 if (Style.ColumnLimit != 0 &&
659 Style.ColumnLimit < NewLeft + NewAnchor + NewRight) {
660 AlignCurrentSequence();
661 StartOfSequence = i;
662 WidthLeft = ChangeWidthLeft;
663 WidthAnchor = ChangeWidthAnchor;
664 WidthRight = ChangeWidthRight;
665 } else {
666 WidthLeft = NewLeft;
667 WidthAnchor = NewAnchor;
668 WidthRight = NewRight;
669 }
670 }
671
672 EndOfSequence = i;
673 AlignCurrentSequence();
674 return i;
675 }
676
677 // Aligns a sequence of matching tokens, on the MinColumn column.
678 //
679 // Sequences start from the first matching token to align, and end at the
680 // first token of the first line that doesn't need to be aligned.
681 //
682 // We need to adjust the StartOfTokenColumn of each Change that is on a line
683 // containing any matching token to be aligned and located after such token.
AlignMacroSequence(unsigned & StartOfSequence,unsigned & EndOfSequence,unsigned & MinColumn,unsigned & MaxColumn,bool & FoundMatchOnLine,std::function<bool (const WhitespaceManager::Change & C)> AlignMacrosMatches,SmallVector<WhitespaceManager::Change,16> & Changes)684 static void AlignMacroSequence(
685 unsigned &StartOfSequence, unsigned &EndOfSequence, unsigned &MinColumn,
686 unsigned &MaxColumn, bool &FoundMatchOnLine,
687 std::function<bool(const WhitespaceManager::Change &C)> AlignMacrosMatches,
688 SmallVector<WhitespaceManager::Change, 16> &Changes) {
689 if (StartOfSequence > 0 && StartOfSequence < EndOfSequence) {
690
691 FoundMatchOnLine = false;
692 int Shift = 0;
693
694 for (unsigned I = StartOfSequence; I != EndOfSequence; ++I) {
695 if (Changes[I].NewlinesBefore > 0) {
696 Shift = 0;
697 FoundMatchOnLine = false;
698 }
699
700 // If this is the first matching token to be aligned, remember by how many
701 // spaces it has to be shifted, so the rest of the changes on the line are
702 // shifted by the same amount
703 if (!FoundMatchOnLine && AlignMacrosMatches(Changes[I])) {
704 FoundMatchOnLine = true;
705 Shift = MinColumn - Changes[I].StartOfTokenColumn;
706 Changes[I].Spaces += Shift;
707 }
708
709 assert(Shift >= 0);
710 Changes[I].StartOfTokenColumn += Shift;
711 if (I + 1 != Changes.size())
712 Changes[I + 1].PreviousEndOfTokenColumn += Shift;
713 }
714 }
715
716 MinColumn = 0;
717 MaxColumn = UINT_MAX;
718 StartOfSequence = 0;
719 EndOfSequence = 0;
720 }
721
alignConsecutiveMacros()722 void WhitespaceManager::alignConsecutiveMacros() {
723 if (!Style.AlignConsecutiveMacros.Enabled)
724 return;
725
726 auto AlignMacrosMatches = [](const Change &C) {
727 const FormatToken *Current = C.Tok;
728 unsigned SpacesRequiredBefore = 1;
729
730 if (Current->SpacesRequiredBefore == 0 || !Current->Previous)
731 return false;
732
733 Current = Current->Previous;
734
735 // If token is a ")", skip over the parameter list, to the
736 // token that precedes the "("
737 if (Current->is(tok::r_paren) && Current->MatchingParen) {
738 Current = Current->MatchingParen->Previous;
739 SpacesRequiredBefore = 0;
740 }
741
742 if (!Current || !Current->is(tok::identifier))
743 return false;
744
745 if (!Current->Previous || !Current->Previous->is(tok::pp_define))
746 return false;
747
748 // For a macro function, 0 spaces are required between the
749 // identifier and the lparen that opens the parameter list.
750 // For a simple macro, 1 space is required between the
751 // identifier and the first token of the defined value.
752 return Current->Next->SpacesRequiredBefore == SpacesRequiredBefore;
753 };
754
755 unsigned MinColumn = 0;
756 unsigned MaxColumn = UINT_MAX;
757
758 // Start and end of the token sequence we're processing.
759 unsigned StartOfSequence = 0;
760 unsigned EndOfSequence = 0;
761
762 // Whether a matching token has been found on the current line.
763 bool FoundMatchOnLine = false;
764
765 // Whether the current line consists only of comments
766 bool LineIsComment = true;
767
768 unsigned I = 0;
769 for (unsigned E = Changes.size(); I != E; ++I) {
770 if (Changes[I].NewlinesBefore != 0) {
771 EndOfSequence = I;
772
773 // Whether to break the alignment sequence because of an empty line.
774 bool EmptyLineBreak = (Changes[I].NewlinesBefore > 1) &&
775 !Style.AlignConsecutiveMacros.AcrossEmptyLines;
776
777 // Whether to break the alignment sequence because of a line without a
778 // match.
779 bool NoMatchBreak =
780 !FoundMatchOnLine &&
781 !(LineIsComment && Style.AlignConsecutiveMacros.AcrossComments);
782
783 if (EmptyLineBreak || NoMatchBreak) {
784 AlignMacroSequence(StartOfSequence, EndOfSequence, MinColumn, MaxColumn,
785 FoundMatchOnLine, AlignMacrosMatches, Changes);
786 }
787
788 // A new line starts, re-initialize line status tracking bools.
789 FoundMatchOnLine = false;
790 LineIsComment = true;
791 }
792
793 if (!Changes[I].Tok->is(tok::comment))
794 LineIsComment = false;
795
796 if (!AlignMacrosMatches(Changes[I]))
797 continue;
798
799 FoundMatchOnLine = true;
800
801 if (StartOfSequence == 0)
802 StartOfSequence = I;
803
804 unsigned ChangeMinColumn = Changes[I].StartOfTokenColumn;
805 int LineLengthAfter = -Changes[I].Spaces;
806 for (unsigned j = I; j != E && Changes[j].NewlinesBefore == 0; ++j)
807 LineLengthAfter += Changes[j].Spaces + Changes[j].TokenLength;
808 unsigned ChangeMaxColumn = Style.ColumnLimit - LineLengthAfter;
809
810 MinColumn = std::max(MinColumn, ChangeMinColumn);
811 MaxColumn = std::min(MaxColumn, ChangeMaxColumn);
812 }
813
814 EndOfSequence = I;
815 AlignMacroSequence(StartOfSequence, EndOfSequence, MinColumn, MaxColumn,
816 FoundMatchOnLine, AlignMacrosMatches, Changes);
817 }
818
alignConsecutiveAssignments()819 void WhitespaceManager::alignConsecutiveAssignments() {
820 if (!Style.AlignConsecutiveAssignments.Enabled)
821 return;
822
823 AlignTokens(
824 Style,
825 [&](const Change &C) {
826 // Do not align on equal signs that are first on a line.
827 if (C.NewlinesBefore > 0)
828 return false;
829
830 // Do not align on equal signs that are last on a line.
831 if (&C != &Changes.back() && (&C + 1)->NewlinesBefore > 0)
832 return false;
833
834 // Do not align operator= overloads.
835 FormatToken *Previous = C.Tok->getPreviousNonComment();
836 if (Previous && Previous->is(tok::kw_operator))
837 return false;
838
839 return Style.AlignConsecutiveAssignments.AlignCompound
840 ? C.Tok->getPrecedence() == prec::Assignment
841 : C.Tok->is(tok::equal);
842 },
843 Changes, /*StartAt=*/0, Style.AlignConsecutiveAssignments,
844 /*RightJustify=*/true);
845 }
846
alignConsecutiveBitFields()847 void WhitespaceManager::alignConsecutiveBitFields() {
848 if (!Style.AlignConsecutiveBitFields.Enabled)
849 return;
850
851 AlignTokens(
852 Style,
853 [&](Change const &C) {
854 // Do not align on ':' that is first on a line.
855 if (C.NewlinesBefore > 0)
856 return false;
857
858 // Do not align on ':' that is last on a line.
859 if (&C != &Changes.back() && (&C + 1)->NewlinesBefore > 0)
860 return false;
861
862 return C.Tok->is(TT_BitFieldColon);
863 },
864 Changes, /*StartAt=*/0, Style.AlignConsecutiveBitFields);
865 }
866
alignConsecutiveDeclarations()867 void WhitespaceManager::alignConsecutiveDeclarations() {
868 if (!Style.AlignConsecutiveDeclarations.Enabled)
869 return;
870
871 AlignTokens(
872 Style,
873 [](Change const &C) {
874 if (C.Tok->is(TT_FunctionDeclarationName))
875 return true;
876 if (C.Tok->isNot(TT_StartOfName))
877 return false;
878 if (C.Tok->Previous &&
879 C.Tok->Previous->is(TT_StatementAttributeLikeMacro))
880 return false;
881 // Check if there is a subsequent name that starts the same declaration.
882 for (FormatToken *Next = C.Tok->Next; Next; Next = Next->Next) {
883 if (Next->is(tok::comment))
884 continue;
885 if (Next->is(TT_PointerOrReference))
886 return false;
887 if (!Next->Tok.getIdentifierInfo())
888 break;
889 if (Next->isOneOf(TT_StartOfName, TT_FunctionDeclarationName,
890 tok::kw_operator)) {
891 return false;
892 }
893 }
894 return true;
895 },
896 Changes, /*StartAt=*/0, Style.AlignConsecutiveDeclarations);
897 }
898
alignChainedConditionals()899 void WhitespaceManager::alignChainedConditionals() {
900 if (Style.BreakBeforeTernaryOperators) {
901 AlignTokens(
902 Style,
903 [](Change const &C) {
904 // Align question operators and last colon
905 return C.Tok->is(TT_ConditionalExpr) &&
906 ((C.Tok->is(tok::question) && !C.NewlinesBefore) ||
907 (C.Tok->is(tok::colon) && C.Tok->Next &&
908 (C.Tok->Next->FakeLParens.size() == 0 ||
909 C.Tok->Next->FakeLParens.back() != prec::Conditional)));
910 },
911 Changes, /*StartAt=*/0);
912 } else {
913 static auto AlignWrappedOperand = [](Change const &C) {
914 FormatToken *Previous = C.Tok->getPreviousNonComment();
915 return C.NewlinesBefore && Previous && Previous->is(TT_ConditionalExpr) &&
916 (Previous->is(tok::colon) &&
917 (C.Tok->FakeLParens.size() == 0 ||
918 C.Tok->FakeLParens.back() != prec::Conditional));
919 };
920 // Ensure we keep alignment of wrapped operands with non-wrapped operands
921 // Since we actually align the operators, the wrapped operands need the
922 // extra offset to be properly aligned.
923 for (Change &C : Changes)
924 if (AlignWrappedOperand(C))
925 C.StartOfTokenColumn -= 2;
926 AlignTokens(
927 Style,
928 [this](Change const &C) {
929 // Align question operators if next operand is not wrapped, as
930 // well as wrapped operands after question operator or last
931 // colon in conditional sequence
932 return (C.Tok->is(TT_ConditionalExpr) && C.Tok->is(tok::question) &&
933 &C != &Changes.back() && (&C + 1)->NewlinesBefore == 0 &&
934 !(&C + 1)->IsTrailingComment) ||
935 AlignWrappedOperand(C);
936 },
937 Changes, /*StartAt=*/0);
938 }
939 }
940
alignTrailingComments()941 void WhitespaceManager::alignTrailingComments() {
942 unsigned MinColumn = 0;
943 unsigned MaxColumn = UINT_MAX;
944 unsigned StartOfSequence = 0;
945 bool BreakBeforeNext = false;
946 unsigned Newlines = 0;
947 unsigned int NewLineThreshold = 1;
948 if (Style.AlignTrailingComments.Kind == FormatStyle::TCAS_Always)
949 NewLineThreshold = Style.AlignTrailingComments.OverEmptyLines + 1;
950
951 for (unsigned i = 0, e = Changes.size(); i != e; ++i) {
952 if (Changes[i].StartOfBlockComment)
953 continue;
954 Newlines += Changes[i].NewlinesBefore;
955 if (!Changes[i].IsTrailingComment)
956 continue;
957
958 if (Style.AlignTrailingComments.Kind == FormatStyle::TCAS_Leave) {
959 auto OriginalSpaces =
960 Changes[i].OriginalWhitespaceRange.getEnd().getRawEncoding() -
961 Changes[i].OriginalWhitespaceRange.getBegin().getRawEncoding() -
962 Changes[i].Tok->NewlinesBefore;
963 unsigned RestoredLineLength = Changes[i].StartOfTokenColumn +
964 Changes[i].TokenLength + OriginalSpaces;
965 // If leaving comments makes the line exceed the column limit, give up to
966 // leave the comments.
967 if (RestoredLineLength >= Style.ColumnLimit && Style.ColumnLimit != 0)
968 break;
969 Changes[i].Spaces = OriginalSpaces;
970 continue;
971 }
972
973 unsigned ChangeMinColumn = Changes[i].StartOfTokenColumn;
974 unsigned ChangeMaxColumn;
975
976 if (Style.ColumnLimit == 0)
977 ChangeMaxColumn = UINT_MAX;
978 else if (Style.ColumnLimit >= Changes[i].TokenLength)
979 ChangeMaxColumn = Style.ColumnLimit - Changes[i].TokenLength;
980 else
981 ChangeMaxColumn = ChangeMinColumn;
982
983 // If we don't create a replacement for this change, we have to consider
984 // it to be immovable.
985 if (!Changes[i].CreateReplacement)
986 ChangeMaxColumn = ChangeMinColumn;
987
988 if (i + 1 != e && Changes[i + 1].ContinuesPPDirective)
989 ChangeMaxColumn -= 2;
990 // If this comment follows an } in column 0, it probably documents the
991 // closing of a namespace and we don't want to align it.
992 bool FollowsRBraceInColumn0 = i > 0 && Changes[i].NewlinesBefore == 0 &&
993 Changes[i - 1].Tok->is(tok::r_brace) &&
994 Changes[i - 1].StartOfTokenColumn == 0;
995 bool WasAlignedWithStartOfNextLine = false;
996 if (Changes[i].NewlinesBefore >= 1) { // A comment on its own line.
997 unsigned CommentColumn = SourceMgr.getSpellingColumnNumber(
998 Changes[i].OriginalWhitespaceRange.getEnd());
999 for (unsigned j = i + 1; j != e; ++j) {
1000 if (Changes[j].Tok->is(tok::comment))
1001 continue;
1002
1003 unsigned NextColumn = SourceMgr.getSpellingColumnNumber(
1004 Changes[j].OriginalWhitespaceRange.getEnd());
1005 // The start of the next token was previously aligned with the
1006 // start of this comment.
1007 WasAlignedWithStartOfNextLine =
1008 CommentColumn == NextColumn ||
1009 CommentColumn == NextColumn + Style.IndentWidth;
1010 break;
1011 }
1012 }
1013 if (Style.AlignTrailingComments.Kind == FormatStyle::TCAS_Never ||
1014 FollowsRBraceInColumn0) {
1015 alignTrailingComments(StartOfSequence, i, MinColumn);
1016 MinColumn = ChangeMinColumn;
1017 MaxColumn = ChangeMinColumn;
1018 StartOfSequence = i;
1019 } else if (BreakBeforeNext || Newlines > NewLineThreshold ||
1020 (ChangeMinColumn > MaxColumn || ChangeMaxColumn < MinColumn) ||
1021 // Break the comment sequence if the previous line did not end
1022 // in a trailing comment.
1023 (Changes[i].NewlinesBefore == 1 && i > 0 &&
1024 !Changes[i - 1].IsTrailingComment) ||
1025 WasAlignedWithStartOfNextLine) {
1026 alignTrailingComments(StartOfSequence, i, MinColumn);
1027 MinColumn = ChangeMinColumn;
1028 MaxColumn = ChangeMaxColumn;
1029 StartOfSequence = i;
1030 } else {
1031 MinColumn = std::max(MinColumn, ChangeMinColumn);
1032 MaxColumn = std::min(MaxColumn, ChangeMaxColumn);
1033 }
1034 BreakBeforeNext = (i == 0) || (Changes[i].NewlinesBefore > 1) ||
1035 // Never start a sequence with a comment at the beginning
1036 // of the line.
1037 (Changes[i].NewlinesBefore == 1 && StartOfSequence == i);
1038 Newlines = 0;
1039 }
1040 alignTrailingComments(StartOfSequence, Changes.size(), MinColumn);
1041 }
1042
alignTrailingComments(unsigned Start,unsigned End,unsigned Column)1043 void WhitespaceManager::alignTrailingComments(unsigned Start, unsigned End,
1044 unsigned Column) {
1045 for (unsigned i = Start; i != End; ++i) {
1046 int Shift = 0;
1047 if (Changes[i].IsTrailingComment)
1048 Shift = Column - Changes[i].StartOfTokenColumn;
1049 if (Changes[i].StartOfBlockComment) {
1050 Shift = Changes[i].IndentationOffset +
1051 Changes[i].StartOfBlockComment->StartOfTokenColumn -
1052 Changes[i].StartOfTokenColumn;
1053 }
1054 if (Shift <= 0)
1055 continue;
1056 Changes[i].Spaces += Shift;
1057 if (i + 1 != Changes.size())
1058 Changes[i + 1].PreviousEndOfTokenColumn += Shift;
1059 Changes[i].StartOfTokenColumn += Shift;
1060 }
1061 }
1062
alignEscapedNewlines()1063 void WhitespaceManager::alignEscapedNewlines() {
1064 if (Style.AlignEscapedNewlines == FormatStyle::ENAS_DontAlign)
1065 return;
1066
1067 bool AlignLeft = Style.AlignEscapedNewlines == FormatStyle::ENAS_Left;
1068 unsigned MaxEndOfLine = AlignLeft ? 0 : Style.ColumnLimit;
1069 unsigned StartOfMacro = 0;
1070 for (unsigned i = 1, e = Changes.size(); i < e; ++i) {
1071 Change &C = Changes[i];
1072 if (C.NewlinesBefore > 0) {
1073 if (C.ContinuesPPDirective) {
1074 MaxEndOfLine = std::max(C.PreviousEndOfTokenColumn + 2, MaxEndOfLine);
1075 } else {
1076 alignEscapedNewlines(StartOfMacro + 1, i, MaxEndOfLine);
1077 MaxEndOfLine = AlignLeft ? 0 : Style.ColumnLimit;
1078 StartOfMacro = i;
1079 }
1080 }
1081 }
1082 alignEscapedNewlines(StartOfMacro + 1, Changes.size(), MaxEndOfLine);
1083 }
1084
alignEscapedNewlines(unsigned Start,unsigned End,unsigned Column)1085 void WhitespaceManager::alignEscapedNewlines(unsigned Start, unsigned End,
1086 unsigned Column) {
1087 for (unsigned i = Start; i < End; ++i) {
1088 Change &C = Changes[i];
1089 if (C.NewlinesBefore > 0) {
1090 assert(C.ContinuesPPDirective);
1091 if (C.PreviousEndOfTokenColumn + 1 > Column)
1092 C.EscapedNewlineColumn = 0;
1093 else
1094 C.EscapedNewlineColumn = Column;
1095 }
1096 }
1097 }
1098
alignArrayInitializers()1099 void WhitespaceManager::alignArrayInitializers() {
1100 if (Style.AlignArrayOfStructures == FormatStyle::AIAS_None)
1101 return;
1102
1103 for (unsigned ChangeIndex = 1U, ChangeEnd = Changes.size();
1104 ChangeIndex < ChangeEnd; ++ChangeIndex) {
1105 auto &C = Changes[ChangeIndex];
1106 if (C.Tok->IsArrayInitializer) {
1107 bool FoundComplete = false;
1108 for (unsigned InsideIndex = ChangeIndex + 1; InsideIndex < ChangeEnd;
1109 ++InsideIndex) {
1110 if (Changes[InsideIndex].Tok == C.Tok->MatchingParen) {
1111 alignArrayInitializers(ChangeIndex, InsideIndex + 1);
1112 ChangeIndex = InsideIndex + 1;
1113 FoundComplete = true;
1114 break;
1115 }
1116 }
1117 if (!FoundComplete)
1118 ChangeIndex = ChangeEnd;
1119 }
1120 }
1121 }
1122
alignArrayInitializers(unsigned Start,unsigned End)1123 void WhitespaceManager::alignArrayInitializers(unsigned Start, unsigned End) {
1124
1125 if (Style.AlignArrayOfStructures == FormatStyle::AIAS_Right)
1126 alignArrayInitializersRightJustified(getCells(Start, End));
1127 else if (Style.AlignArrayOfStructures == FormatStyle::AIAS_Left)
1128 alignArrayInitializersLeftJustified(getCells(Start, End));
1129 }
1130
alignArrayInitializersRightJustified(CellDescriptions && CellDescs)1131 void WhitespaceManager::alignArrayInitializersRightJustified(
1132 CellDescriptions &&CellDescs) {
1133 if (!CellDescs.isRectangular())
1134 return;
1135
1136 auto &Cells = CellDescs.Cells;
1137 // Now go through and fixup the spaces.
1138 auto *CellIter = Cells.begin();
1139 for (auto i = 0U; i < CellDescs.CellCounts[0]; ++i, ++CellIter) {
1140 unsigned NetWidth = 0U;
1141 if (isSplitCell(*CellIter))
1142 NetWidth = getNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces);
1143 auto CellWidth = getMaximumCellWidth(CellIter, NetWidth);
1144
1145 if (Changes[CellIter->Index].Tok->is(tok::r_brace)) {
1146 // So in here we want to see if there is a brace that falls
1147 // on a line that was split. If so on that line we make sure that
1148 // the spaces in front of the brace are enough.
1149 const auto *Next = CellIter;
1150 do {
1151 const FormatToken *Previous = Changes[Next->Index].Tok->Previous;
1152 if (Previous && Previous->isNot(TT_LineComment)) {
1153 Changes[Next->Index].Spaces = 0;
1154 Changes[Next->Index].NewlinesBefore = 0;
1155 }
1156 Next = Next->NextColumnElement;
1157 } while (Next);
1158 // Unless the array is empty, we need the position of all the
1159 // immediately adjacent cells
1160 if (CellIter != Cells.begin()) {
1161 auto ThisNetWidth =
1162 getNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces);
1163 auto MaxNetWidth = getMaximumNetWidth(
1164 Cells.begin(), CellIter, CellDescs.InitialSpaces,
1165 CellDescs.CellCounts[0], CellDescs.CellCounts.size());
1166 if (ThisNetWidth < MaxNetWidth)
1167 Changes[CellIter->Index].Spaces = (MaxNetWidth - ThisNetWidth);
1168 auto RowCount = 1U;
1169 auto Offset = std::distance(Cells.begin(), CellIter);
1170 for (const auto *Next = CellIter->NextColumnElement; Next != nullptr;
1171 Next = Next->NextColumnElement) {
1172 auto *Start = (Cells.begin() + RowCount * CellDescs.CellCounts[0]);
1173 auto *End = Start + Offset;
1174 ThisNetWidth = getNetWidth(Start, End, CellDescs.InitialSpaces);
1175 if (ThisNetWidth < MaxNetWidth)
1176 Changes[Next->Index].Spaces = (MaxNetWidth - ThisNetWidth);
1177 ++RowCount;
1178 }
1179 }
1180 } else {
1181 auto ThisWidth =
1182 calculateCellWidth(CellIter->Index, CellIter->EndIndex, true) +
1183 NetWidth;
1184 if (Changes[CellIter->Index].NewlinesBefore == 0) {
1185 Changes[CellIter->Index].Spaces = (CellWidth - (ThisWidth + NetWidth));
1186 Changes[CellIter->Index].Spaces += (i > 0) ? 1 : 0;
1187 }
1188 alignToStartOfCell(CellIter->Index, CellIter->EndIndex);
1189 for (const auto *Next = CellIter->NextColumnElement; Next != nullptr;
1190 Next = Next->NextColumnElement) {
1191 ThisWidth =
1192 calculateCellWidth(Next->Index, Next->EndIndex, true) + NetWidth;
1193 if (Changes[Next->Index].NewlinesBefore == 0) {
1194 Changes[Next->Index].Spaces = (CellWidth - ThisWidth);
1195 Changes[Next->Index].Spaces += (i > 0) ? 1 : 0;
1196 }
1197 alignToStartOfCell(Next->Index, Next->EndIndex);
1198 }
1199 }
1200 }
1201 }
1202
alignArrayInitializersLeftJustified(CellDescriptions && CellDescs)1203 void WhitespaceManager::alignArrayInitializersLeftJustified(
1204 CellDescriptions &&CellDescs) {
1205
1206 if (!CellDescs.isRectangular())
1207 return;
1208
1209 auto &Cells = CellDescs.Cells;
1210 // Now go through and fixup the spaces.
1211 auto *CellIter = Cells.begin();
1212 // The first cell needs to be against the left brace.
1213 if (Changes[CellIter->Index].NewlinesBefore == 0)
1214 Changes[CellIter->Index].Spaces = 0;
1215 else
1216 Changes[CellIter->Index].Spaces = CellDescs.InitialSpaces;
1217 ++CellIter;
1218 for (auto i = 1U; i < CellDescs.CellCounts[0]; i++, ++CellIter) {
1219 auto MaxNetWidth = getMaximumNetWidth(
1220 Cells.begin(), CellIter, CellDescs.InitialSpaces,
1221 CellDescs.CellCounts[0], CellDescs.CellCounts.size());
1222 auto ThisNetWidth =
1223 getNetWidth(Cells.begin(), CellIter, CellDescs.InitialSpaces);
1224 if (Changes[CellIter->Index].NewlinesBefore == 0) {
1225 Changes[CellIter->Index].Spaces =
1226 MaxNetWidth - ThisNetWidth +
1227 (Changes[CellIter->Index].Tok->isNot(tok::r_brace) ? 1 : 0);
1228 }
1229 auto RowCount = 1U;
1230 auto Offset = std::distance(Cells.begin(), CellIter);
1231 for (const auto *Next = CellIter->NextColumnElement; Next != nullptr;
1232 Next = Next->NextColumnElement) {
1233 if (RowCount > CellDescs.CellCounts.size())
1234 break;
1235 auto *Start = (Cells.begin() + RowCount * CellDescs.CellCounts[0]);
1236 auto *End = Start + Offset;
1237 auto ThisNetWidth = getNetWidth(Start, End, CellDescs.InitialSpaces);
1238 if (Changes[Next->Index].NewlinesBefore == 0) {
1239 Changes[Next->Index].Spaces =
1240 MaxNetWidth - ThisNetWidth +
1241 (Changes[Next->Index].Tok->isNot(tok::r_brace) ? 1 : 0);
1242 }
1243 ++RowCount;
1244 }
1245 }
1246 }
1247
isSplitCell(const CellDescription & Cell)1248 bool WhitespaceManager::isSplitCell(const CellDescription &Cell) {
1249 if (Cell.HasSplit)
1250 return true;
1251 for (const auto *Next = Cell.NextColumnElement; Next != nullptr;
1252 Next = Next->NextColumnElement) {
1253 if (Next->HasSplit)
1254 return true;
1255 }
1256 return false;
1257 }
1258
getCells(unsigned Start,unsigned End)1259 WhitespaceManager::CellDescriptions WhitespaceManager::getCells(unsigned Start,
1260 unsigned End) {
1261
1262 unsigned Depth = 0;
1263 unsigned Cell = 0;
1264 SmallVector<unsigned> CellCounts;
1265 unsigned InitialSpaces = 0;
1266 unsigned InitialTokenLength = 0;
1267 unsigned EndSpaces = 0;
1268 SmallVector<CellDescription> Cells;
1269 const FormatToken *MatchingParen = nullptr;
1270 for (unsigned i = Start; i < End; ++i) {
1271 auto &C = Changes[i];
1272 if (C.Tok->is(tok::l_brace))
1273 ++Depth;
1274 else if (C.Tok->is(tok::r_brace))
1275 --Depth;
1276 if (Depth == 2) {
1277 if (C.Tok->is(tok::l_brace)) {
1278 Cell = 0;
1279 MatchingParen = C.Tok->MatchingParen;
1280 if (InitialSpaces == 0) {
1281 InitialSpaces = C.Spaces + C.TokenLength;
1282 InitialTokenLength = C.TokenLength;
1283 auto j = i - 1;
1284 for (; Changes[j].NewlinesBefore == 0 && j > Start; --j) {
1285 InitialSpaces += Changes[j].Spaces + Changes[j].TokenLength;
1286 InitialTokenLength += Changes[j].TokenLength;
1287 }
1288 if (C.NewlinesBefore == 0) {
1289 InitialSpaces += Changes[j].Spaces + Changes[j].TokenLength;
1290 InitialTokenLength += Changes[j].TokenLength;
1291 }
1292 }
1293 } else if (C.Tok->is(tok::comma)) {
1294 if (!Cells.empty())
1295 Cells.back().EndIndex = i;
1296 if (C.Tok->getNextNonComment()->isNot(tok::r_brace)) // dangling comma
1297 ++Cell;
1298 }
1299 } else if (Depth == 1) {
1300 if (C.Tok == MatchingParen) {
1301 if (!Cells.empty())
1302 Cells.back().EndIndex = i;
1303 Cells.push_back(CellDescription{i, ++Cell, i + 1, false, nullptr});
1304 CellCounts.push_back(C.Tok->Previous->isNot(tok::comma) ? Cell + 1
1305 : Cell);
1306 // Go to the next non-comment and ensure there is a break in front
1307 const auto *NextNonComment = C.Tok->getNextNonComment();
1308 while (NextNonComment->is(tok::comma))
1309 NextNonComment = NextNonComment->getNextNonComment();
1310 auto j = i;
1311 while (Changes[j].Tok != NextNonComment && j < End)
1312 ++j;
1313 if (j < End && Changes[j].NewlinesBefore == 0 &&
1314 Changes[j].Tok->isNot(tok::r_brace)) {
1315 Changes[j].NewlinesBefore = 1;
1316 // Account for the added token lengths
1317 Changes[j].Spaces = InitialSpaces - InitialTokenLength;
1318 }
1319 } else if (C.Tok->is(tok::comment)) {
1320 // Trailing comments stay at a space past the last token
1321 C.Spaces = Changes[i - 1].Tok->is(tok::comma) ? 1 : 2;
1322 } else if (C.Tok->is(tok::l_brace)) {
1323 // We need to make sure that the ending braces is aligned to the
1324 // start of our initializer
1325 auto j = i - 1;
1326 for (; j > 0 && !Changes[j].Tok->ArrayInitializerLineStart; --j)
1327 ; // Nothing the loop does the work
1328 EndSpaces = Changes[j].Spaces;
1329 }
1330 } else if (Depth == 0 && C.Tok->is(tok::r_brace)) {
1331 C.NewlinesBefore = 1;
1332 C.Spaces = EndSpaces;
1333 }
1334 if (C.Tok->StartsColumn) {
1335 // This gets us past tokens that have been split over multiple
1336 // lines
1337 bool HasSplit = false;
1338 if (Changes[i].NewlinesBefore > 0) {
1339 // So if we split a line previously and the tail line + this token is
1340 // less then the column limit we remove the split here and just put
1341 // the column start at a space past the comma
1342 //
1343 // FIXME This if branch covers the cases where the column is not
1344 // the first column. This leads to weird pathologies like the formatting
1345 // auto foo = Items{
1346 // Section{
1347 // 0, bar(),
1348 // }
1349 // };
1350 // Well if it doesn't lead to that it's indicative that the line
1351 // breaking should be revisited. Unfortunately alot of other options
1352 // interact with this
1353 auto j = i - 1;
1354 if ((j - 1) > Start && Changes[j].Tok->is(tok::comma) &&
1355 Changes[j - 1].NewlinesBefore > 0) {
1356 --j;
1357 auto LineLimit = Changes[j].Spaces + Changes[j].TokenLength;
1358 if (LineLimit < Style.ColumnLimit) {
1359 Changes[i].NewlinesBefore = 0;
1360 Changes[i].Spaces = 1;
1361 }
1362 }
1363 }
1364 while (Changes[i].NewlinesBefore > 0 && Changes[i].Tok == C.Tok) {
1365 Changes[i].Spaces = InitialSpaces;
1366 ++i;
1367 HasSplit = true;
1368 }
1369 if (Changes[i].Tok != C.Tok)
1370 --i;
1371 Cells.push_back(CellDescription{i, Cell, i, HasSplit, nullptr});
1372 }
1373 }
1374
1375 return linkCells({Cells, CellCounts, InitialSpaces});
1376 }
1377
calculateCellWidth(unsigned Start,unsigned End,bool WithSpaces) const1378 unsigned WhitespaceManager::calculateCellWidth(unsigned Start, unsigned End,
1379 bool WithSpaces) const {
1380 unsigned CellWidth = 0;
1381 for (auto i = Start; i < End; i++) {
1382 if (Changes[i].NewlinesBefore > 0)
1383 CellWidth = 0;
1384 CellWidth += Changes[i].TokenLength;
1385 CellWidth += (WithSpaces ? Changes[i].Spaces : 0);
1386 }
1387 return CellWidth;
1388 }
1389
alignToStartOfCell(unsigned Start,unsigned End)1390 void WhitespaceManager::alignToStartOfCell(unsigned Start, unsigned End) {
1391 if ((End - Start) <= 1)
1392 return;
1393 // If the line is broken anywhere in there make sure everything
1394 // is aligned to the parent
1395 for (auto i = Start + 1; i < End; i++)
1396 if (Changes[i].NewlinesBefore > 0)
1397 Changes[i].Spaces = Changes[Start].Spaces;
1398 }
1399
1400 WhitespaceManager::CellDescriptions
linkCells(CellDescriptions && CellDesc)1401 WhitespaceManager::linkCells(CellDescriptions &&CellDesc) {
1402 auto &Cells = CellDesc.Cells;
1403 for (auto *CellIter = Cells.begin(); CellIter != Cells.end(); ++CellIter) {
1404 if (CellIter->NextColumnElement == nullptr &&
1405 ((CellIter + 1) != Cells.end())) {
1406 for (auto *NextIter = CellIter + 1; NextIter != Cells.end(); ++NextIter) {
1407 if (NextIter->Cell == CellIter->Cell) {
1408 CellIter->NextColumnElement = &(*NextIter);
1409 break;
1410 }
1411 }
1412 }
1413 }
1414 return std::move(CellDesc);
1415 }
1416
generateChanges()1417 void WhitespaceManager::generateChanges() {
1418 for (unsigned i = 0, e = Changes.size(); i != e; ++i) {
1419 const Change &C = Changes[i];
1420 if (i > 0 && Changes[i - 1].OriginalWhitespaceRange.getBegin() ==
1421 C.OriginalWhitespaceRange.getBegin()) {
1422 // Do not generate two replacements for the same location.
1423 continue;
1424 }
1425 if (C.CreateReplacement) {
1426 std::string ReplacementText = C.PreviousLinePostfix;
1427 if (C.ContinuesPPDirective) {
1428 appendEscapedNewlineText(ReplacementText, C.NewlinesBefore,
1429 C.PreviousEndOfTokenColumn,
1430 C.EscapedNewlineColumn);
1431 } else {
1432 appendNewlineText(ReplacementText, C.NewlinesBefore);
1433 }
1434 // FIXME: This assert should hold if we computed the column correctly.
1435 // assert((int)C.StartOfTokenColumn >= C.Spaces);
1436 appendIndentText(
1437 ReplacementText, C.Tok->IndentLevel, std::max(0, C.Spaces),
1438 std::max((int)C.StartOfTokenColumn, C.Spaces) - std::max(0, C.Spaces),
1439 C.IsAligned);
1440 ReplacementText.append(C.CurrentLinePrefix);
1441 storeReplacement(C.OriginalWhitespaceRange, ReplacementText);
1442 }
1443 }
1444 }
1445
storeReplacement(SourceRange Range,StringRef Text)1446 void WhitespaceManager::storeReplacement(SourceRange Range, StringRef Text) {
1447 unsigned WhitespaceLength = SourceMgr.getFileOffset(Range.getEnd()) -
1448 SourceMgr.getFileOffset(Range.getBegin());
1449 // Don't create a replacement, if it does not change anything.
1450 if (StringRef(SourceMgr.getCharacterData(Range.getBegin()),
1451 WhitespaceLength) == Text) {
1452 return;
1453 }
1454 auto Err = Replaces.add(tooling::Replacement(
1455 SourceMgr, CharSourceRange::getCharRange(Range), Text));
1456 // FIXME: better error handling. For now, just print an error message in the
1457 // release version.
1458 if (Err) {
1459 llvm::errs() << llvm::toString(std::move(Err)) << "\n";
1460 assert(false);
1461 }
1462 }
1463
appendNewlineText(std::string & Text,unsigned Newlines)1464 void WhitespaceManager::appendNewlineText(std::string &Text,
1465 unsigned Newlines) {
1466 if (UseCRLF) {
1467 Text.reserve(Text.size() + 2 * Newlines);
1468 for (unsigned i = 0; i < Newlines; ++i)
1469 Text.append("\r\n");
1470 } else {
1471 Text.append(Newlines, '\n');
1472 }
1473 }
1474
appendEscapedNewlineText(std::string & Text,unsigned Newlines,unsigned PreviousEndOfTokenColumn,unsigned EscapedNewlineColumn)1475 void WhitespaceManager::appendEscapedNewlineText(
1476 std::string &Text, unsigned Newlines, unsigned PreviousEndOfTokenColumn,
1477 unsigned EscapedNewlineColumn) {
1478 if (Newlines > 0) {
1479 unsigned Spaces =
1480 std::max<int>(1, EscapedNewlineColumn - PreviousEndOfTokenColumn - 1);
1481 for (unsigned i = 0; i < Newlines; ++i) {
1482 Text.append(Spaces, ' ');
1483 Text.append(UseCRLF ? "\\\r\n" : "\\\n");
1484 Spaces = std::max<int>(0, EscapedNewlineColumn - 1);
1485 }
1486 }
1487 }
1488
appendIndentText(std::string & Text,unsigned IndentLevel,unsigned Spaces,unsigned WhitespaceStartColumn,bool IsAligned)1489 void WhitespaceManager::appendIndentText(std::string &Text,
1490 unsigned IndentLevel, unsigned Spaces,
1491 unsigned WhitespaceStartColumn,
1492 bool IsAligned) {
1493 switch (Style.UseTab) {
1494 case FormatStyle::UT_Never:
1495 Text.append(Spaces, ' ');
1496 break;
1497 case FormatStyle::UT_Always: {
1498 if (Style.TabWidth) {
1499 unsigned FirstTabWidth =
1500 Style.TabWidth - WhitespaceStartColumn % Style.TabWidth;
1501
1502 // Insert only spaces when we want to end up before the next tab.
1503 if (Spaces < FirstTabWidth || Spaces == 1) {
1504 Text.append(Spaces, ' ');
1505 break;
1506 }
1507 // Align to the next tab.
1508 Spaces -= FirstTabWidth;
1509 Text.append("\t");
1510
1511 Text.append(Spaces / Style.TabWidth, '\t');
1512 Text.append(Spaces % Style.TabWidth, ' ');
1513 } else if (Spaces == 1) {
1514 Text.append(Spaces, ' ');
1515 }
1516 break;
1517 }
1518 case FormatStyle::UT_ForIndentation:
1519 if (WhitespaceStartColumn == 0) {
1520 unsigned Indentation = IndentLevel * Style.IndentWidth;
1521 Spaces = appendTabIndent(Text, Spaces, Indentation);
1522 }
1523 Text.append(Spaces, ' ');
1524 break;
1525 case FormatStyle::UT_ForContinuationAndIndentation:
1526 if (WhitespaceStartColumn == 0)
1527 Spaces = appendTabIndent(Text, Spaces, Spaces);
1528 Text.append(Spaces, ' ');
1529 break;
1530 case FormatStyle::UT_AlignWithSpaces:
1531 if (WhitespaceStartColumn == 0) {
1532 unsigned Indentation =
1533 IsAligned ? IndentLevel * Style.IndentWidth : Spaces;
1534 Spaces = appendTabIndent(Text, Spaces, Indentation);
1535 }
1536 Text.append(Spaces, ' ');
1537 break;
1538 }
1539 }
1540
appendTabIndent(std::string & Text,unsigned Spaces,unsigned Indentation)1541 unsigned WhitespaceManager::appendTabIndent(std::string &Text, unsigned Spaces,
1542 unsigned Indentation) {
1543 // This happens, e.g. when a line in a block comment is indented less than the
1544 // first one.
1545 if (Indentation > Spaces)
1546 Indentation = Spaces;
1547 if (Style.TabWidth) {
1548 unsigned Tabs = Indentation / Style.TabWidth;
1549 Text.append(Tabs, '\t');
1550 Spaces -= Tabs * Style.TabWidth;
1551 }
1552 return Spaces;
1553 }
1554
1555 } // namespace format
1556 } // namespace clang
1557