1 //===--- UnwrappedLineFormatter.cpp - Format C++ code ---------------------===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9
10 #include "UnwrappedLineFormatter.h"
11 #include "WhitespaceManager.h"
12 #include "llvm/Support/Debug.h"
13
14 #define DEBUG_TYPE "format-formatter"
15
16 namespace clang {
17 namespace format {
18
19 namespace {
20
startsExternCBlock(const AnnotatedLine & Line)21 bool startsExternCBlock(const AnnotatedLine &Line) {
22 const FormatToken *Next = Line.First->getNextNonComment();
23 const FormatToken *NextNext = Next ? Next->getNextNonComment() : nullptr;
24 return Line.startsWith(tok::kw_extern) && Next && Next->isStringLiteral() &&
25 NextNext && NextNext->is(tok::l_brace);
26 }
27
28 /// \brief Tracks the indent level of \c AnnotatedLines across levels.
29 ///
30 /// \c nextLine must be called for each \c AnnotatedLine, after which \c
31 /// getIndent() will return the indent for the last line \c nextLine was called
32 /// with.
33 /// If the line is not formatted (and thus the indent does not change), calling
34 /// \c adjustToUnmodifiedLine after the call to \c nextLine will cause
35 /// subsequent lines on the same level to be indented at the same level as the
36 /// given line.
37 class LevelIndentTracker {
38 public:
LevelIndentTracker(const FormatStyle & Style,const AdditionalKeywords & Keywords,unsigned StartLevel,int AdditionalIndent)39 LevelIndentTracker(const FormatStyle &Style,
40 const AdditionalKeywords &Keywords, unsigned StartLevel,
41 int AdditionalIndent)
42 : Style(Style), Keywords(Keywords), AdditionalIndent(AdditionalIndent) {
43 for (unsigned i = 0; i != StartLevel; ++i)
44 IndentForLevel.push_back(Style.IndentWidth * i + AdditionalIndent);
45 }
46
47 /// \brief Returns the indent for the current line.
getIndent() const48 unsigned getIndent() const { return Indent; }
49
50 /// \brief Update the indent state given that \p Line is going to be formatted
51 /// next.
nextLine(const AnnotatedLine & Line)52 void nextLine(const AnnotatedLine &Line) {
53 Offset = getIndentOffset(*Line.First);
54 // Update the indent level cache size so that we can rely on it
55 // having the right size in adjustToUnmodifiedline.
56 while (IndentForLevel.size() <= Line.Level)
57 IndentForLevel.push_back(-1);
58 if (Line.InPPDirective) {
59 Indent = Line.Level * Style.IndentWidth + AdditionalIndent;
60 } else {
61 IndentForLevel.resize(Line.Level + 1);
62 Indent = getIndent(IndentForLevel, Line.Level);
63 }
64 if (static_cast<int>(Indent) + Offset >= 0)
65 Indent += Offset;
66 }
67
68 /// \brief Update the level indent to adapt to the given \p Line.
69 ///
70 /// When a line is not formatted, we move the subsequent lines on the same
71 /// level to the same indent.
72 /// Note that \c nextLine must have been called before this method.
adjustToUnmodifiedLine(const AnnotatedLine & Line)73 void adjustToUnmodifiedLine(const AnnotatedLine &Line) {
74 unsigned LevelIndent = Line.First->OriginalColumn;
75 if (static_cast<int>(LevelIndent) - Offset >= 0)
76 LevelIndent -= Offset;
77 if ((!Line.First->is(tok::comment) || IndentForLevel[Line.Level] == -1) &&
78 !Line.InPPDirective)
79 IndentForLevel[Line.Level] = LevelIndent;
80 }
81
82 private:
83 /// \brief Get the offset of the line relatively to the level.
84 ///
85 /// For example, 'public:' labels in classes are offset by 1 or 2
86 /// characters to the left from their level.
getIndentOffset(const FormatToken & RootToken)87 int getIndentOffset(const FormatToken &RootToken) {
88 if (Style.Language == FormatStyle::LK_Java ||
89 Style.Language == FormatStyle::LK_JavaScript)
90 return 0;
91 if (RootToken.isAccessSpecifier(false) ||
92 RootToken.isObjCAccessSpecifier() ||
93 (RootToken.isOneOf(Keywords.kw_signals, Keywords.kw_qsignals) &&
94 RootToken.Next && RootToken.Next->is(tok::colon)))
95 return Style.AccessModifierOffset;
96 return 0;
97 }
98
99 /// \brief Get the indent of \p Level from \p IndentForLevel.
100 ///
101 /// \p IndentForLevel must contain the indent for the level \c l
102 /// at \p IndentForLevel[l], or a value < 0 if the indent for
103 /// that level is unknown.
getIndent(ArrayRef<int> IndentForLevel,unsigned Level)104 unsigned getIndent(ArrayRef<int> IndentForLevel, unsigned Level) {
105 if (IndentForLevel[Level] != -1)
106 return IndentForLevel[Level];
107 if (Level == 0)
108 return 0;
109 return getIndent(IndentForLevel, Level - 1) + Style.IndentWidth;
110 }
111
112 const FormatStyle &Style;
113 const AdditionalKeywords &Keywords;
114 const unsigned AdditionalIndent;
115
116 /// \brief The indent in characters for each level.
117 std::vector<int> IndentForLevel;
118
119 /// \brief Offset of the current line relative to the indent level.
120 ///
121 /// For example, the 'public' keywords is often indented with a negative
122 /// offset.
123 int Offset = 0;
124
125 /// \brief The current line's indent.
126 unsigned Indent = 0;
127 };
128
129 class LineJoiner {
130 public:
LineJoiner(const FormatStyle & Style,const AdditionalKeywords & Keywords,const SmallVectorImpl<AnnotatedLine * > & Lines)131 LineJoiner(const FormatStyle &Style, const AdditionalKeywords &Keywords,
132 const SmallVectorImpl<AnnotatedLine *> &Lines)
133 : Style(Style), Keywords(Keywords), End(Lines.end()),
134 Next(Lines.begin()) {}
135
136 /// \brief Returns the next line, merging multiple lines into one if possible.
getNextMergedLine(bool DryRun,LevelIndentTracker & IndentTracker)137 const AnnotatedLine *getNextMergedLine(bool DryRun,
138 LevelIndentTracker &IndentTracker) {
139 if (Next == End)
140 return nullptr;
141 const AnnotatedLine *Current = *Next;
142 IndentTracker.nextLine(*Current);
143 unsigned MergedLines =
144 tryFitMultipleLinesInOne(IndentTracker.getIndent(), Next, End);
145 if (MergedLines > 0 && Style.ColumnLimit == 0)
146 // Disallow line merging if there is a break at the start of one of the
147 // input lines.
148 for (unsigned i = 0; i < MergedLines; ++i)
149 if (Next[i + 1]->First->NewlinesBefore > 0)
150 MergedLines = 0;
151 if (!DryRun)
152 for (unsigned i = 0; i < MergedLines; ++i)
153 join(*Next[i], *Next[i + 1]);
154 Next = Next + MergedLines + 1;
155 return Current;
156 }
157
158 private:
159 /// \brief Calculates how many lines can be merged into 1 starting at \p I.
160 unsigned
tryFitMultipleLinesInOne(unsigned Indent,SmallVectorImpl<AnnotatedLine * >::const_iterator I,SmallVectorImpl<AnnotatedLine * >::const_iterator E)161 tryFitMultipleLinesInOne(unsigned Indent,
162 SmallVectorImpl<AnnotatedLine *>::const_iterator I,
163 SmallVectorImpl<AnnotatedLine *>::const_iterator E) {
164 // Can't join the last line with anything.
165 if (I + 1 == E)
166 return 0;
167 // We can never merge stuff if there are trailing line comments.
168 const AnnotatedLine *TheLine = *I;
169 if (TheLine->Last->is(TT_LineComment))
170 return 0;
171 if (I[1]->Type == LT_Invalid || I[1]->First->MustBreakBefore)
172 return 0;
173 if (TheLine->InPPDirective &&
174 (!I[1]->InPPDirective || I[1]->First->HasUnescapedNewline))
175 return 0;
176
177 if (Style.ColumnLimit > 0 && Indent > Style.ColumnLimit)
178 return 0;
179
180 unsigned Limit =
181 Style.ColumnLimit == 0 ? UINT_MAX : Style.ColumnLimit - Indent;
182 // If we already exceed the column limit, we set 'Limit' to 0. The different
183 // tryMerge..() functions can then decide whether to still do merging.
184 Limit = TheLine->Last->TotalLength > Limit
185 ? 0
186 : Limit - TheLine->Last->TotalLength;
187
188 // FIXME: TheLine->Level != 0 might or might not be the right check to do.
189 // If necessary, change to something smarter.
190 bool MergeShortFunctions =
191 Style.AllowShortFunctionsOnASingleLine == FormatStyle::SFS_All ||
192 (Style.AllowShortFunctionsOnASingleLine >= FormatStyle::SFS_Empty &&
193 I[1]->First->is(tok::r_brace)) ||
194 (Style.AllowShortFunctionsOnASingleLine == FormatStyle::SFS_Inline &&
195 TheLine->Level != 0);
196
197 if (TheLine->Last->is(TT_FunctionLBrace) &&
198 TheLine->First != TheLine->Last) {
199 return MergeShortFunctions ? tryMergeSimpleBlock(I, E, Limit) : 0;
200 }
201 if (TheLine->Last->is(tok::l_brace)) {
202 return !Style.BraceWrapping.AfterFunction
203 ? tryMergeSimpleBlock(I, E, Limit)
204 : 0;
205 }
206 if (I[1]->First->is(TT_FunctionLBrace) &&
207 Style.BraceWrapping.AfterFunction) {
208 if (I[1]->Last->is(TT_LineComment))
209 return 0;
210
211 // Check for Limit <= 2 to account for the " {".
212 if (Limit <= 2 || (Style.ColumnLimit == 0 && containsMustBreak(TheLine)))
213 return 0;
214 Limit -= 2;
215
216 unsigned MergedLines = 0;
217 if (MergeShortFunctions) {
218 MergedLines = tryMergeSimpleBlock(I + 1, E, Limit);
219 // If we managed to merge the block, count the function header, which is
220 // on a separate line.
221 if (MergedLines > 0)
222 ++MergedLines;
223 }
224 return MergedLines;
225 }
226 if (TheLine->First->is(tok::kw_if)) {
227 return Style.AllowShortIfStatementsOnASingleLine
228 ? tryMergeSimpleControlStatement(I, E, Limit)
229 : 0;
230 }
231 if (TheLine->First->isOneOf(tok::kw_for, tok::kw_while)) {
232 return Style.AllowShortLoopsOnASingleLine
233 ? tryMergeSimpleControlStatement(I, E, Limit)
234 : 0;
235 }
236 if (TheLine->First->isOneOf(tok::kw_case, tok::kw_default)) {
237 return Style.AllowShortCaseLabelsOnASingleLine
238 ? tryMergeShortCaseLabels(I, E, Limit)
239 : 0;
240 }
241 if (TheLine->InPPDirective &&
242 (TheLine->First->HasUnescapedNewline || TheLine->First->IsFirst)) {
243 return tryMergeSimplePPDirective(I, E, Limit);
244 }
245 return 0;
246 }
247
248 unsigned
tryMergeSimplePPDirective(SmallVectorImpl<AnnotatedLine * >::const_iterator I,SmallVectorImpl<AnnotatedLine * >::const_iterator E,unsigned Limit)249 tryMergeSimplePPDirective(SmallVectorImpl<AnnotatedLine *>::const_iterator I,
250 SmallVectorImpl<AnnotatedLine *>::const_iterator E,
251 unsigned Limit) {
252 if (Limit == 0)
253 return 0;
254 if (I + 2 != E && I[2]->InPPDirective && !I[2]->First->HasUnescapedNewline)
255 return 0;
256 if (1 + I[1]->Last->TotalLength > Limit)
257 return 0;
258 return 1;
259 }
260
tryMergeSimpleControlStatement(SmallVectorImpl<AnnotatedLine * >::const_iterator I,SmallVectorImpl<AnnotatedLine * >::const_iterator E,unsigned Limit)261 unsigned tryMergeSimpleControlStatement(
262 SmallVectorImpl<AnnotatedLine *>::const_iterator I,
263 SmallVectorImpl<AnnotatedLine *>::const_iterator E, unsigned Limit) {
264 if (Limit == 0)
265 return 0;
266 if (Style.BraceWrapping.AfterControlStatement &&
267 (I[1]->First->is(tok::l_brace) && !Style.AllowShortBlocksOnASingleLine))
268 return 0;
269 if (I[1]->InPPDirective != (*I)->InPPDirective ||
270 (I[1]->InPPDirective && I[1]->First->HasUnescapedNewline))
271 return 0;
272 Limit = limitConsideringMacros(I + 1, E, Limit);
273 AnnotatedLine &Line = **I;
274 if (Line.Last->isNot(tok::r_paren))
275 return 0;
276 if (1 + I[1]->Last->TotalLength > Limit)
277 return 0;
278 if (I[1]->First->isOneOf(tok::semi, tok::kw_if, tok::kw_for, tok::kw_while,
279 TT_LineComment))
280 return 0;
281 // Only inline simple if's (no nested if or else).
282 if (I + 2 != E && Line.startsWith(tok::kw_if) &&
283 I[2]->First->is(tok::kw_else))
284 return 0;
285 return 1;
286 }
287
288 unsigned
tryMergeShortCaseLabels(SmallVectorImpl<AnnotatedLine * >::const_iterator I,SmallVectorImpl<AnnotatedLine * >::const_iterator E,unsigned Limit)289 tryMergeShortCaseLabels(SmallVectorImpl<AnnotatedLine *>::const_iterator I,
290 SmallVectorImpl<AnnotatedLine *>::const_iterator E,
291 unsigned Limit) {
292 if (Limit == 0 || I + 1 == E ||
293 I[1]->First->isOneOf(tok::kw_case, tok::kw_default))
294 return 0;
295 unsigned NumStmts = 0;
296 unsigned Length = 0;
297 bool InPPDirective = I[0]->InPPDirective;
298 for (; NumStmts < 3; ++NumStmts) {
299 if (I + 1 + NumStmts == E)
300 break;
301 const AnnotatedLine *Line = I[1 + NumStmts];
302 if (Line->InPPDirective != InPPDirective)
303 break;
304 if (Line->First->isOneOf(tok::kw_case, tok::kw_default, tok::r_brace))
305 break;
306 if (Line->First->isOneOf(tok::kw_if, tok::kw_for, tok::kw_switch,
307 tok::kw_while, tok::comment) ||
308 Line->Last->is(tok::comment))
309 return 0;
310 Length += I[1 + NumStmts]->Last->TotalLength + 1; // 1 for the space.
311 }
312 if (NumStmts == 0 || NumStmts == 3 || Length > Limit)
313 return 0;
314 return NumStmts;
315 }
316
317 unsigned
tryMergeSimpleBlock(SmallVectorImpl<AnnotatedLine * >::const_iterator I,SmallVectorImpl<AnnotatedLine * >::const_iterator E,unsigned Limit)318 tryMergeSimpleBlock(SmallVectorImpl<AnnotatedLine *>::const_iterator I,
319 SmallVectorImpl<AnnotatedLine *>::const_iterator E,
320 unsigned Limit) {
321 AnnotatedLine &Line = **I;
322
323 // Don't merge ObjC @ keywords and methods.
324 // FIXME: If an option to allow short exception handling clauses on a single
325 // line is added, change this to not return for @try and friends.
326 if (Style.Language != FormatStyle::LK_Java &&
327 Line.First->isOneOf(tok::at, tok::minus, tok::plus))
328 return 0;
329
330 // Check that the current line allows merging. This depends on whether we
331 // are in a control flow statements as well as several style flags.
332 if (Line.First->isOneOf(tok::kw_else, tok::kw_case) ||
333 (Line.First->Next && Line.First->Next->is(tok::kw_else)))
334 return 0;
335 if (Line.First->isOneOf(tok::kw_if, tok::kw_while, tok::kw_do, tok::kw_try,
336 tok::kw___try, tok::kw_catch, tok::kw___finally,
337 tok::kw_for, tok::r_brace, Keywords.kw___except)) {
338 if (!Style.AllowShortBlocksOnASingleLine)
339 return 0;
340 if (!Style.AllowShortIfStatementsOnASingleLine &&
341 Line.startsWith(tok::kw_if))
342 return 0;
343 if (!Style.AllowShortLoopsOnASingleLine &&
344 Line.First->isOneOf(tok::kw_while, tok::kw_do, tok::kw_for))
345 return 0;
346 // FIXME: Consider an option to allow short exception handling clauses on
347 // a single line.
348 // FIXME: This isn't covered by tests.
349 // FIXME: For catch, __except, __finally the first token on the line
350 // is '}', so this isn't correct here.
351 if (Line.First->isOneOf(tok::kw_try, tok::kw___try, tok::kw_catch,
352 Keywords.kw___except, tok::kw___finally))
353 return 0;
354 }
355
356 FormatToken *Tok = I[1]->First;
357 if (Tok->is(tok::r_brace) && !Tok->MustBreakBefore &&
358 (Tok->getNextNonComment() == nullptr ||
359 Tok->getNextNonComment()->is(tok::semi))) {
360 // We merge empty blocks even if the line exceeds the column limit.
361 Tok->SpacesRequiredBefore = 0;
362 Tok->CanBreakBefore = true;
363 return 1;
364 } else if (Limit != 0 && !Line.startsWith(tok::kw_namespace) &&
365 !startsExternCBlock(Line)) {
366 // We don't merge short records.
367 if (Line.First->isOneOf(tok::kw_class, tok::kw_union, tok::kw_struct,
368 Keywords.kw_interface))
369 return 0;
370
371 // Check that we still have three lines and they fit into the limit.
372 if (I + 2 == E || I[2]->Type == LT_Invalid)
373 return 0;
374 Limit = limitConsideringMacros(I + 2, E, Limit);
375
376 if (!nextTwoLinesFitInto(I, Limit))
377 return 0;
378
379 // Second, check that the next line does not contain any braces - if it
380 // does, readability declines when putting it into a single line.
381 if (I[1]->Last->is(TT_LineComment))
382 return 0;
383 do {
384 if (Tok->is(tok::l_brace) && Tok->BlockKind != BK_BracedInit)
385 return 0;
386 Tok = Tok->Next;
387 } while (Tok);
388
389 // Last, check that the third line starts with a closing brace.
390 Tok = I[2]->First;
391 if (Tok->isNot(tok::r_brace))
392 return 0;
393
394 // Don't merge "if (a) { .. } else {".
395 if (Tok->Next && Tok->Next->is(tok::kw_else))
396 return 0;
397
398 return 2;
399 }
400 return 0;
401 }
402
403 /// Returns the modified column limit for \p I if it is inside a macro and
404 /// needs a trailing '\'.
405 unsigned
limitConsideringMacros(SmallVectorImpl<AnnotatedLine * >::const_iterator I,SmallVectorImpl<AnnotatedLine * >::const_iterator E,unsigned Limit)406 limitConsideringMacros(SmallVectorImpl<AnnotatedLine *>::const_iterator I,
407 SmallVectorImpl<AnnotatedLine *>::const_iterator E,
408 unsigned Limit) {
409 if (I[0]->InPPDirective && I + 1 != E &&
410 !I[1]->First->HasUnescapedNewline && !I[1]->First->is(tok::eof)) {
411 return Limit < 2 ? 0 : Limit - 2;
412 }
413 return Limit;
414 }
415
nextTwoLinesFitInto(SmallVectorImpl<AnnotatedLine * >::const_iterator I,unsigned Limit)416 bool nextTwoLinesFitInto(SmallVectorImpl<AnnotatedLine *>::const_iterator I,
417 unsigned Limit) {
418 if (I[1]->First->MustBreakBefore || I[2]->First->MustBreakBefore)
419 return false;
420 return 1 + I[1]->Last->TotalLength + 1 + I[2]->Last->TotalLength <= Limit;
421 }
422
containsMustBreak(const AnnotatedLine * Line)423 bool containsMustBreak(const AnnotatedLine *Line) {
424 for (const FormatToken *Tok = Line->First; Tok; Tok = Tok->Next) {
425 if (Tok->MustBreakBefore)
426 return true;
427 }
428 return false;
429 }
430
join(AnnotatedLine & A,const AnnotatedLine & B)431 void join(AnnotatedLine &A, const AnnotatedLine &B) {
432 assert(!A.Last->Next);
433 assert(!B.First->Previous);
434 if (B.Affected)
435 A.Affected = true;
436 A.Last->Next = B.First;
437 B.First->Previous = A.Last;
438 B.First->CanBreakBefore = true;
439 unsigned LengthA = A.Last->TotalLength + B.First->SpacesRequiredBefore;
440 for (FormatToken *Tok = B.First; Tok; Tok = Tok->Next) {
441 Tok->TotalLength += LengthA;
442 A.Last = Tok;
443 }
444 }
445
446 const FormatStyle &Style;
447 const AdditionalKeywords &Keywords;
448 const SmallVectorImpl<AnnotatedLine *>::const_iterator End;
449
450 SmallVectorImpl<AnnotatedLine *>::const_iterator Next;
451 };
452
markFinalized(FormatToken * Tok)453 static void markFinalized(FormatToken *Tok) {
454 for (; Tok; Tok = Tok->Next) {
455 Tok->Finalized = true;
456 for (AnnotatedLine *Child : Tok->Children)
457 markFinalized(Child->First);
458 }
459 }
460
461 #ifndef NDEBUG
printLineState(const LineState & State)462 static void printLineState(const LineState &State) {
463 llvm::dbgs() << "State: ";
464 for (const ParenState &P : State.Stack) {
465 llvm::dbgs() << P.Indent << "|" << P.LastSpace << "|" << P.NestedBlockIndent
466 << " ";
467 }
468 llvm::dbgs() << State.NextToken->TokenText << "\n";
469 }
470 #endif
471
472 /// \brief Base class for classes that format one \c AnnotatedLine.
473 class LineFormatter {
474 public:
LineFormatter(ContinuationIndenter * Indenter,WhitespaceManager * Whitespaces,const FormatStyle & Style,UnwrappedLineFormatter * BlockFormatter)475 LineFormatter(ContinuationIndenter *Indenter, WhitespaceManager *Whitespaces,
476 const FormatStyle &Style,
477 UnwrappedLineFormatter *BlockFormatter)
478 : Indenter(Indenter), Whitespaces(Whitespaces), Style(Style),
479 BlockFormatter(BlockFormatter) {}
~LineFormatter()480 virtual ~LineFormatter() {}
481
482 /// \brief Formats an \c AnnotatedLine and returns the penalty.
483 ///
484 /// If \p DryRun is \c false, directly applies the changes.
485 virtual unsigned formatLine(const AnnotatedLine &Line, unsigned FirstIndent,
486 bool DryRun) = 0;
487
488 protected:
489 /// \brief If the \p State's next token is an r_brace closing a nested block,
490 /// format the nested block before it.
491 ///
492 /// Returns \c true if all children could be placed successfully and adapts
493 /// \p Penalty as well as \p State. If \p DryRun is false, also directly
494 /// creates changes using \c Whitespaces.
495 ///
496 /// The crucial idea here is that children always get formatted upon
497 /// encountering the closing brace right after the nested block. Now, if we
498 /// are currently trying to keep the "}" on the same line (i.e. \p NewLine is
499 /// \c false), the entire block has to be kept on the same line (which is only
500 /// possible if it fits on the line, only contains a single statement, etc.
501 ///
502 /// If \p NewLine is true, we format the nested block on separate lines, i.e.
503 /// break after the "{", format all lines with correct indentation and the put
504 /// the closing "}" on yet another new line.
505 ///
506 /// This enables us to keep the simple structure of the
507 /// \c UnwrappedLineFormatter, where we only have two options for each token:
508 /// break or don't break.
formatChildren(LineState & State,bool NewLine,bool DryRun,unsigned & Penalty)509 bool formatChildren(LineState &State, bool NewLine, bool DryRun,
510 unsigned &Penalty) {
511 const FormatToken *LBrace = State.NextToken->getPreviousNonComment();
512 FormatToken &Previous = *State.NextToken->Previous;
513 if (!LBrace || LBrace->isNot(tok::l_brace) ||
514 LBrace->BlockKind != BK_Block || Previous.Children.size() == 0)
515 // The previous token does not open a block. Nothing to do. We don't
516 // assert so that we can simply call this function for all tokens.
517 return true;
518
519 if (NewLine) {
520 int AdditionalIndent = State.Stack.back().Indent -
521 Previous.Children[0]->Level * Style.IndentWidth;
522
523 Penalty +=
524 BlockFormatter->format(Previous.Children, DryRun, AdditionalIndent,
525 /*FixBadIndentation=*/true);
526 return true;
527 }
528
529 if (Previous.Children[0]->First->MustBreakBefore)
530 return false;
531
532 // Cannot merge multiple statements into a single line.
533 if (Previous.Children.size() > 1)
534 return false;
535
536 // Cannot merge into one line if this line ends on a comment.
537 if (Previous.is(tok::comment))
538 return false;
539
540 // We can't put the closing "}" on a line with a trailing comment.
541 if (Previous.Children[0]->Last->isTrailingComment())
542 return false;
543
544 // If the child line exceeds the column limit, we wouldn't want to merge it.
545 // We add +2 for the trailing " }".
546 if (Style.ColumnLimit > 0 &&
547 Previous.Children[0]->Last->TotalLength + State.Column + 2 >
548 Style.ColumnLimit)
549 return false;
550
551 if (!DryRun) {
552 Whitespaces->replaceWhitespace(
553 *Previous.Children[0]->First,
554 /*Newlines=*/0, /*IndentLevel=*/0, /*Spaces=*/1,
555 /*StartOfTokenColumn=*/State.Column, State.Line->InPPDirective);
556 }
557 Penalty += formatLine(*Previous.Children[0], State.Column + 1, DryRun);
558
559 State.Column += 1 + Previous.Children[0]->Last->TotalLength;
560 return true;
561 }
562
563 ContinuationIndenter *Indenter;
564
565 private:
566 WhitespaceManager *Whitespaces;
567 const FormatStyle &Style;
568 UnwrappedLineFormatter *BlockFormatter;
569 };
570
571 /// \brief Formatter that keeps the existing line breaks.
572 class NoColumnLimitLineFormatter : public LineFormatter {
573 public:
NoColumnLimitLineFormatter(ContinuationIndenter * Indenter,WhitespaceManager * Whitespaces,const FormatStyle & Style,UnwrappedLineFormatter * BlockFormatter)574 NoColumnLimitLineFormatter(ContinuationIndenter *Indenter,
575 WhitespaceManager *Whitespaces,
576 const FormatStyle &Style,
577 UnwrappedLineFormatter *BlockFormatter)
578 : LineFormatter(Indenter, Whitespaces, Style, BlockFormatter) {}
579
580 /// \brief Formats the line, simply keeping all of the input's line breaking
581 /// decisions.
formatLine(const AnnotatedLine & Line,unsigned FirstIndent,bool DryRun)582 unsigned formatLine(const AnnotatedLine &Line, unsigned FirstIndent,
583 bool DryRun) override {
584 assert(!DryRun);
585 LineState State =
586 Indenter->getInitialState(FirstIndent, &Line, /*DryRun=*/false);
587 while (State.NextToken) {
588 bool Newline =
589 Indenter->mustBreak(State) ||
590 (Indenter->canBreak(State) && State.NextToken->NewlinesBefore > 0);
591 unsigned Penalty = 0;
592 formatChildren(State, Newline, /*DryRun=*/false, Penalty);
593 Indenter->addTokenToState(State, Newline, /*DryRun=*/false);
594 }
595 return 0;
596 }
597 };
598
599 /// \brief Formatter that puts all tokens into a single line without breaks.
600 class NoLineBreakFormatter : public LineFormatter {
601 public:
NoLineBreakFormatter(ContinuationIndenter * Indenter,WhitespaceManager * Whitespaces,const FormatStyle & Style,UnwrappedLineFormatter * BlockFormatter)602 NoLineBreakFormatter(ContinuationIndenter *Indenter,
603 WhitespaceManager *Whitespaces, const FormatStyle &Style,
604 UnwrappedLineFormatter *BlockFormatter)
605 : LineFormatter(Indenter, Whitespaces, Style, BlockFormatter) {}
606
607 /// \brief Puts all tokens into a single line.
formatLine(const AnnotatedLine & Line,unsigned FirstIndent,bool DryRun)608 unsigned formatLine(const AnnotatedLine &Line, unsigned FirstIndent,
609 bool DryRun) override {
610 unsigned Penalty = 0;
611 LineState State = Indenter->getInitialState(FirstIndent, &Line, DryRun);
612 while (State.NextToken) {
613 formatChildren(State, /*Newline=*/false, DryRun, Penalty);
614 Indenter->addTokenToState(State, /*Newline=*/false, DryRun);
615 }
616 return Penalty;
617 }
618 };
619
620 /// \brief Finds the best way to break lines.
621 class OptimizingLineFormatter : public LineFormatter {
622 public:
OptimizingLineFormatter(ContinuationIndenter * Indenter,WhitespaceManager * Whitespaces,const FormatStyle & Style,UnwrappedLineFormatter * BlockFormatter)623 OptimizingLineFormatter(ContinuationIndenter *Indenter,
624 WhitespaceManager *Whitespaces,
625 const FormatStyle &Style,
626 UnwrappedLineFormatter *BlockFormatter)
627 : LineFormatter(Indenter, Whitespaces, Style, BlockFormatter) {}
628
629 /// \brief Formats the line by finding the best line breaks with line lengths
630 /// below the column limit.
formatLine(const AnnotatedLine & Line,unsigned FirstIndent,bool DryRun)631 unsigned formatLine(const AnnotatedLine &Line, unsigned FirstIndent,
632 bool DryRun) override {
633 LineState State = Indenter->getInitialState(FirstIndent, &Line, DryRun);
634
635 // If the ObjC method declaration does not fit on a line, we should format
636 // it with one arg per line.
637 if (State.Line->Type == LT_ObjCMethodDecl)
638 State.Stack.back().BreakBeforeParameter = true;
639
640 // Find best solution in solution space.
641 return analyzeSolutionSpace(State, DryRun);
642 }
643
644 private:
645 struct CompareLineStatePointers {
operator ()clang::format::__anon5061c3a60111::OptimizingLineFormatter::CompareLineStatePointers646 bool operator()(LineState *obj1, LineState *obj2) const {
647 return *obj1 < *obj2;
648 }
649 };
650
651 /// \brief A pair of <penalty, count> that is used to prioritize the BFS on.
652 ///
653 /// In case of equal penalties, we want to prefer states that were inserted
654 /// first. During state generation we make sure that we insert states first
655 /// that break the line as late as possible.
656 typedef std::pair<unsigned, unsigned> OrderedPenalty;
657
658 /// \brief An edge in the solution space from \c Previous->State to \c State,
659 /// inserting a newline dependent on the \c NewLine.
660 struct StateNode {
StateNodeclang::format::__anon5061c3a60111::OptimizingLineFormatter::StateNode661 StateNode(const LineState &State, bool NewLine, StateNode *Previous)
662 : State(State), NewLine(NewLine), Previous(Previous) {}
663 LineState State;
664 bool NewLine;
665 StateNode *Previous;
666 };
667
668 /// \brief An item in the prioritized BFS search queue. The \c StateNode's
669 /// \c State has the given \c OrderedPenalty.
670 typedef std::pair<OrderedPenalty, StateNode *> QueueItem;
671
672 /// \brief The BFS queue type.
673 typedef std::priority_queue<QueueItem, std::vector<QueueItem>,
674 std::greater<QueueItem>> QueueType;
675
676 /// \brief Analyze the entire solution space starting from \p InitialState.
677 ///
678 /// This implements a variant of Dijkstra's algorithm on the graph that spans
679 /// the solution space (\c LineStates are the nodes). The algorithm tries to
680 /// find the shortest path (the one with lowest penalty) from \p InitialState
681 /// to a state where all tokens are placed. Returns the penalty.
682 ///
683 /// If \p DryRun is \c false, directly applies the changes.
analyzeSolutionSpace(LineState & InitialState,bool DryRun)684 unsigned analyzeSolutionSpace(LineState &InitialState, bool DryRun) {
685 std::set<LineState *, CompareLineStatePointers> Seen;
686
687 // Increasing count of \c StateNode items we have created. This is used to
688 // create a deterministic order independent of the container.
689 unsigned Count = 0;
690 QueueType Queue;
691
692 // Insert start element into queue.
693 StateNode *Node =
694 new (Allocator.Allocate()) StateNode(InitialState, false, nullptr);
695 Queue.push(QueueItem(OrderedPenalty(0, Count), Node));
696 ++Count;
697
698 unsigned Penalty = 0;
699
700 // While not empty, take first element and follow edges.
701 while (!Queue.empty()) {
702 Penalty = Queue.top().first.first;
703 StateNode *Node = Queue.top().second;
704 if (!Node->State.NextToken) {
705 DEBUG(llvm::dbgs() << "\n---\nPenalty for line: " << Penalty << "\n");
706 break;
707 }
708 Queue.pop();
709
710 // Cut off the analysis of certain solutions if the analysis gets too
711 // complex. See description of IgnoreStackForComparison.
712 if (Count > 50000)
713 Node->State.IgnoreStackForComparison = true;
714
715 if (!Seen.insert(&Node->State).second)
716 // State already examined with lower penalty.
717 continue;
718
719 FormatDecision LastFormat = Node->State.NextToken->Decision;
720 if (LastFormat == FD_Unformatted || LastFormat == FD_Continue)
721 addNextStateToQueue(Penalty, Node, /*NewLine=*/false, &Count, &Queue);
722 if (LastFormat == FD_Unformatted || LastFormat == FD_Break)
723 addNextStateToQueue(Penalty, Node, /*NewLine=*/true, &Count, &Queue);
724 }
725
726 if (Queue.empty()) {
727 // We were unable to find a solution, do nothing.
728 // FIXME: Add diagnostic?
729 DEBUG(llvm::dbgs() << "Could not find a solution.\n");
730 return 0;
731 }
732
733 // Reconstruct the solution.
734 if (!DryRun)
735 reconstructPath(InitialState, Queue.top().second);
736
737 DEBUG(llvm::dbgs() << "Total number of analyzed states: " << Count << "\n");
738 DEBUG(llvm::dbgs() << "---\n");
739
740 return Penalty;
741 }
742
743 /// \brief Add the following state to the analysis queue \c Queue.
744 ///
745 /// Assume the current state is \p PreviousNode and has been reached with a
746 /// penalty of \p Penalty. Insert a line break if \p NewLine is \c true.
addNextStateToQueue(unsigned Penalty,StateNode * PreviousNode,bool NewLine,unsigned * Count,QueueType * Queue)747 void addNextStateToQueue(unsigned Penalty, StateNode *PreviousNode,
748 bool NewLine, unsigned *Count, QueueType *Queue) {
749 if (NewLine && !Indenter->canBreak(PreviousNode->State))
750 return;
751 if (!NewLine && Indenter->mustBreak(PreviousNode->State))
752 return;
753
754 StateNode *Node = new (Allocator.Allocate())
755 StateNode(PreviousNode->State, NewLine, PreviousNode);
756 if (!formatChildren(Node->State, NewLine, /*DryRun=*/true, Penalty))
757 return;
758
759 Penalty += Indenter->addTokenToState(Node->State, NewLine, true);
760
761 Queue->push(QueueItem(OrderedPenalty(Penalty, *Count), Node));
762 ++(*Count);
763 }
764
765 /// \brief Applies the best formatting by reconstructing the path in the
766 /// solution space that leads to \c Best.
reconstructPath(LineState & State,StateNode * Best)767 void reconstructPath(LineState &State, StateNode *Best) {
768 std::deque<StateNode *> Path;
769 // We do not need a break before the initial token.
770 while (Best->Previous) {
771 Path.push_front(Best);
772 Best = Best->Previous;
773 }
774 for (std::deque<StateNode *>::iterator I = Path.begin(), E = Path.end();
775 I != E; ++I) {
776 unsigned Penalty = 0;
777 formatChildren(State, (*I)->NewLine, /*DryRun=*/false, Penalty);
778 Penalty += Indenter->addTokenToState(State, (*I)->NewLine, false);
779
780 DEBUG({
781 printLineState((*I)->Previous->State);
782 if ((*I)->NewLine) {
783 llvm::dbgs() << "Penalty for placing "
784 << (*I)->Previous->State.NextToken->Tok.getName() << ": "
785 << Penalty << "\n";
786 }
787 });
788 }
789 }
790
791 llvm::SpecificBumpPtrAllocator<StateNode> Allocator;
792 };
793
794 } // anonymous namespace
795
796 unsigned
format(const SmallVectorImpl<AnnotatedLine * > & Lines,bool DryRun,int AdditionalIndent,bool FixBadIndentation)797 UnwrappedLineFormatter::format(const SmallVectorImpl<AnnotatedLine *> &Lines,
798 bool DryRun, int AdditionalIndent,
799 bool FixBadIndentation) {
800 LineJoiner Joiner(Style, Keywords, Lines);
801
802 // Try to look up already computed penalty in DryRun-mode.
803 std::pair<const SmallVectorImpl<AnnotatedLine *> *, unsigned> CacheKey(
804 &Lines, AdditionalIndent);
805 auto CacheIt = PenaltyCache.find(CacheKey);
806 if (DryRun && CacheIt != PenaltyCache.end())
807 return CacheIt->second;
808
809 assert(!Lines.empty());
810 unsigned Penalty = 0;
811 LevelIndentTracker IndentTracker(Style, Keywords, Lines[0]->Level,
812 AdditionalIndent);
813 const AnnotatedLine *PreviousLine = nullptr;
814 const AnnotatedLine *NextLine = nullptr;
815
816 // The minimum level of consecutive lines that have been formatted.
817 unsigned RangeMinLevel = UINT_MAX;
818
819 for (const AnnotatedLine *Line =
820 Joiner.getNextMergedLine(DryRun, IndentTracker);
821 Line; Line = NextLine) {
822 const AnnotatedLine &TheLine = *Line;
823 unsigned Indent = IndentTracker.getIndent();
824
825 // We continue formatting unchanged lines to adjust their indent, e.g. if a
826 // scope was added. However, we need to carefully stop doing this when we
827 // exit the scope of affected lines to prevent indenting a the entire
828 // remaining file if it currently missing a closing brace.
829 bool ContinueFormatting =
830 TheLine.Level > RangeMinLevel ||
831 (TheLine.Level == RangeMinLevel && !TheLine.startsWith(tok::r_brace));
832
833 bool FixIndentation = (FixBadIndentation || ContinueFormatting) &&
834 Indent != TheLine.First->OriginalColumn;
835 bool ShouldFormat = TheLine.Affected || FixIndentation;
836 // We cannot format this line; if the reason is that the line had a
837 // parsing error, remember that.
838 if (ShouldFormat && TheLine.Type == LT_Invalid && IncompleteFormat)
839 *IncompleteFormat = true;
840
841 if (ShouldFormat && TheLine.Type != LT_Invalid) {
842 if (!DryRun)
843 formatFirstToken(*TheLine.First, PreviousLine, TheLine.Level, Indent,
844 TheLine.InPPDirective);
845
846 NextLine = Joiner.getNextMergedLine(DryRun, IndentTracker);
847 unsigned ColumnLimit = getColumnLimit(TheLine.InPPDirective, NextLine);
848 bool FitsIntoOneLine =
849 TheLine.Last->TotalLength + Indent <= ColumnLimit ||
850 TheLine.Type == LT_ImportStatement;
851
852 if (Style.ColumnLimit == 0)
853 NoColumnLimitLineFormatter(Indenter, Whitespaces, Style, this)
854 .formatLine(TheLine, Indent, DryRun);
855 else if (FitsIntoOneLine)
856 Penalty += NoLineBreakFormatter(Indenter, Whitespaces, Style, this)
857 .formatLine(TheLine, Indent, DryRun);
858 else
859 Penalty += OptimizingLineFormatter(Indenter, Whitespaces, Style, this)
860 .formatLine(TheLine, Indent, DryRun);
861 RangeMinLevel = std::min(RangeMinLevel, TheLine.Level);
862 } else {
863 // If no token in the current line is affected, we still need to format
864 // affected children.
865 if (TheLine.ChildrenAffected)
866 format(TheLine.Children, DryRun);
867
868 // Adapt following lines on the current indent level to the same level
869 // unless the current \c AnnotatedLine is not at the beginning of a line.
870 bool StartsNewLine =
871 TheLine.First->NewlinesBefore > 0 || TheLine.First->IsFirst;
872 if (StartsNewLine)
873 IndentTracker.adjustToUnmodifiedLine(TheLine);
874 if (!DryRun) {
875 bool ReformatLeadingWhitespace =
876 StartsNewLine && ((PreviousLine && PreviousLine->Affected) ||
877 TheLine.LeadingEmptyLinesAffected);
878 // Format the first token.
879 if (ReformatLeadingWhitespace)
880 formatFirstToken(*TheLine.First, PreviousLine, TheLine.Level,
881 TheLine.First->OriginalColumn,
882 TheLine.InPPDirective);
883 else
884 Whitespaces->addUntouchableToken(*TheLine.First,
885 TheLine.InPPDirective);
886
887 // Notify the WhitespaceManager about the unchanged whitespace.
888 for (FormatToken *Tok = TheLine.First->Next; Tok; Tok = Tok->Next)
889 Whitespaces->addUntouchableToken(*Tok, TheLine.InPPDirective);
890 }
891 NextLine = Joiner.getNextMergedLine(DryRun, IndentTracker);
892 RangeMinLevel = UINT_MAX;
893 }
894 if (!DryRun)
895 markFinalized(TheLine.First);
896 PreviousLine = &TheLine;
897 }
898 PenaltyCache[CacheKey] = Penalty;
899 return Penalty;
900 }
901
formatFirstToken(FormatToken & RootToken,const AnnotatedLine * PreviousLine,unsigned IndentLevel,unsigned Indent,bool InPPDirective)902 void UnwrappedLineFormatter::formatFirstToken(FormatToken &RootToken,
903 const AnnotatedLine *PreviousLine,
904 unsigned IndentLevel,
905 unsigned Indent,
906 bool InPPDirective) {
907 if (RootToken.is(tok::eof)) {
908 unsigned Newlines = std::min(RootToken.NewlinesBefore, 1u);
909 Whitespaces->replaceWhitespace(RootToken, Newlines, /*IndentLevel=*/0,
910 /*Spaces=*/0, /*TargetColumn=*/0);
911 return;
912 }
913 unsigned Newlines =
914 std::min(RootToken.NewlinesBefore, Style.MaxEmptyLinesToKeep + 1);
915 // Remove empty lines before "}" where applicable.
916 if (RootToken.is(tok::r_brace) &&
917 (!RootToken.Next ||
918 (RootToken.Next->is(tok::semi) && !RootToken.Next->Next)))
919 Newlines = std::min(Newlines, 1u);
920 if (Newlines == 0 && !RootToken.IsFirst)
921 Newlines = 1;
922 if (RootToken.IsFirst && !RootToken.HasUnescapedNewline)
923 Newlines = 0;
924
925 // Remove empty lines after "{".
926 if (!Style.KeepEmptyLinesAtTheStartOfBlocks && PreviousLine &&
927 PreviousLine->Last->is(tok::l_brace) &&
928 PreviousLine->First->isNot(tok::kw_namespace) &&
929 !startsExternCBlock(*PreviousLine))
930 Newlines = 1;
931
932 // Insert extra new line before access specifiers.
933 if (PreviousLine && PreviousLine->Last->isOneOf(tok::semi, tok::r_brace) &&
934 RootToken.isAccessSpecifier() && RootToken.NewlinesBefore == 1)
935 ++Newlines;
936
937 // Remove empty lines after access specifiers.
938 if (PreviousLine && PreviousLine->First->isAccessSpecifier() &&
939 (!PreviousLine->InPPDirective || !RootToken.HasUnescapedNewline))
940 Newlines = std::min(1u, Newlines);
941
942 Whitespaces->replaceWhitespace(RootToken, Newlines, IndentLevel, Indent,
943 Indent, InPPDirective &&
944 !RootToken.HasUnescapedNewline);
945 }
946
947 unsigned
getColumnLimit(bool InPPDirective,const AnnotatedLine * NextLine) const948 UnwrappedLineFormatter::getColumnLimit(bool InPPDirective,
949 const AnnotatedLine *NextLine) const {
950 // In preprocessor directives reserve two chars for trailing " \" if the
951 // next line continues the preprocessor directive.
952 bool ContinuesPPDirective =
953 InPPDirective &&
954 // If there is no next line, this is likely a child line and the parent
955 // continues the preprocessor directive.
956 (!NextLine ||
957 (NextLine->InPPDirective &&
958 // If there is an unescaped newline between this line and the next, the
959 // next line starts a new preprocessor directive.
960 !NextLine->First->HasUnescapedNewline));
961 return Style.ColumnLimit - (ContinuesPPDirective ? 2 : 0);
962 }
963
964 } // namespace format
965 } // namespace clang
966