150
|
1 //===--- UnwrappedLineFormatter.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 #include "UnwrappedLineFormatter.h"
|
|
10 #include "NamespaceEndCommentsFixer.h"
|
|
11 #include "WhitespaceManager.h"
|
|
12 #include "llvm/Support/Debug.h"
|
|
13 #include <queue>
|
|
14
|
|
15 #define DEBUG_TYPE "format-formatter"
|
|
16
|
|
17 namespace clang {
|
|
18 namespace format {
|
|
19
|
|
20 namespace {
|
|
21
|
|
22 bool startsExternCBlock(const AnnotatedLine &Line) {
|
|
23 const FormatToken *Next = Line.First->getNextNonComment();
|
|
24 const FormatToken *NextNext = Next ? Next->getNextNonComment() : nullptr;
|
|
25 return Line.startsWith(tok::kw_extern) && Next && Next->isStringLiteral() &&
|
|
26 NextNext && NextNext->is(tok::l_brace);
|
|
27 }
|
|
28
|
|
29 /// Tracks the indent level of \c AnnotatedLines across levels.
|
|
30 ///
|
|
31 /// \c nextLine must be called for each \c AnnotatedLine, after which \c
|
|
32 /// getIndent() will return the indent for the last line \c nextLine was called
|
|
33 /// with.
|
|
34 /// If the line is not formatted (and thus the indent does not change), calling
|
|
35 /// \c adjustToUnmodifiedLine after the call to \c nextLine will cause
|
|
36 /// subsequent lines on the same level to be indented at the same level as the
|
|
37 /// given line.
|
|
38 class LevelIndentTracker {
|
|
39 public:
|
|
40 LevelIndentTracker(const FormatStyle &Style,
|
|
41 const AdditionalKeywords &Keywords, unsigned StartLevel,
|
|
42 int AdditionalIndent)
|
|
43 : Style(Style), Keywords(Keywords), AdditionalIndent(AdditionalIndent) {
|
|
44 for (unsigned i = 0; i != StartLevel; ++i)
|
|
45 IndentForLevel.push_back(Style.IndentWidth * i + AdditionalIndent);
|
|
46 }
|
|
47
|
|
48 /// Returns the indent for the current line.
|
|
49 unsigned getIndent() const { return Indent; }
|
|
50
|
|
51 /// Update the indent state given that \p Line is going to be formatted
|
|
52 /// next.
|
|
53 void nextLine(const AnnotatedLine &Line) {
|
|
54 Offset = getIndentOffset(*Line.First);
|
|
55 // Update the indent level cache size so that we can rely on it
|
|
56 // having the right size in adjustToUnmodifiedline.
|
|
57 while (IndentForLevel.size() <= Line.Level)
|
|
58 IndentForLevel.push_back(-1);
|
|
59 if (Line.InPPDirective) {
|
|
60 Indent = Line.Level * Style.IndentWidth + AdditionalIndent;
|
|
61 } else {
|
|
62 IndentForLevel.resize(Line.Level + 1);
|
|
63 Indent = getIndent(IndentForLevel, Line.Level);
|
|
64 }
|
|
65 if (static_cast<int>(Indent) + Offset >= 0)
|
|
66 Indent += Offset;
|
173
|
67 if (Line.First->is(TT_CSharpGenericTypeConstraint))
|
|
68 Indent = Line.Level * Style.IndentWidth + Style.ContinuationIndentWidth;
|
150
|
69 }
|
|
70
|
|
71 /// Update the indent state given that \p Line indent should be
|
|
72 /// skipped.
|
|
73 void skipLine(const AnnotatedLine &Line) {
|
|
74 while (IndentForLevel.size() <= Line.Level)
|
|
75 IndentForLevel.push_back(Indent);
|
|
76 }
|
|
77
|
|
78 /// Update the level indent to adapt to the given \p Line.
|
|
79 ///
|
|
80 /// When a line is not formatted, we move the subsequent lines on the same
|
|
81 /// level to the same indent.
|
|
82 /// Note that \c nextLine must have been called before this method.
|
|
83 void adjustToUnmodifiedLine(const AnnotatedLine &Line) {
|
|
84 unsigned LevelIndent = Line.First->OriginalColumn;
|
|
85 if (static_cast<int>(LevelIndent) - Offset >= 0)
|
|
86 LevelIndent -= Offset;
|
|
87 if ((!Line.First->is(tok::comment) || IndentForLevel[Line.Level] == -1) &&
|
|
88 !Line.InPPDirective)
|
|
89 IndentForLevel[Line.Level] = LevelIndent;
|
|
90 }
|
|
91
|
|
92 private:
|
|
93 /// Get the offset of the line relatively to the level.
|
|
94 ///
|
|
95 /// For example, 'public:' labels in classes are offset by 1 or 2
|
|
96 /// characters to the left from their level.
|
|
97 int getIndentOffset(const FormatToken &RootToken) {
|
|
98 if (Style.Language == FormatStyle::LK_Java ||
|
|
99 Style.Language == FormatStyle::LK_JavaScript || Style.isCSharp())
|
|
100 return 0;
|
|
101 if (RootToken.isAccessSpecifier(false) ||
|
|
102 RootToken.isObjCAccessSpecifier() ||
|
|
103 (RootToken.isOneOf(Keywords.kw_signals, Keywords.kw_qsignals) &&
|
|
104 RootToken.Next && RootToken.Next->is(tok::colon)))
|
|
105 return Style.AccessModifierOffset;
|
|
106 return 0;
|
|
107 }
|
|
108
|
|
109 /// Get the indent of \p Level from \p IndentForLevel.
|
|
110 ///
|
|
111 /// \p IndentForLevel must contain the indent for the level \c l
|
|
112 /// at \p IndentForLevel[l], or a value < 0 if the indent for
|
|
113 /// that level is unknown.
|
|
114 unsigned getIndent(ArrayRef<int> IndentForLevel, unsigned Level) {
|
|
115 if (IndentForLevel[Level] != -1)
|
|
116 return IndentForLevel[Level];
|
|
117 if (Level == 0)
|
|
118 return 0;
|
|
119 return getIndent(IndentForLevel, Level - 1) + Style.IndentWidth;
|
|
120 }
|
|
121
|
|
122 const FormatStyle &Style;
|
|
123 const AdditionalKeywords &Keywords;
|
|
124 const unsigned AdditionalIndent;
|
|
125
|
|
126 /// The indent in characters for each level.
|
|
127 std::vector<int> IndentForLevel;
|
|
128
|
|
129 /// Offset of the current line relative to the indent level.
|
|
130 ///
|
|
131 /// For example, the 'public' keywords is often indented with a negative
|
|
132 /// offset.
|
|
133 int Offset = 0;
|
|
134
|
|
135 /// The current line's indent.
|
|
136 unsigned Indent = 0;
|
|
137 };
|
|
138
|
|
139 const FormatToken *getMatchingNamespaceToken(
|
|
140 const AnnotatedLine *Line,
|
|
141 const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
|
|
142 if (!Line->startsWith(tok::r_brace))
|
|
143 return nullptr;
|
|
144 size_t StartLineIndex = Line->MatchingOpeningBlockLineIndex;
|
|
145 if (StartLineIndex == UnwrappedLine::kInvalidIndex)
|
|
146 return nullptr;
|
|
147 assert(StartLineIndex < AnnotatedLines.size());
|
|
148 return AnnotatedLines[StartLineIndex]->First->getNamespaceToken();
|
|
149 }
|
|
150
|
|
151 StringRef getNamespaceTokenText(const AnnotatedLine *Line) {
|
|
152 const FormatToken *NamespaceToken = Line->First->getNamespaceToken();
|
|
153 return NamespaceToken ? NamespaceToken->TokenText : StringRef();
|
|
154 }
|
|
155
|
|
156 StringRef getMatchingNamespaceTokenText(
|
|
157 const AnnotatedLine *Line,
|
|
158 const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
|
|
159 const FormatToken *NamespaceToken =
|
|
160 getMatchingNamespaceToken(Line, AnnotatedLines);
|
|
161 return NamespaceToken ? NamespaceToken->TokenText : StringRef();
|
|
162 }
|
|
163
|
|
164 class LineJoiner {
|
|
165 public:
|
|
166 LineJoiner(const FormatStyle &Style, const AdditionalKeywords &Keywords,
|
|
167 const SmallVectorImpl<AnnotatedLine *> &Lines)
|
|
168 : Style(Style), Keywords(Keywords), End(Lines.end()), Next(Lines.begin()),
|
|
169 AnnotatedLines(Lines) {}
|
|
170
|
|
171 /// Returns the next line, merging multiple lines into one if possible.
|
|
172 const AnnotatedLine *getNextMergedLine(bool DryRun,
|
|
173 LevelIndentTracker &IndentTracker) {
|
|
174 if (Next == End)
|
|
175 return nullptr;
|
|
176 const AnnotatedLine *Current = *Next;
|
|
177 IndentTracker.nextLine(*Current);
|
|
178 unsigned MergedLines = tryFitMultipleLinesInOne(IndentTracker, Next, End);
|
|
179 if (MergedLines > 0 && Style.ColumnLimit == 0)
|
|
180 // Disallow line merging if there is a break at the start of one of the
|
|
181 // input lines.
|
|
182 for (unsigned i = 0; i < MergedLines; ++i)
|
|
183 if (Next[i + 1]->First->NewlinesBefore > 0)
|
|
184 MergedLines = 0;
|
|
185 if (!DryRun)
|
|
186 for (unsigned i = 0; i < MergedLines; ++i)
|
|
187 join(*Next[0], *Next[i + 1]);
|
|
188 Next = Next + MergedLines + 1;
|
|
189 return Current;
|
|
190 }
|
|
191
|
|
192 private:
|
|
193 /// Calculates how many lines can be merged into 1 starting at \p I.
|
|
194 unsigned
|
|
195 tryFitMultipleLinesInOne(LevelIndentTracker &IndentTracker,
|
|
196 SmallVectorImpl<AnnotatedLine *>::const_iterator I,
|
|
197 SmallVectorImpl<AnnotatedLine *>::const_iterator E) {
|
|
198 const unsigned Indent = IndentTracker.getIndent();
|
|
199
|
|
200 // Can't join the last line with anything.
|
|
201 if (I + 1 == E)
|
|
202 return 0;
|
|
203 // We can never merge stuff if there are trailing line comments.
|
|
204 const AnnotatedLine *TheLine = *I;
|
|
205 if (TheLine->Last->is(TT_LineComment))
|
|
206 return 0;
|
|
207 if (I[1]->Type == LT_Invalid || I[1]->First->MustBreakBefore)
|
|
208 return 0;
|
|
209 if (TheLine->InPPDirective &&
|
|
210 (!I[1]->InPPDirective || I[1]->First->HasUnescapedNewline))
|
|
211 return 0;
|
|
212
|
|
213 if (Style.ColumnLimit > 0 && Indent > Style.ColumnLimit)
|
|
214 return 0;
|
|
215
|
|
216 unsigned Limit =
|
|
217 Style.ColumnLimit == 0 ? UINT_MAX : Style.ColumnLimit - Indent;
|
|
218 // If we already exceed the column limit, we set 'Limit' to 0. The different
|
|
219 // tryMerge..() functions can then decide whether to still do merging.
|
|
220 Limit = TheLine->Last->TotalLength > Limit
|
|
221 ? 0
|
|
222 : Limit - TheLine->Last->TotalLength;
|
|
223
|
|
224 if (TheLine->Last->is(TT_FunctionLBrace) &&
|
|
225 TheLine->First == TheLine->Last &&
|
|
226 !Style.BraceWrapping.SplitEmptyFunction &&
|
|
227 I[1]->First->is(tok::r_brace))
|
|
228 return tryMergeSimpleBlock(I, E, Limit);
|
|
229
|
|
230 // Handle empty record blocks where the brace has already been wrapped
|
|
231 if (TheLine->Last->is(tok::l_brace) && TheLine->First == TheLine->Last &&
|
|
232 I != AnnotatedLines.begin()) {
|
|
233 bool EmptyBlock = I[1]->First->is(tok::r_brace);
|
|
234
|
|
235 const FormatToken *Tok = I[-1]->First;
|
|
236 if (Tok && Tok->is(tok::comment))
|
|
237 Tok = Tok->getNextNonComment();
|
|
238
|
|
239 if (Tok && Tok->getNamespaceToken())
|
|
240 return !Style.BraceWrapping.SplitEmptyNamespace && EmptyBlock
|
|
241 ? tryMergeSimpleBlock(I, E, Limit)
|
|
242 : 0;
|
|
243
|
|
244 if (Tok && Tok->is(tok::kw_typedef))
|
|
245 Tok = Tok->getNextNonComment();
|
|
246 if (Tok && Tok->isOneOf(tok::kw_class, tok::kw_struct, tok::kw_union,
|
|
247 tok::kw_extern, Keywords.kw_interface))
|
|
248 return !Style.BraceWrapping.SplitEmptyRecord && EmptyBlock
|
|
249 ? tryMergeSimpleBlock(I, E, Limit)
|
|
250 : 0;
|
|
251 }
|
|
252
|
|
253 // FIXME: TheLine->Level != 0 might or might not be the right check to do.
|
|
254 // If necessary, change to something smarter.
|
|
255 bool MergeShortFunctions =
|
|
256 Style.AllowShortFunctionsOnASingleLine == FormatStyle::SFS_All ||
|
|
257 (Style.AllowShortFunctionsOnASingleLine >= FormatStyle::SFS_Empty &&
|
|
258 I[1]->First->is(tok::r_brace)) ||
|
|
259 (Style.AllowShortFunctionsOnASingleLine & FormatStyle::SFS_InlineOnly &&
|
|
260 TheLine->Level != 0);
|
|
261
|
|
262 if (Style.CompactNamespaces) {
|
|
263 if (auto nsToken = TheLine->First->getNamespaceToken()) {
|
|
264 int i = 0;
|
|
265 unsigned closingLine = TheLine->MatchingClosingBlockLineIndex - 1;
|
|
266 for (; I + 1 + i != E &&
|
|
267 nsToken->TokenText == getNamespaceTokenText(I[i + 1]) &&
|
|
268 closingLine == I[i + 1]->MatchingClosingBlockLineIndex &&
|
|
269 I[i + 1]->Last->TotalLength < Limit;
|
|
270 i++, closingLine--) {
|
|
271 // No extra indent for compacted namespaces
|
|
272 IndentTracker.skipLine(*I[i + 1]);
|
|
273
|
|
274 Limit -= I[i + 1]->Last->TotalLength;
|
|
275 }
|
|
276 return i;
|
|
277 }
|
|
278
|
|
279 if (auto nsToken = getMatchingNamespaceToken(TheLine, AnnotatedLines)) {
|
|
280 int i = 0;
|
|
281 unsigned openingLine = TheLine->MatchingOpeningBlockLineIndex - 1;
|
|
282 for (; I + 1 + i != E &&
|
|
283 nsToken->TokenText ==
|
|
284 getMatchingNamespaceTokenText(I[i + 1], AnnotatedLines) &&
|
|
285 openingLine == I[i + 1]->MatchingOpeningBlockLineIndex;
|
|
286 i++, openingLine--) {
|
|
287 // No space between consecutive braces
|
|
288 I[i + 1]->First->SpacesRequiredBefore = !I[i]->Last->is(tok::r_brace);
|
|
289
|
|
290 // Indent like the outer-most namespace
|
|
291 IndentTracker.nextLine(*I[i + 1]);
|
|
292 }
|
|
293 return i;
|
|
294 }
|
|
295 }
|
|
296
|
|
297 // Try to merge a function block with left brace unwrapped
|
|
298 if (TheLine->Last->is(TT_FunctionLBrace) &&
|
|
299 TheLine->First != TheLine->Last) {
|
|
300 return MergeShortFunctions ? tryMergeSimpleBlock(I, E, Limit) : 0;
|
|
301 }
|
|
302 // Try to merge a control statement block with left brace unwrapped
|
|
303 if (TheLine->Last->is(tok::l_brace) && TheLine->First != TheLine->Last &&
|
|
304 TheLine->First->isOneOf(tok::kw_if, tok::kw_while, tok::kw_for)) {
|
|
305 return Style.AllowShortBlocksOnASingleLine != FormatStyle::SBS_Never
|
|
306 ? tryMergeSimpleBlock(I, E, Limit)
|
|
307 : 0;
|
|
308 }
|
|
309 // Try to merge a control statement block with left brace wrapped
|
|
310 if (I[1]->First->is(tok::l_brace) &&
|
|
311 (TheLine->First->isOneOf(tok::kw_if, tok::kw_while, tok::kw_for,
|
|
312 tok::kw_switch, tok::kw_try, tok::kw_do) ||
|
|
313 (TheLine->First->is(tok::r_brace) && TheLine->First->Next &&
|
|
314 TheLine->First->Next->isOneOf(tok::kw_else, tok::kw_catch))) &&
|
|
315 Style.BraceWrapping.AfterControlStatement ==
|
|
316 FormatStyle::BWACS_MultiLine) {
|
|
317 // If possible, merge the next line's wrapped left brace with the current
|
|
318 // line. Otherwise, leave it on the next line, as this is a multi-line
|
|
319 // control statement.
|
|
320 return (Style.ColumnLimit == 0 ||
|
|
321 TheLine->Last->TotalLength <= Style.ColumnLimit)
|
|
322 ? 1
|
|
323 : 0;
|
|
324 } else if (I[1]->First->is(tok::l_brace) &&
|
|
325 TheLine->First->isOneOf(tok::kw_if, tok::kw_while,
|
|
326 tok::kw_for)) {
|
|
327 return (Style.BraceWrapping.AfterControlStatement ==
|
|
328 FormatStyle::BWACS_Always)
|
|
329 ? tryMergeSimpleBlock(I, E, Limit)
|
|
330 : 0;
|
|
331 } else if (I[1]->First->is(tok::l_brace) &&
|
|
332 TheLine->First->isOneOf(tok::kw_else, tok::kw_catch) &&
|
|
333 Style.BraceWrapping.AfterControlStatement ==
|
|
334 FormatStyle::BWACS_MultiLine) {
|
|
335 // This case if different from the upper BWACS_MultiLine processing
|
|
336 // in that a preceding r_brace is not on the same line as else/catch
|
|
337 // most likely because of BeforeElse/BeforeCatch set to true.
|
|
338 // If the line length doesn't fit ColumnLimit, leave l_brace on the
|
|
339 // next line to respect the BWACS_MultiLine.
|
|
340 return (Style.ColumnLimit == 0 ||
|
|
341 TheLine->Last->TotalLength <= Style.ColumnLimit)
|
|
342 ? 1
|
|
343 : 0;
|
|
344 }
|
|
345 // Try to merge either empty or one-line block if is precedeed by control
|
|
346 // statement token
|
|
347 if (TheLine->First->is(tok::l_brace) && TheLine->First == TheLine->Last &&
|
|
348 I != AnnotatedLines.begin() &&
|
|
349 I[-1]->First->isOneOf(tok::kw_if, tok::kw_while, tok::kw_for)) {
|
|
350 unsigned MergedLines = 0;
|
|
351 if (Style.AllowShortBlocksOnASingleLine != FormatStyle::SBS_Never) {
|
|
352 MergedLines = tryMergeSimpleBlock(I - 1, E, Limit);
|
|
353 // If we managed to merge the block, discard the first merged line
|
|
354 // since we are merging starting from I.
|
|
355 if (MergedLines > 0)
|
|
356 --MergedLines;
|
|
357 }
|
|
358 return MergedLines;
|
|
359 }
|
|
360 // Don't merge block with left brace wrapped after ObjC special blocks
|
|
361 if (TheLine->First->is(tok::l_brace) && I != AnnotatedLines.begin() &&
|
|
362 I[-1]->First->is(tok::at) && I[-1]->First->Next) {
|
|
363 tok::ObjCKeywordKind kwId = I[-1]->First->Next->Tok.getObjCKeywordID();
|
|
364 if (kwId == clang::tok::objc_autoreleasepool ||
|
|
365 kwId == clang::tok::objc_synchronized)
|
|
366 return 0;
|
|
367 }
|
|
368 // Don't merge block with left brace wrapped after case labels
|
|
369 if (TheLine->First->is(tok::l_brace) && I != AnnotatedLines.begin() &&
|
|
370 I[-1]->First->isOneOf(tok::kw_case, tok::kw_default))
|
|
371 return 0;
|
|
372 // Try to merge a block with left brace wrapped that wasn't yet covered
|
|
373 if (TheLine->Last->is(tok::l_brace)) {
|
|
374 return !Style.BraceWrapping.AfterFunction ||
|
|
375 (I[1]->First->is(tok::r_brace) &&
|
|
376 !Style.BraceWrapping.SplitEmptyRecord)
|
|
377 ? tryMergeSimpleBlock(I, E, Limit)
|
|
378 : 0;
|
|
379 }
|
|
380 // Try to merge a function block with left brace wrapped
|
|
381 if (I[1]->First->is(TT_FunctionLBrace) &&
|
|
382 Style.BraceWrapping.AfterFunction) {
|
|
383 if (I[1]->Last->is(TT_LineComment))
|
|
384 return 0;
|
|
385
|
|
386 // Check for Limit <= 2 to account for the " {".
|
|
387 if (Limit <= 2 || (Style.ColumnLimit == 0 && containsMustBreak(TheLine)))
|
|
388 return 0;
|
|
389 Limit -= 2;
|
|
390
|
|
391 unsigned MergedLines = 0;
|
|
392 if (MergeShortFunctions ||
|
|
393 (Style.AllowShortFunctionsOnASingleLine >= FormatStyle::SFS_Empty &&
|
|
394 I[1]->First == I[1]->Last && I + 2 != E &&
|
|
395 I[2]->First->is(tok::r_brace))) {
|
|
396 MergedLines = tryMergeSimpleBlock(I + 1, E, Limit);
|
|
397 // If we managed to merge the block, count the function header, which is
|
|
398 // on a separate line.
|
|
399 if (MergedLines > 0)
|
|
400 ++MergedLines;
|
|
401 }
|
|
402 return MergedLines;
|
|
403 }
|
|
404 if (TheLine->First->is(tok::kw_if)) {
|
|
405 return Style.AllowShortIfStatementsOnASingleLine
|
|
406 ? tryMergeSimpleControlStatement(I, E, Limit)
|
|
407 : 0;
|
|
408 }
|
173
|
409 if (TheLine->First->isOneOf(tok::kw_for, tok::kw_while, tok::kw_do)) {
|
150
|
410 return Style.AllowShortLoopsOnASingleLine
|
|
411 ? tryMergeSimpleControlStatement(I, E, Limit)
|
|
412 : 0;
|
|
413 }
|
|
414 if (TheLine->First->isOneOf(tok::kw_case, tok::kw_default)) {
|
|
415 return Style.AllowShortCaseLabelsOnASingleLine
|
|
416 ? tryMergeShortCaseLabels(I, E, Limit)
|
|
417 : 0;
|
|
418 }
|
|
419 if (TheLine->InPPDirective &&
|
|
420 (TheLine->First->HasUnescapedNewline || TheLine->First->IsFirst)) {
|
|
421 return tryMergeSimplePPDirective(I, E, Limit);
|
|
422 }
|
|
423 return 0;
|
|
424 }
|
|
425
|
|
426 unsigned
|
|
427 tryMergeSimplePPDirective(SmallVectorImpl<AnnotatedLine *>::const_iterator I,
|
|
428 SmallVectorImpl<AnnotatedLine *>::const_iterator E,
|
|
429 unsigned Limit) {
|
|
430 if (Limit == 0)
|
|
431 return 0;
|
|
432 if (I + 2 != E && I[2]->InPPDirective && !I[2]->First->HasUnescapedNewline)
|
|
433 return 0;
|
|
434 if (1 + I[1]->Last->TotalLength > Limit)
|
|
435 return 0;
|
|
436 return 1;
|
|
437 }
|
|
438
|
|
439 unsigned tryMergeSimpleControlStatement(
|
|
440 SmallVectorImpl<AnnotatedLine *>::const_iterator I,
|
|
441 SmallVectorImpl<AnnotatedLine *>::const_iterator E, unsigned Limit) {
|
|
442 if (Limit == 0)
|
|
443 return 0;
|
|
444 if (Style.BraceWrapping.AfterControlStatement ==
|
|
445 FormatStyle::BWACS_Always &&
|
|
446 I[1]->First->is(tok::l_brace) &&
|
|
447 Style.AllowShortBlocksOnASingleLine == FormatStyle::SBS_Never)
|
|
448 return 0;
|
|
449 if (I[1]->InPPDirective != (*I)->InPPDirective ||
|
|
450 (I[1]->InPPDirective && I[1]->First->HasUnescapedNewline))
|
|
451 return 0;
|
|
452 Limit = limitConsideringMacros(I + 1, E, Limit);
|
|
453 AnnotatedLine &Line = **I;
|
173
|
454 if (!Line.First->is(tok::kw_do) && Line.Last->isNot(tok::r_paren))
|
|
455 return 0;
|
|
456 // Only merge do while if do is the only statement on the line.
|
|
457 if (Line.First->is(tok::kw_do) && !Line.Last->is(tok::kw_do))
|
150
|
458 return 0;
|
|
459 if (1 + I[1]->Last->TotalLength > Limit)
|
|
460 return 0;
|
|
461 if (I[1]->First->isOneOf(tok::semi, tok::kw_if, tok::kw_for, tok::kw_while,
|
|
462 TT_LineComment))
|
|
463 return 0;
|
|
464 // Only inline simple if's (no nested if or else), unless specified
|
|
465 if (Style.AllowShortIfStatementsOnASingleLine != FormatStyle::SIS_Always) {
|
|
466 if (I + 2 != E && Line.startsWith(tok::kw_if) &&
|
|
467 I[2]->First->is(tok::kw_else))
|
|
468 return 0;
|
|
469 }
|
|
470 return 1;
|
|
471 }
|
|
472
|
|
473 unsigned
|
|
474 tryMergeShortCaseLabels(SmallVectorImpl<AnnotatedLine *>::const_iterator I,
|
|
475 SmallVectorImpl<AnnotatedLine *>::const_iterator E,
|
|
476 unsigned Limit) {
|
|
477 if (Limit == 0 || I + 1 == E ||
|
|
478 I[1]->First->isOneOf(tok::kw_case, tok::kw_default))
|
|
479 return 0;
|
|
480 if (I[0]->Last->is(tok::l_brace) || I[1]->First->is(tok::l_brace))
|
|
481 return 0;
|
|
482 unsigned NumStmts = 0;
|
|
483 unsigned Length = 0;
|
|
484 bool EndsWithComment = false;
|
|
485 bool InPPDirective = I[0]->InPPDirective;
|
|
486 const unsigned Level = I[0]->Level;
|
|
487 for (; NumStmts < 3; ++NumStmts) {
|
|
488 if (I + 1 + NumStmts == E)
|
|
489 break;
|
|
490 const AnnotatedLine *Line = I[1 + NumStmts];
|
|
491 if (Line->InPPDirective != InPPDirective)
|
|
492 break;
|
|
493 if (Line->First->isOneOf(tok::kw_case, tok::kw_default, tok::r_brace))
|
|
494 break;
|
|
495 if (Line->First->isOneOf(tok::kw_if, tok::kw_for, tok::kw_switch,
|
|
496 tok::kw_while) ||
|
|
497 EndsWithComment)
|
|
498 return 0;
|
|
499 if (Line->First->is(tok::comment)) {
|
|
500 if (Level != Line->Level)
|
|
501 return 0;
|
|
502 SmallVectorImpl<AnnotatedLine *>::const_iterator J = I + 2 + NumStmts;
|
|
503 for (; J != E; ++J) {
|
|
504 Line = *J;
|
|
505 if (Line->InPPDirective != InPPDirective)
|
|
506 break;
|
|
507 if (Line->First->isOneOf(tok::kw_case, tok::kw_default, tok::r_brace))
|
|
508 break;
|
|
509 if (Line->First->isNot(tok::comment) || Level != Line->Level)
|
|
510 return 0;
|
|
511 }
|
|
512 break;
|
|
513 }
|
|
514 if (Line->Last->is(tok::comment))
|
|
515 EndsWithComment = true;
|
|
516 Length += I[1 + NumStmts]->Last->TotalLength + 1; // 1 for the space.
|
|
517 }
|
|
518 if (NumStmts == 0 || NumStmts == 3 || Length > Limit)
|
|
519 return 0;
|
|
520 return NumStmts;
|
|
521 }
|
|
522
|
|
523 unsigned
|
|
524 tryMergeSimpleBlock(SmallVectorImpl<AnnotatedLine *>::const_iterator I,
|
|
525 SmallVectorImpl<AnnotatedLine *>::const_iterator E,
|
|
526 unsigned Limit) {
|
|
527 AnnotatedLine &Line = **I;
|
|
528
|
|
529 // Don't merge ObjC @ keywords and methods.
|
|
530 // FIXME: If an option to allow short exception handling clauses on a single
|
|
531 // line is added, change this to not return for @try and friends.
|
|
532 if (Style.Language != FormatStyle::LK_Java &&
|
|
533 Line.First->isOneOf(tok::at, tok::minus, tok::plus))
|
|
534 return 0;
|
|
535
|
|
536 // Check that the current line allows merging. This depends on whether we
|
|
537 // are in a control flow statements as well as several style flags.
|
|
538 if (Line.First->isOneOf(tok::kw_else, tok::kw_case) ||
|
|
539 (Line.First->Next && Line.First->Next->is(tok::kw_else)))
|
|
540 return 0;
|
|
541 // default: in switch statement
|
|
542 if (Line.First->is(tok::kw_default)) {
|
|
543 const FormatToken *Tok = Line.First->getNextNonComment();
|
|
544 if (Tok && Tok->is(tok::colon))
|
|
545 return 0;
|
|
546 }
|
|
547 if (Line.First->isOneOf(tok::kw_if, tok::kw_while, tok::kw_do, tok::kw_try,
|
|
548 tok::kw___try, tok::kw_catch, tok::kw___finally,
|
|
549 tok::kw_for, tok::r_brace, Keywords.kw___except)) {
|
|
550 if (Style.AllowShortBlocksOnASingleLine == FormatStyle::SBS_Never)
|
|
551 return 0;
|
|
552 // Don't merge when we can't except the case when
|
|
553 // the control statement block is empty
|
|
554 if (!Style.AllowShortIfStatementsOnASingleLine &&
|
|
555 Line.startsWith(tok::kw_if) &&
|
|
556 !Style.BraceWrapping.AfterControlStatement &&
|
|
557 !I[1]->First->is(tok::r_brace))
|
|
558 return 0;
|
|
559 if (!Style.AllowShortIfStatementsOnASingleLine &&
|
|
560 Line.startsWith(tok::kw_if) &&
|
|
561 Style.BraceWrapping.AfterControlStatement ==
|
|
562 FormatStyle::BWACS_Always &&
|
|
563 I + 2 != E && !I[2]->First->is(tok::r_brace))
|
|
564 return 0;
|
|
565 if (!Style.AllowShortLoopsOnASingleLine &&
|
|
566 Line.First->isOneOf(tok::kw_while, tok::kw_do, tok::kw_for) &&
|
|
567 !Style.BraceWrapping.AfterControlStatement &&
|
|
568 !I[1]->First->is(tok::r_brace))
|
|
569 return 0;
|
|
570 if (!Style.AllowShortLoopsOnASingleLine &&
|
|
571 Line.First->isOneOf(tok::kw_while, tok::kw_do, tok::kw_for) &&
|
|
572 Style.BraceWrapping.AfterControlStatement ==
|
|
573 FormatStyle::BWACS_Always &&
|
|
574 I + 2 != E && !I[2]->First->is(tok::r_brace))
|
|
575 return 0;
|
|
576 // FIXME: Consider an option to allow short exception handling clauses on
|
|
577 // a single line.
|
|
578 // FIXME: This isn't covered by tests.
|
|
579 // FIXME: For catch, __except, __finally the first token on the line
|
|
580 // is '}', so this isn't correct here.
|
|
581 if (Line.First->isOneOf(tok::kw_try, tok::kw___try, tok::kw_catch,
|
|
582 Keywords.kw___except, tok::kw___finally))
|
|
583 return 0;
|
|
584 }
|
|
585
|
|
586 if (Line.Last->is(tok::l_brace)) {
|
|
587 FormatToken *Tok = I[1]->First;
|
|
588 if (Tok->is(tok::r_brace) && !Tok->MustBreakBefore &&
|
|
589 (Tok->getNextNonComment() == nullptr ||
|
|
590 Tok->getNextNonComment()->is(tok::semi))) {
|
|
591 // We merge empty blocks even if the line exceeds the column limit.
|
|
592 Tok->SpacesRequiredBefore = Style.SpaceInEmptyBlock ? 1 : 0;
|
|
593 Tok->CanBreakBefore = true;
|
|
594 return 1;
|
|
595 } else if (Limit != 0 && !Line.startsWithNamespace() &&
|
|
596 !startsExternCBlock(Line)) {
|
|
597 // We don't merge short records.
|
|
598 FormatToken *RecordTok = Line.First;
|
|
599 // Skip record modifiers.
|
|
600 while (RecordTok->Next &&
|
|
601 RecordTok->isOneOf(
|
|
602 tok::kw_typedef, tok::kw_export, Keywords.kw_declare,
|
|
603 Keywords.kw_abstract, tok::kw_default, tok::kw_public,
|
|
604 tok::kw_private, tok::kw_protected, Keywords.kw_internal))
|
|
605 RecordTok = RecordTok->Next;
|
|
606 if (RecordTok &&
|
|
607 RecordTok->isOneOf(tok::kw_class, tok::kw_union, tok::kw_struct,
|
|
608 Keywords.kw_interface))
|
|
609 return 0;
|
|
610
|
|
611 // Check that we still have three lines and they fit into the limit.
|
|
612 if (I + 2 == E || I[2]->Type == LT_Invalid)
|
|
613 return 0;
|
|
614 Limit = limitConsideringMacros(I + 2, E, Limit);
|
|
615
|
|
616 if (!nextTwoLinesFitInto(I, Limit))
|
|
617 return 0;
|
|
618
|
|
619 // Second, check that the next line does not contain any braces - if it
|
|
620 // does, readability declines when putting it into a single line.
|
|
621 if (I[1]->Last->is(TT_LineComment))
|
|
622 return 0;
|
|
623 do {
|
|
624 if (Tok->is(tok::l_brace) && Tok->BlockKind != BK_BracedInit)
|
|
625 return 0;
|
|
626 Tok = Tok->Next;
|
|
627 } while (Tok);
|
|
628
|
|
629 // Last, check that the third line starts with a closing brace.
|
|
630 Tok = I[2]->First;
|
|
631 if (Tok->isNot(tok::r_brace))
|
|
632 return 0;
|
|
633
|
|
634 // Don't merge "if (a) { .. } else {".
|
|
635 if (Tok->Next && Tok->Next->is(tok::kw_else))
|
|
636 return 0;
|
|
637
|
|
638 // Don't merge a trailing multi-line control statement block like:
|
|
639 // } else if (foo &&
|
|
640 // bar)
|
|
641 // { <-- current Line
|
|
642 // baz();
|
|
643 // }
|
|
644 if (Line.First == Line.Last &&
|
|
645 Style.BraceWrapping.AfterControlStatement ==
|
|
646 FormatStyle::BWACS_MultiLine)
|
|
647 return 0;
|
|
648
|
|
649 return 2;
|
|
650 }
|
|
651 } else if (I[1]->First->is(tok::l_brace)) {
|
|
652 if (I[1]->Last->is(TT_LineComment))
|
|
653 return 0;
|
|
654
|
|
655 // Check for Limit <= 2 to account for the " {".
|
|
656 if (Limit <= 2 || (Style.ColumnLimit == 0 && containsMustBreak(*I)))
|
|
657 return 0;
|
|
658 Limit -= 2;
|
|
659 unsigned MergedLines = 0;
|
|
660 if (Style.AllowShortBlocksOnASingleLine != FormatStyle::SBS_Never ||
|
|
661 (I[1]->First == I[1]->Last && I + 2 != E &&
|
|
662 I[2]->First->is(tok::r_brace))) {
|
|
663 MergedLines = tryMergeSimpleBlock(I + 1, E, Limit);
|
|
664 // If we managed to merge the block, count the statement header, which
|
|
665 // is on a separate line.
|
|
666 if (MergedLines > 0)
|
|
667 ++MergedLines;
|
|
668 }
|
|
669 return MergedLines;
|
|
670 }
|
|
671 return 0;
|
|
672 }
|
|
673
|
|
674 /// Returns the modified column limit for \p I if it is inside a macro and
|
|
675 /// needs a trailing '\'.
|
|
676 unsigned
|
|
677 limitConsideringMacros(SmallVectorImpl<AnnotatedLine *>::const_iterator I,
|
|
678 SmallVectorImpl<AnnotatedLine *>::const_iterator E,
|
|
679 unsigned Limit) {
|
|
680 if (I[0]->InPPDirective && I + 1 != E &&
|
|
681 !I[1]->First->HasUnescapedNewline && !I[1]->First->is(tok::eof)) {
|
|
682 return Limit < 2 ? 0 : Limit - 2;
|
|
683 }
|
|
684 return Limit;
|
|
685 }
|
|
686
|
|
687 bool nextTwoLinesFitInto(SmallVectorImpl<AnnotatedLine *>::const_iterator I,
|
|
688 unsigned Limit) {
|
|
689 if (I[1]->First->MustBreakBefore || I[2]->First->MustBreakBefore)
|
|
690 return false;
|
|
691 return 1 + I[1]->Last->TotalLength + 1 + I[2]->Last->TotalLength <= Limit;
|
|
692 }
|
|
693
|
|
694 bool containsMustBreak(const AnnotatedLine *Line) {
|
|
695 for (const FormatToken *Tok = Line->First; Tok; Tok = Tok->Next) {
|
|
696 if (Tok->MustBreakBefore)
|
|
697 return true;
|
|
698 }
|
|
699 return false;
|
|
700 }
|
|
701
|
|
702 void join(AnnotatedLine &A, const AnnotatedLine &B) {
|
|
703 assert(!A.Last->Next);
|
|
704 assert(!B.First->Previous);
|
|
705 if (B.Affected)
|
|
706 A.Affected = true;
|
|
707 A.Last->Next = B.First;
|
|
708 B.First->Previous = A.Last;
|
|
709 B.First->CanBreakBefore = true;
|
|
710 unsigned LengthA = A.Last->TotalLength + B.First->SpacesRequiredBefore;
|
|
711 for (FormatToken *Tok = B.First; Tok; Tok = Tok->Next) {
|
|
712 Tok->TotalLength += LengthA;
|
|
713 A.Last = Tok;
|
|
714 }
|
|
715 }
|
|
716
|
|
717 const FormatStyle &Style;
|
|
718 const AdditionalKeywords &Keywords;
|
|
719 const SmallVectorImpl<AnnotatedLine *>::const_iterator End;
|
|
720
|
|
721 SmallVectorImpl<AnnotatedLine *>::const_iterator Next;
|
|
722 const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines;
|
|
723 };
|
|
724
|
|
725 static void markFinalized(FormatToken *Tok) {
|
|
726 for (; Tok; Tok = Tok->Next) {
|
|
727 Tok->Finalized = true;
|
|
728 for (AnnotatedLine *Child : Tok->Children)
|
|
729 markFinalized(Child->First);
|
|
730 }
|
|
731 }
|
|
732
|
|
733 #ifndef NDEBUG
|
|
734 static void printLineState(const LineState &State) {
|
|
735 llvm::dbgs() << "State: ";
|
|
736 for (const ParenState &P : State.Stack) {
|
|
737 llvm::dbgs() << (P.Tok ? P.Tok->TokenText : "F") << "|" << P.Indent << "|"
|
|
738 << P.LastSpace << "|" << P.NestedBlockIndent << " ";
|
|
739 }
|
|
740 llvm::dbgs() << State.NextToken->TokenText << "\n";
|
|
741 }
|
|
742 #endif
|
|
743
|
|
744 /// Base class for classes that format one \c AnnotatedLine.
|
|
745 class LineFormatter {
|
|
746 public:
|
|
747 LineFormatter(ContinuationIndenter *Indenter, WhitespaceManager *Whitespaces,
|
|
748 const FormatStyle &Style,
|
|
749 UnwrappedLineFormatter *BlockFormatter)
|
|
750 : Indenter(Indenter), Whitespaces(Whitespaces), Style(Style),
|
|
751 BlockFormatter(BlockFormatter) {}
|
|
752 virtual ~LineFormatter() {}
|
|
753
|
|
754 /// Formats an \c AnnotatedLine and returns the penalty.
|
|
755 ///
|
|
756 /// If \p DryRun is \c false, directly applies the changes.
|
|
757 virtual unsigned formatLine(const AnnotatedLine &Line, unsigned FirstIndent,
|
|
758 unsigned FirstStartColumn, bool DryRun) = 0;
|
|
759
|
|
760 protected:
|
|
761 /// If the \p State's next token is an r_brace closing a nested block,
|
|
762 /// format the nested block before it.
|
|
763 ///
|
|
764 /// Returns \c true if all children could be placed successfully and adapts
|
|
765 /// \p Penalty as well as \p State. If \p DryRun is false, also directly
|
|
766 /// creates changes using \c Whitespaces.
|
|
767 ///
|
|
768 /// The crucial idea here is that children always get formatted upon
|
|
769 /// encountering the closing brace right after the nested block. Now, if we
|
|
770 /// are currently trying to keep the "}" on the same line (i.e. \p NewLine is
|
|
771 /// \c false), the entire block has to be kept on the same line (which is only
|
|
772 /// possible if it fits on the line, only contains a single statement, etc.
|
|
773 ///
|
|
774 /// If \p NewLine is true, we format the nested block on separate lines, i.e.
|
|
775 /// break after the "{", format all lines with correct indentation and the put
|
|
776 /// the closing "}" on yet another new line.
|
|
777 ///
|
|
778 /// This enables us to keep the simple structure of the
|
|
779 /// \c UnwrappedLineFormatter, where we only have two options for each token:
|
|
780 /// break or don't break.
|
|
781 bool formatChildren(LineState &State, bool NewLine, bool DryRun,
|
|
782 unsigned &Penalty) {
|
|
783 const FormatToken *LBrace = State.NextToken->getPreviousNonComment();
|
|
784 FormatToken &Previous = *State.NextToken->Previous;
|
|
785 if (!LBrace || LBrace->isNot(tok::l_brace) ||
|
|
786 LBrace->BlockKind != BK_Block || Previous.Children.size() == 0)
|
|
787 // The previous token does not open a block. Nothing to do. We don't
|
|
788 // assert so that we can simply call this function for all tokens.
|
|
789 return true;
|
|
790
|
|
791 if (NewLine) {
|
|
792 int AdditionalIndent = State.Stack.back().Indent -
|
|
793 Previous.Children[0]->Level * Style.IndentWidth;
|
|
794
|
|
795 Penalty +=
|
|
796 BlockFormatter->format(Previous.Children, DryRun, AdditionalIndent,
|
|
797 /*FixBadIndentation=*/true);
|
|
798 return true;
|
|
799 }
|
|
800
|
|
801 if (Previous.Children[0]->First->MustBreakBefore)
|
|
802 return false;
|
|
803
|
|
804 // Cannot merge into one line if this line ends on a comment.
|
|
805 if (Previous.is(tok::comment))
|
|
806 return false;
|
|
807
|
|
808 // Cannot merge multiple statements into a single line.
|
|
809 if (Previous.Children.size() > 1)
|
|
810 return false;
|
|
811
|
|
812 const AnnotatedLine *Child = Previous.Children[0];
|
|
813 // We can't put the closing "}" on a line with a trailing comment.
|
|
814 if (Child->Last->isTrailingComment())
|
|
815 return false;
|
|
816
|
|
817 // If the child line exceeds the column limit, we wouldn't want to merge it.
|
|
818 // We add +2 for the trailing " }".
|
|
819 if (Style.ColumnLimit > 0 &&
|
|
820 Child->Last->TotalLength + State.Column + 2 > Style.ColumnLimit)
|
|
821 return false;
|
|
822
|
|
823 if (!DryRun) {
|
|
824 Whitespaces->replaceWhitespace(
|
|
825 *Child->First, /*Newlines=*/0, /*Spaces=*/1,
|
173
|
826 /*StartOfTokenColumn=*/State.Column, /*IsAligned=*/false,
|
|
827 State.Line->InPPDirective);
|
150
|
828 }
|
|
829 Penalty +=
|
|
830 formatLine(*Child, State.Column + 1, /*FirstStartColumn=*/0, DryRun);
|
|
831
|
|
832 State.Column += 1 + Child->Last->TotalLength;
|
|
833 return true;
|
|
834 }
|
|
835
|
|
836 ContinuationIndenter *Indenter;
|
|
837
|
|
838 private:
|
|
839 WhitespaceManager *Whitespaces;
|
|
840 const FormatStyle &Style;
|
|
841 UnwrappedLineFormatter *BlockFormatter;
|
|
842 };
|
|
843
|
|
844 /// Formatter that keeps the existing line breaks.
|
|
845 class NoColumnLimitLineFormatter : public LineFormatter {
|
|
846 public:
|
|
847 NoColumnLimitLineFormatter(ContinuationIndenter *Indenter,
|
|
848 WhitespaceManager *Whitespaces,
|
|
849 const FormatStyle &Style,
|
|
850 UnwrappedLineFormatter *BlockFormatter)
|
|
851 : LineFormatter(Indenter, Whitespaces, Style, BlockFormatter) {}
|
|
852
|
|
853 /// Formats the line, simply keeping all of the input's line breaking
|
|
854 /// decisions.
|
|
855 unsigned formatLine(const AnnotatedLine &Line, unsigned FirstIndent,
|
|
856 unsigned FirstStartColumn, bool DryRun) override {
|
|
857 assert(!DryRun);
|
|
858 LineState State = Indenter->getInitialState(FirstIndent, FirstStartColumn,
|
|
859 &Line, /*DryRun=*/false);
|
|
860 while (State.NextToken) {
|
|
861 bool Newline =
|
|
862 Indenter->mustBreak(State) ||
|
|
863 (Indenter->canBreak(State) && State.NextToken->NewlinesBefore > 0);
|
|
864 unsigned Penalty = 0;
|
|
865 formatChildren(State, Newline, /*DryRun=*/false, Penalty);
|
|
866 Indenter->addTokenToState(State, Newline, /*DryRun=*/false);
|
|
867 }
|
|
868 return 0;
|
|
869 }
|
|
870 };
|
|
871
|
|
872 /// Formatter that puts all tokens into a single line without breaks.
|
|
873 class NoLineBreakFormatter : public LineFormatter {
|
|
874 public:
|
|
875 NoLineBreakFormatter(ContinuationIndenter *Indenter,
|
|
876 WhitespaceManager *Whitespaces, const FormatStyle &Style,
|
|
877 UnwrappedLineFormatter *BlockFormatter)
|
|
878 : LineFormatter(Indenter, Whitespaces, Style, BlockFormatter) {}
|
|
879
|
|
880 /// Puts all tokens into a single line.
|
|
881 unsigned formatLine(const AnnotatedLine &Line, unsigned FirstIndent,
|
|
882 unsigned FirstStartColumn, bool DryRun) override {
|
|
883 unsigned Penalty = 0;
|
|
884 LineState State =
|
|
885 Indenter->getInitialState(FirstIndent, FirstStartColumn, &Line, DryRun);
|
|
886 while (State.NextToken) {
|
|
887 formatChildren(State, /*NewLine=*/false, DryRun, Penalty);
|
|
888 Indenter->addTokenToState(
|
|
889 State, /*Newline=*/State.NextToken->MustBreakBefore, DryRun);
|
|
890 }
|
|
891 return Penalty;
|
|
892 }
|
|
893 };
|
|
894
|
|
895 /// Finds the best way to break lines.
|
|
896 class OptimizingLineFormatter : public LineFormatter {
|
|
897 public:
|
|
898 OptimizingLineFormatter(ContinuationIndenter *Indenter,
|
|
899 WhitespaceManager *Whitespaces,
|
|
900 const FormatStyle &Style,
|
|
901 UnwrappedLineFormatter *BlockFormatter)
|
|
902 : LineFormatter(Indenter, Whitespaces, Style, BlockFormatter) {}
|
|
903
|
|
904 /// Formats the line by finding the best line breaks with line lengths
|
|
905 /// below the column limit.
|
|
906 unsigned formatLine(const AnnotatedLine &Line, unsigned FirstIndent,
|
|
907 unsigned FirstStartColumn, bool DryRun) override {
|
|
908 LineState State =
|
|
909 Indenter->getInitialState(FirstIndent, FirstStartColumn, &Line, DryRun);
|
|
910
|
|
911 // If the ObjC method declaration does not fit on a line, we should format
|
|
912 // it with one arg per line.
|
|
913 if (State.Line->Type == LT_ObjCMethodDecl)
|
|
914 State.Stack.back().BreakBeforeParameter = true;
|
|
915
|
|
916 // Find best solution in solution space.
|
|
917 return analyzeSolutionSpace(State, DryRun);
|
|
918 }
|
|
919
|
|
920 private:
|
|
921 struct CompareLineStatePointers {
|
|
922 bool operator()(LineState *obj1, LineState *obj2) const {
|
|
923 return *obj1 < *obj2;
|
|
924 }
|
|
925 };
|
|
926
|
|
927 /// A pair of <penalty, count> that is used to prioritize the BFS on.
|
|
928 ///
|
|
929 /// In case of equal penalties, we want to prefer states that were inserted
|
|
930 /// first. During state generation we make sure that we insert states first
|
|
931 /// that break the line as late as possible.
|
|
932 typedef std::pair<unsigned, unsigned> OrderedPenalty;
|
|
933
|
|
934 /// An edge in the solution space from \c Previous->State to \c State,
|
|
935 /// inserting a newline dependent on the \c NewLine.
|
|
936 struct StateNode {
|
|
937 StateNode(const LineState &State, bool NewLine, StateNode *Previous)
|
|
938 : State(State), NewLine(NewLine), Previous(Previous) {}
|
|
939 LineState State;
|
|
940 bool NewLine;
|
|
941 StateNode *Previous;
|
|
942 };
|
|
943
|
|
944 /// An item in the prioritized BFS search queue. The \c StateNode's
|
|
945 /// \c State has the given \c OrderedPenalty.
|
|
946 typedef std::pair<OrderedPenalty, StateNode *> QueueItem;
|
|
947
|
|
948 /// The BFS queue type.
|
|
949 typedef std::priority_queue<QueueItem, std::vector<QueueItem>,
|
|
950 std::greater<QueueItem>>
|
|
951 QueueType;
|
|
952
|
|
953 /// Analyze the entire solution space starting from \p InitialState.
|
|
954 ///
|
|
955 /// This implements a variant of Dijkstra's algorithm on the graph that spans
|
|
956 /// the solution space (\c LineStates are the nodes). The algorithm tries to
|
|
957 /// find the shortest path (the one with lowest penalty) from \p InitialState
|
|
958 /// to a state where all tokens are placed. Returns the penalty.
|
|
959 ///
|
|
960 /// If \p DryRun is \c false, directly applies the changes.
|
|
961 unsigned analyzeSolutionSpace(LineState &InitialState, bool DryRun) {
|
|
962 std::set<LineState *, CompareLineStatePointers> Seen;
|
|
963
|
|
964 // Increasing count of \c StateNode items we have created. This is used to
|
|
965 // create a deterministic order independent of the container.
|
|
966 unsigned Count = 0;
|
|
967 QueueType Queue;
|
|
968
|
|
969 // Insert start element into queue.
|
|
970 StateNode *Node =
|
|
971 new (Allocator.Allocate()) StateNode(InitialState, false, nullptr);
|
|
972 Queue.push(QueueItem(OrderedPenalty(0, Count), Node));
|
|
973 ++Count;
|
|
974
|
|
975 unsigned Penalty = 0;
|
|
976
|
|
977 // While not empty, take first element and follow edges.
|
|
978 while (!Queue.empty()) {
|
|
979 Penalty = Queue.top().first.first;
|
|
980 StateNode *Node = Queue.top().second;
|
|
981 if (!Node->State.NextToken) {
|
|
982 LLVM_DEBUG(llvm::dbgs()
|
|
983 << "\n---\nPenalty for line: " << Penalty << "\n");
|
|
984 break;
|
|
985 }
|
|
986 Queue.pop();
|
|
987
|
|
988 // Cut off the analysis of certain solutions if the analysis gets too
|
|
989 // complex. See description of IgnoreStackForComparison.
|
|
990 if (Count > 50000)
|
|
991 Node->State.IgnoreStackForComparison = true;
|
|
992
|
|
993 if (!Seen.insert(&Node->State).second)
|
|
994 // State already examined with lower penalty.
|
|
995 continue;
|
|
996
|
|
997 FormatDecision LastFormat = Node->State.NextToken->Decision;
|
|
998 if (LastFormat == FD_Unformatted || LastFormat == FD_Continue)
|
|
999 addNextStateToQueue(Penalty, Node, /*NewLine=*/false, &Count, &Queue);
|
|
1000 if (LastFormat == FD_Unformatted || LastFormat == FD_Break)
|
|
1001 addNextStateToQueue(Penalty, Node, /*NewLine=*/true, &Count, &Queue);
|
|
1002 }
|
|
1003
|
|
1004 if (Queue.empty()) {
|
|
1005 // We were unable to find a solution, do nothing.
|
|
1006 // FIXME: Add diagnostic?
|
|
1007 LLVM_DEBUG(llvm::dbgs() << "Could not find a solution.\n");
|
|
1008 return 0;
|
|
1009 }
|
|
1010
|
|
1011 // Reconstruct the solution.
|
|
1012 if (!DryRun)
|
|
1013 reconstructPath(InitialState, Queue.top().second);
|
|
1014
|
|
1015 LLVM_DEBUG(llvm::dbgs()
|
|
1016 << "Total number of analyzed states: " << Count << "\n");
|
|
1017 LLVM_DEBUG(llvm::dbgs() << "---\n");
|
|
1018
|
|
1019 return Penalty;
|
|
1020 }
|
|
1021
|
|
1022 /// Add the following state to the analysis queue \c Queue.
|
|
1023 ///
|
|
1024 /// Assume the current state is \p PreviousNode and has been reached with a
|
|
1025 /// penalty of \p Penalty. Insert a line break if \p NewLine is \c true.
|
|
1026 void addNextStateToQueue(unsigned Penalty, StateNode *PreviousNode,
|
|
1027 bool NewLine, unsigned *Count, QueueType *Queue) {
|
|
1028 if (NewLine && !Indenter->canBreak(PreviousNode->State))
|
|
1029 return;
|
|
1030 if (!NewLine && Indenter->mustBreak(PreviousNode->State))
|
|
1031 return;
|
|
1032
|
|
1033 StateNode *Node = new (Allocator.Allocate())
|
|
1034 StateNode(PreviousNode->State, NewLine, PreviousNode);
|
|
1035 if (!formatChildren(Node->State, NewLine, /*DryRun=*/true, Penalty))
|
|
1036 return;
|
|
1037
|
|
1038 Penalty += Indenter->addTokenToState(Node->State, NewLine, true);
|
|
1039
|
|
1040 Queue->push(QueueItem(OrderedPenalty(Penalty, *Count), Node));
|
|
1041 ++(*Count);
|
|
1042 }
|
|
1043
|
|
1044 /// Applies the best formatting by reconstructing the path in the
|
|
1045 /// solution space that leads to \c Best.
|
|
1046 void reconstructPath(LineState &State, StateNode *Best) {
|
|
1047 std::deque<StateNode *> Path;
|
|
1048 // We do not need a break before the initial token.
|
|
1049 while (Best->Previous) {
|
|
1050 Path.push_front(Best);
|
|
1051 Best = Best->Previous;
|
|
1052 }
|
|
1053 for (auto I = Path.begin(), E = Path.end(); I != E; ++I) {
|
|
1054 unsigned Penalty = 0;
|
|
1055 formatChildren(State, (*I)->NewLine, /*DryRun=*/false, Penalty);
|
|
1056 Penalty += Indenter->addTokenToState(State, (*I)->NewLine, false);
|
|
1057
|
|
1058 LLVM_DEBUG({
|
|
1059 printLineState((*I)->Previous->State);
|
|
1060 if ((*I)->NewLine) {
|
|
1061 llvm::dbgs() << "Penalty for placing "
|
|
1062 << (*I)->Previous->State.NextToken->Tok.getName()
|
|
1063 << " on a new line: " << Penalty << "\n";
|
|
1064 }
|
|
1065 });
|
|
1066 }
|
|
1067 }
|
|
1068
|
|
1069 llvm::SpecificBumpPtrAllocator<StateNode> Allocator;
|
|
1070 };
|
|
1071
|
|
1072 } // anonymous namespace
|
|
1073
|
|
1074 unsigned UnwrappedLineFormatter::format(
|
|
1075 const SmallVectorImpl<AnnotatedLine *> &Lines, bool DryRun,
|
|
1076 int AdditionalIndent, bool FixBadIndentation, unsigned FirstStartColumn,
|
|
1077 unsigned NextStartColumn, unsigned LastStartColumn) {
|
|
1078 LineJoiner Joiner(Style, Keywords, Lines);
|
|
1079
|
|
1080 // Try to look up already computed penalty in DryRun-mode.
|
|
1081 std::pair<const SmallVectorImpl<AnnotatedLine *> *, unsigned> CacheKey(
|
|
1082 &Lines, AdditionalIndent);
|
|
1083 auto CacheIt = PenaltyCache.find(CacheKey);
|
|
1084 if (DryRun && CacheIt != PenaltyCache.end())
|
|
1085 return CacheIt->second;
|
|
1086
|
|
1087 assert(!Lines.empty());
|
|
1088 unsigned Penalty = 0;
|
|
1089 LevelIndentTracker IndentTracker(Style, Keywords, Lines[0]->Level,
|
|
1090 AdditionalIndent);
|
|
1091 const AnnotatedLine *PreviousLine = nullptr;
|
|
1092 const AnnotatedLine *NextLine = nullptr;
|
|
1093
|
|
1094 // The minimum level of consecutive lines that have been formatted.
|
|
1095 unsigned RangeMinLevel = UINT_MAX;
|
|
1096
|
|
1097 bool FirstLine = true;
|
|
1098 for (const AnnotatedLine *Line =
|
|
1099 Joiner.getNextMergedLine(DryRun, IndentTracker);
|
|
1100 Line; Line = NextLine, FirstLine = false) {
|
|
1101 const AnnotatedLine &TheLine = *Line;
|
|
1102 unsigned Indent = IndentTracker.getIndent();
|
|
1103
|
|
1104 // We continue formatting unchanged lines to adjust their indent, e.g. if a
|
|
1105 // scope was added. However, we need to carefully stop doing this when we
|
|
1106 // exit the scope of affected lines to prevent indenting a the entire
|
|
1107 // remaining file if it currently missing a closing brace.
|
|
1108 bool PreviousRBrace =
|
|
1109 PreviousLine && PreviousLine->startsWith(tok::r_brace);
|
|
1110 bool ContinueFormatting =
|
|
1111 TheLine.Level > RangeMinLevel ||
|
|
1112 (TheLine.Level == RangeMinLevel && !PreviousRBrace &&
|
|
1113 !TheLine.startsWith(tok::r_brace));
|
|
1114
|
|
1115 bool FixIndentation = (FixBadIndentation || ContinueFormatting) &&
|
|
1116 Indent != TheLine.First->OriginalColumn;
|
|
1117 bool ShouldFormat = TheLine.Affected || FixIndentation;
|
|
1118 // We cannot format this line; if the reason is that the line had a
|
|
1119 // parsing error, remember that.
|
|
1120 if (ShouldFormat && TheLine.Type == LT_Invalid && Status) {
|
|
1121 Status->FormatComplete = false;
|
|
1122 Status->Line =
|
|
1123 SourceMgr.getSpellingLineNumber(TheLine.First->Tok.getLocation());
|
|
1124 }
|
|
1125
|
|
1126 if (ShouldFormat && TheLine.Type != LT_Invalid) {
|
|
1127 if (!DryRun) {
|
|
1128 bool LastLine = Line->First->is(tok::eof);
|
|
1129 formatFirstToken(TheLine, PreviousLine, Lines, Indent,
|
|
1130 LastLine ? LastStartColumn : NextStartColumn + Indent);
|
|
1131 }
|
|
1132
|
|
1133 NextLine = Joiner.getNextMergedLine(DryRun, IndentTracker);
|
|
1134 unsigned ColumnLimit = getColumnLimit(TheLine.InPPDirective, NextLine);
|
|
1135 bool FitsIntoOneLine =
|
|
1136 TheLine.Last->TotalLength + Indent <= ColumnLimit ||
|
|
1137 (TheLine.Type == LT_ImportStatement &&
|
|
1138 (Style.Language != FormatStyle::LK_JavaScript ||
|
|
1139 !Style.JavaScriptWrapImports)) ||
|
|
1140 (Style.isCSharp() &&
|
|
1141 TheLine.InPPDirective); // don't split #regions in C#
|
|
1142 if (Style.ColumnLimit == 0)
|
|
1143 NoColumnLimitLineFormatter(Indenter, Whitespaces, Style, this)
|
|
1144 .formatLine(TheLine, NextStartColumn + Indent,
|
|
1145 FirstLine ? FirstStartColumn : 0, DryRun);
|
|
1146 else if (FitsIntoOneLine)
|
|
1147 Penalty += NoLineBreakFormatter(Indenter, Whitespaces, Style, this)
|
|
1148 .formatLine(TheLine, NextStartColumn + Indent,
|
|
1149 FirstLine ? FirstStartColumn : 0, DryRun);
|
|
1150 else
|
|
1151 Penalty += OptimizingLineFormatter(Indenter, Whitespaces, Style, this)
|
|
1152 .formatLine(TheLine, NextStartColumn + Indent,
|
|
1153 FirstLine ? FirstStartColumn : 0, DryRun);
|
|
1154 RangeMinLevel = std::min(RangeMinLevel, TheLine.Level);
|
|
1155 } else {
|
|
1156 // If no token in the current line is affected, we still need to format
|
|
1157 // affected children.
|
|
1158 if (TheLine.ChildrenAffected)
|
|
1159 for (const FormatToken *Tok = TheLine.First; Tok; Tok = Tok->Next)
|
|
1160 if (!Tok->Children.empty())
|
|
1161 format(Tok->Children, DryRun);
|
|
1162
|
|
1163 // Adapt following lines on the current indent level to the same level
|
|
1164 // unless the current \c AnnotatedLine is not at the beginning of a line.
|
|
1165 bool StartsNewLine =
|
|
1166 TheLine.First->NewlinesBefore > 0 || TheLine.First->IsFirst;
|
|
1167 if (StartsNewLine)
|
|
1168 IndentTracker.adjustToUnmodifiedLine(TheLine);
|
|
1169 if (!DryRun) {
|
|
1170 bool ReformatLeadingWhitespace =
|
|
1171 StartsNewLine && ((PreviousLine && PreviousLine->Affected) ||
|
|
1172 TheLine.LeadingEmptyLinesAffected);
|
|
1173 // Format the first token.
|
|
1174 if (ReformatLeadingWhitespace)
|
|
1175 formatFirstToken(TheLine, PreviousLine, Lines,
|
|
1176 TheLine.First->OriginalColumn,
|
|
1177 TheLine.First->OriginalColumn);
|
|
1178 else
|
|
1179 Whitespaces->addUntouchableToken(*TheLine.First,
|
|
1180 TheLine.InPPDirective);
|
|
1181
|
|
1182 // Notify the WhitespaceManager about the unchanged whitespace.
|
|
1183 for (FormatToken *Tok = TheLine.First->Next; Tok; Tok = Tok->Next)
|
|
1184 Whitespaces->addUntouchableToken(*Tok, TheLine.InPPDirective);
|
|
1185 }
|
|
1186 NextLine = Joiner.getNextMergedLine(DryRun, IndentTracker);
|
|
1187 RangeMinLevel = UINT_MAX;
|
|
1188 }
|
|
1189 if (!DryRun)
|
|
1190 markFinalized(TheLine.First);
|
|
1191 PreviousLine = &TheLine;
|
|
1192 }
|
|
1193 PenaltyCache[CacheKey] = Penalty;
|
|
1194 return Penalty;
|
|
1195 }
|
|
1196
|
|
1197 void UnwrappedLineFormatter::formatFirstToken(
|
|
1198 const AnnotatedLine &Line, const AnnotatedLine *PreviousLine,
|
|
1199 const SmallVectorImpl<AnnotatedLine *> &Lines, unsigned Indent,
|
|
1200 unsigned NewlineIndent) {
|
|
1201 FormatToken &RootToken = *Line.First;
|
|
1202 if (RootToken.is(tok::eof)) {
|
|
1203 unsigned Newlines = std::min(RootToken.NewlinesBefore, 1u);
|
|
1204 unsigned TokenIndent = Newlines ? NewlineIndent : 0;
|
|
1205 Whitespaces->replaceWhitespace(RootToken, Newlines, TokenIndent,
|
|
1206 TokenIndent);
|
|
1207 return;
|
|
1208 }
|
|
1209 unsigned Newlines =
|
|
1210 std::min(RootToken.NewlinesBefore, Style.MaxEmptyLinesToKeep + 1);
|
|
1211 // Remove empty lines before "}" where applicable.
|
|
1212 if (RootToken.is(tok::r_brace) &&
|
|
1213 (!RootToken.Next ||
|
|
1214 (RootToken.Next->is(tok::semi) && !RootToken.Next->Next)) &&
|
|
1215 // Do not remove empty lines before namespace closing "}".
|
|
1216 !getNamespaceToken(&Line, Lines))
|
|
1217 Newlines = std::min(Newlines, 1u);
|
|
1218 // Remove empty lines at the start of nested blocks (lambdas/arrow functions)
|
|
1219 if (PreviousLine == nullptr && Line.Level > 0)
|
|
1220 Newlines = std::min(Newlines, 1u);
|
|
1221 if (Newlines == 0 && !RootToken.IsFirst)
|
|
1222 Newlines = 1;
|
|
1223 if (RootToken.IsFirst && !RootToken.HasUnescapedNewline)
|
|
1224 Newlines = 0;
|
|
1225
|
|
1226 // Remove empty lines after "{".
|
|
1227 if (!Style.KeepEmptyLinesAtTheStartOfBlocks && PreviousLine &&
|
|
1228 PreviousLine->Last->is(tok::l_brace) &&
|
|
1229 !PreviousLine->startsWithNamespace() &&
|
|
1230 !startsExternCBlock(*PreviousLine))
|
|
1231 Newlines = 1;
|
|
1232
|
|
1233 // Insert extra new line before access specifiers.
|
|
1234 if (PreviousLine && PreviousLine->Last->isOneOf(tok::semi, tok::r_brace) &&
|
|
1235 RootToken.isAccessSpecifier() && RootToken.NewlinesBefore == 1)
|
|
1236 ++Newlines;
|
|
1237
|
|
1238 // Remove empty lines after access specifiers.
|
|
1239 if (PreviousLine && PreviousLine->First->isAccessSpecifier() &&
|
|
1240 (!PreviousLine->InPPDirective || !RootToken.HasUnescapedNewline))
|
|
1241 Newlines = std::min(1u, Newlines);
|
|
1242
|
|
1243 if (Newlines)
|
|
1244 Indent = NewlineIndent;
|
|
1245
|
|
1246 // If in Whitemsmiths mode, indent start and end of blocks
|
|
1247 if (Style.BreakBeforeBraces == FormatStyle::BS_Whitesmiths) {
|
|
1248 if (RootToken.isOneOf(tok::l_brace, tok::r_brace, tok::kw_case))
|
|
1249 Indent += Style.IndentWidth;
|
|
1250 }
|
|
1251
|
|
1252 // Preprocessor directives get indented before the hash only if specified
|
|
1253 if (Style.IndentPPDirectives != FormatStyle::PPDIS_BeforeHash &&
|
|
1254 (Line.Type == LT_PreprocessorDirective ||
|
|
1255 Line.Type == LT_ImportStatement))
|
|
1256 Indent = 0;
|
|
1257
|
|
1258 Whitespaces->replaceWhitespace(RootToken, Newlines, Indent, Indent,
|
173
|
1259 /*IsAligned=*/false,
|
150
|
1260 Line.InPPDirective &&
|
|
1261 !RootToken.HasUnescapedNewline);
|
|
1262 }
|
|
1263
|
|
1264 unsigned
|
|
1265 UnwrappedLineFormatter::getColumnLimit(bool InPPDirective,
|
|
1266 const AnnotatedLine *NextLine) const {
|
|
1267 // In preprocessor directives reserve two chars for trailing " \" if the
|
|
1268 // next line continues the preprocessor directive.
|
|
1269 bool ContinuesPPDirective =
|
|
1270 InPPDirective &&
|
|
1271 // If there is no next line, this is likely a child line and the parent
|
|
1272 // continues the preprocessor directive.
|
|
1273 (!NextLine ||
|
|
1274 (NextLine->InPPDirective &&
|
|
1275 // If there is an unescaped newline between this line and the next, the
|
|
1276 // next line starts a new preprocessor directive.
|
|
1277 !NextLine->First->HasUnescapedNewline));
|
|
1278 return Style.ColumnLimit - (ContinuesPPDirective ? 2 : 0);
|
|
1279 }
|
|
1280
|
|
1281 } // namespace format
|
|
1282 } // namespace clang
|