1 | //===--- ContinuationIndenter.h - Format C++ code ---------------*- C++ -*-===// |
2 | // |
3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
4 | // See https://llvm.org/LICENSE.txt for license information. |
5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
6 | // |
7 | //===----------------------------------------------------------------------===// |
8 | /// |
9 | /// \file |
10 | /// This file implements an indenter that manages the indentation of |
11 | /// continuations. |
12 | /// |
13 | //===----------------------------------------------------------------------===// |
14 | |
15 | #ifndef LLVM_CLANG_LIB_FORMAT_CONTINUATIONINDENTER_H |
16 | #define LLVM_CLANG_LIB_FORMAT_CONTINUATIONINDENTER_H |
17 | |
18 | #include "Encoding.h" |
19 | #include "FormatToken.h" |
20 | #include "clang/Format/Format.h" |
21 | #include "llvm/Support/Regex.h" |
22 | #include <map> |
23 | #include <tuple> |
24 | |
25 | namespace clang { |
26 | class SourceManager; |
27 | |
28 | namespace format { |
29 | |
30 | class AnnotatedLine; |
31 | class BreakableToken; |
32 | struct FormatToken; |
33 | struct LineState; |
34 | struct ParenState; |
35 | struct RawStringFormatStyleManager; |
36 | class WhitespaceManager; |
37 | |
38 | struct RawStringFormatStyleManager { |
39 | llvm::StringMap<FormatStyle> DelimiterStyle; |
40 | llvm::StringMap<FormatStyle> EnclosingFunctionStyle; |
41 | |
42 | RawStringFormatStyleManager(const FormatStyle &CodeStyle); |
43 | |
44 | llvm::Optional<FormatStyle> getDelimiterStyle(StringRef Delimiter) const; |
45 | |
46 | llvm::Optional<FormatStyle> |
47 | getEnclosingFunctionStyle(StringRef EnclosingFunction) const; |
48 | }; |
49 | |
50 | class ContinuationIndenter { |
51 | public: |
52 | /// Constructs a \c ContinuationIndenter to format \p Line starting in |
53 | /// column \p FirstIndent. |
54 | ContinuationIndenter(const FormatStyle &Style, |
55 | const AdditionalKeywords &Keywords, |
56 | const SourceManager &SourceMgr, |
57 | WhitespaceManager &Whitespaces, |
58 | encoding::Encoding Encoding, |
59 | bool BinPackInconclusiveFunctions); |
60 | |
61 | /// Get the initial state, i.e. the state after placing \p Line's |
62 | /// first token at \p FirstIndent. When reformatting a fragment of code, as in |
63 | /// the case of formatting inside raw string literals, \p FirstStartColumn is |
64 | /// the column at which the state of the parent formatter is. |
65 | LineState getInitialState(unsigned FirstIndent, unsigned FirstStartColumn, |
66 | const AnnotatedLine *Line, bool DryRun); |
67 | |
68 | // FIXME: canBreak and mustBreak aren't strictly indentation-related. Find a |
69 | // better home. |
70 | /// Returns \c true, if a line break after \p State is allowed. |
71 | bool canBreak(const LineState &State); |
72 | |
73 | /// Returns \c true, if a line break after \p State is mandatory. |
74 | bool mustBreak(const LineState &State); |
75 | |
76 | /// Appends the next token to \p State and updates information |
77 | /// necessary for indentation. |
78 | /// |
79 | /// Puts the token on the current line if \p Newline is \c false and adds a |
80 | /// line break and necessary indentation otherwise. |
81 | /// |
82 | /// If \p DryRun is \c false, also creates and stores the required |
83 | /// \c Replacement. |
84 | unsigned addTokenToState(LineState &State, bool Newline, bool DryRun, |
85 | unsigned ExtraSpaces = 0); |
86 | |
87 | /// Get the column limit for this line. This is the style's column |
88 | /// limit, potentially reduced for preprocessor definitions. |
89 | unsigned getColumnLimit(const LineState &State) const; |
90 | |
91 | private: |
92 | /// Mark the next token as consumed in \p State and modify its stacks |
93 | /// accordingly. |
94 | unsigned moveStateToNextToken(LineState &State, bool DryRun, bool Newline); |
95 | |
96 | /// Update 'State' according to the next token's fake left parentheses. |
97 | void moveStatePastFakeLParens(LineState &State, bool Newline); |
98 | /// Update 'State' according to the next token's fake r_parens. |
99 | void moveStatePastFakeRParens(LineState &State); |
100 | |
101 | /// Update 'State' according to the next token being one of "(<{[". |
102 | void moveStatePastScopeOpener(LineState &State, bool Newline); |
103 | /// Update 'State' according to the next token being one of ")>}]". |
104 | void moveStatePastScopeCloser(LineState &State); |
105 | /// Update 'State' with the next token opening a nested block. |
106 | void moveStateToNewBlock(LineState &State); |
107 | |
108 | /// Reformats a raw string literal. |
109 | /// |
110 | /// \returns An extra penalty induced by reformatting the token. |
111 | unsigned reformatRawStringLiteral(const FormatToken &Current, |
112 | LineState &State, |
113 | const FormatStyle &RawStringStyle, |
114 | bool DryRun); |
115 | |
116 | /// If the current token is at the end of the current line, handle |
117 | /// the transition to the next line. |
118 | unsigned handleEndOfLine(const FormatToken &Current, LineState &State, |
119 | bool DryRun, bool AllowBreak); |
120 | |
121 | /// If \p Current is a raw string that is configured to be reformatted, |
122 | /// return the style to be used. |
123 | llvm::Optional<FormatStyle> getRawStringStyle(const FormatToken &Current, |
124 | const LineState &State); |
125 | |
126 | /// If the current token sticks out over the end of the line, break |
127 | /// it if possible. |
128 | /// |
129 | /// \returns A pair (penalty, exceeded), where penalty is the extra penalty |
130 | /// when tokens are broken or lines exceed the column limit, and exceeded |
131 | /// indicates whether the algorithm purposefully left lines exceeding the |
132 | /// column limit. |
133 | /// |
134 | /// The returned penalty will cover the cost of the additional line breaks |
135 | /// and column limit violation in all lines except for the last one. The |
136 | /// penalty for the column limit violation in the last line (and in single |
137 | /// line tokens) is handled in \c addNextStateToQueue. |
138 | /// |
139 | /// \p Strict indicates whether reflowing is allowed to leave characters |
140 | /// protruding the column limit; if true, lines will be split strictly within |
141 | /// the column limit where possible; if false, words are allowed to protrude |
142 | /// over the column limit as long as the penalty is less than the penalty |
143 | /// of a break. |
144 | std::pair<unsigned, bool> breakProtrudingToken(const FormatToken &Current, |
145 | LineState &State, |
146 | bool AllowBreak, bool DryRun, |
147 | bool Strict); |
148 | |
149 | /// Returns the \c BreakableToken starting at \p Current, or nullptr |
150 | /// if the current token cannot be broken. |
151 | std::unique_ptr<BreakableToken> |
152 | createBreakableToken(const FormatToken &Current, LineState &State, |
153 | bool AllowBreak); |
154 | |
155 | /// Appends the next token to \p State and updates information |
156 | /// necessary for indentation. |
157 | /// |
158 | /// Puts the token on the current line. |
159 | /// |
160 | /// If \p DryRun is \c false, also creates and stores the required |
161 | /// \c Replacement. |
162 | void addTokenOnCurrentLine(LineState &State, bool DryRun, |
163 | unsigned ExtraSpaces); |
164 | |
165 | /// Appends the next token to \p State and updates information |
166 | /// necessary for indentation. |
167 | /// |
168 | /// Adds a line break and necessary indentation. |
169 | /// |
170 | /// If \p DryRun is \c false, also creates and stores the required |
171 | /// \c Replacement. |
172 | unsigned addTokenOnNewLine(LineState &State, bool DryRun); |
173 | |
174 | /// Calculate the new column for a line wrap before the next token. |
175 | unsigned getNewLineColumn(const LineState &State); |
176 | |
177 | /// Adds a multiline token to the \p State. |
178 | /// |
179 | /// \returns Extra penalty for the first line of the literal: last line is |
180 | /// handled in \c addNextStateToQueue, and the penalty for other lines doesn't |
181 | /// matter, as we don't change them. |
182 | unsigned addMultilineToken(const FormatToken &Current, LineState &State); |
183 | |
184 | /// Returns \c true if the next token starts a multiline string |
185 | /// literal. |
186 | /// |
187 | /// This includes implicitly concatenated strings, strings that will be broken |
188 | /// by clang-format and string literals with escaped newlines. |
189 | bool nextIsMultilineString(const LineState &State); |
190 | |
191 | FormatStyle Style; |
192 | const AdditionalKeywords &Keywords; |
193 | const SourceManager &SourceMgr; |
194 | WhitespaceManager &Whitespaces; |
195 | encoding::Encoding Encoding; |
196 | bool BinPackInconclusiveFunctions; |
197 | llvm::Regex CommentPragmasRegex; |
198 | const RawStringFormatStyleManager RawStringFormats; |
199 | }; |
200 | |
201 | struct ParenState { |
202 | ParenState(const FormatToken *Tok, unsigned Indent, unsigned LastSpace, |
203 | bool AvoidBinPacking, bool NoLineBreak) |
204 | : Tok(Tok), Indent(Indent), LastSpace(LastSpace), |
205 | NestedBlockIndent(Indent), BreakBeforeClosingBrace(false), |
206 | AvoidBinPacking(AvoidBinPacking), BreakBeforeParameter(false), |
207 | NoLineBreak(NoLineBreak), NoLineBreakInOperand(false), |
208 | LastOperatorWrapped(true), ContainsLineBreak(false), |
209 | ContainsUnwrappedBuilder(false), AlignColons(true), |
210 | ObjCSelectorNameFound(false), HasMultipleNestedBlocks(false), |
211 | NestedBlockInlined(false), IsInsideObjCArrayLiteral(false) {} |
212 | |
213 | /// \brief The token opening this parenthesis level, or nullptr if this level |
214 | /// is opened by fake parenthesis. |
215 | /// |
216 | /// Not considered for memoization as it will always have the same value at |
217 | /// the same token. |
218 | const FormatToken *Tok; |
219 | |
220 | /// The position to which a specific parenthesis level needs to be |
221 | /// indented. |
222 | unsigned Indent; |
223 | |
224 | /// The position of the last space on each level. |
225 | /// |
226 | /// Used e.g. to break like: |
227 | /// functionCall(Parameter, otherCall( |
228 | /// OtherParameter)); |
229 | unsigned LastSpace; |
230 | |
231 | /// If a block relative to this parenthesis level gets wrapped, indent |
232 | /// it this much. |
233 | unsigned NestedBlockIndent; |
234 | |
235 | /// The position the first "<<" operator encountered on each level. |
236 | /// |
237 | /// Used to align "<<" operators. 0 if no such operator has been encountered |
238 | /// on a level. |
239 | unsigned FirstLessLess = 0; |
240 | |
241 | /// The column of a \c ? in a conditional expression; |
242 | unsigned QuestionColumn = 0; |
243 | |
244 | /// The position of the colon in an ObjC method declaration/call. |
245 | unsigned ColonPos = 0; |
246 | |
247 | /// The start of the most recent function in a builder-type call. |
248 | unsigned StartOfFunctionCall = 0; |
249 | |
250 | /// Contains the start of array subscript expressions, so that they |
251 | /// can be aligned. |
252 | unsigned StartOfArraySubscripts = 0; |
253 | |
254 | /// If a nested name specifier was broken over multiple lines, this |
255 | /// contains the start column of the second line. Otherwise 0. |
256 | unsigned NestedNameSpecifierContinuation = 0; |
257 | |
258 | /// If a call expression was broken over multiple lines, this |
259 | /// contains the start column of the second line. Otherwise 0. |
260 | unsigned CallContinuation = 0; |
261 | |
262 | /// The column of the first variable name in a variable declaration. |
263 | /// |
264 | /// Used to align further variables if necessary. |
265 | unsigned VariablePos = 0; |
266 | |
267 | /// Whether a newline needs to be inserted before the block's closing |
268 | /// brace. |
269 | /// |
270 | /// We only want to insert a newline before the closing brace if there also |
271 | /// was a newline after the beginning left brace. |
272 | bool BreakBeforeClosingBrace : 1; |
273 | |
274 | /// Avoid bin packing, i.e. multiple parameters/elements on multiple |
275 | /// lines, in this context. |
276 | bool AvoidBinPacking : 1; |
277 | |
278 | /// Break after the next comma (or all the commas in this context if |
279 | /// \c AvoidBinPacking is \c true). |
280 | bool BreakBeforeParameter : 1; |
281 | |
282 | /// Line breaking in this context would break a formatting rule. |
283 | bool NoLineBreak : 1; |
284 | |
285 | /// Same as \c NoLineBreak, but is restricted until the end of the |
286 | /// operand (including the next ","). |
287 | bool NoLineBreakInOperand : 1; |
288 | |
289 | /// True if the last binary operator on this level was wrapped to the |
290 | /// next line. |
291 | bool LastOperatorWrapped : 1; |
292 | |
293 | /// \c true if this \c ParenState already contains a line-break. |
294 | /// |
295 | /// The first line break in a certain \c ParenState causes extra penalty so |
296 | /// that clang-format prefers similar breaks, i.e. breaks in the same |
297 | /// parenthesis. |
298 | bool ContainsLineBreak : 1; |
299 | |
300 | /// \c true if this \c ParenState contains multiple segments of a |
301 | /// builder-type call on one line. |
302 | bool ContainsUnwrappedBuilder : 1; |
303 | |
304 | /// \c true if the colons of the curren ObjC method expression should |
305 | /// be aligned. |
306 | /// |
307 | /// Not considered for memoization as it will always have the same value at |
308 | /// the same token. |
309 | bool AlignColons : 1; |
310 | |
311 | /// \c true if at least one selector name was found in the current |
312 | /// ObjC method expression. |
313 | /// |
314 | /// Not considered for memoization as it will always have the same value at |
315 | /// the same token. |
316 | bool ObjCSelectorNameFound : 1; |
317 | |
318 | /// \c true if there are multiple nested blocks inside these parens. |
319 | /// |
320 | /// Not considered for memoization as it will always have the same value at |
321 | /// the same token. |
322 | bool HasMultipleNestedBlocks : 1; |
323 | |
324 | /// The start of a nested block (e.g. lambda introducer in C++ or |
325 | /// "function" in JavaScript) is not wrapped to a new line. |
326 | bool NestedBlockInlined : 1; |
327 | |
328 | /// \c true if the current \c ParenState represents an Objective-C |
329 | /// array literal. |
330 | bool IsInsideObjCArrayLiteral : 1; |
331 | |
332 | bool operator<(const ParenState &Other) const { |
333 | if (Indent != Other.Indent) |
334 | return Indent < Other.Indent; |
335 | if (LastSpace != Other.LastSpace) |
336 | return LastSpace < Other.LastSpace; |
337 | if (NestedBlockIndent != Other.NestedBlockIndent) |
338 | return NestedBlockIndent < Other.NestedBlockIndent; |
339 | if (FirstLessLess != Other.FirstLessLess) |
340 | return FirstLessLess < Other.FirstLessLess; |
341 | if (BreakBeforeClosingBrace != Other.BreakBeforeClosingBrace) |
342 | return BreakBeforeClosingBrace; |
343 | if (QuestionColumn != Other.QuestionColumn) |
344 | return QuestionColumn < Other.QuestionColumn; |
345 | if (AvoidBinPacking != Other.AvoidBinPacking) |
346 | return AvoidBinPacking; |
347 | if (BreakBeforeParameter != Other.BreakBeforeParameter) |
348 | return BreakBeforeParameter; |
349 | if (NoLineBreak != Other.NoLineBreak) |
350 | return NoLineBreak; |
351 | if (LastOperatorWrapped != Other.LastOperatorWrapped) |
352 | return LastOperatorWrapped; |
353 | if (ColonPos != Other.ColonPos) |
354 | return ColonPos < Other.ColonPos; |
355 | if (StartOfFunctionCall != Other.StartOfFunctionCall) |
356 | return StartOfFunctionCall < Other.StartOfFunctionCall; |
357 | if (StartOfArraySubscripts != Other.StartOfArraySubscripts) |
358 | return StartOfArraySubscripts < Other.StartOfArraySubscripts; |
359 | if (CallContinuation != Other.CallContinuation) |
360 | return CallContinuation < Other.CallContinuation; |
361 | if (VariablePos != Other.VariablePos) |
362 | return VariablePos < Other.VariablePos; |
363 | if (ContainsLineBreak != Other.ContainsLineBreak) |
364 | return ContainsLineBreak; |
365 | if (ContainsUnwrappedBuilder != Other.ContainsUnwrappedBuilder) |
366 | return ContainsUnwrappedBuilder; |
367 | if (NestedBlockInlined != Other.NestedBlockInlined) |
368 | return NestedBlockInlined; |
369 | return false; |
370 | } |
371 | }; |
372 | |
373 | /// The current state when indenting a unwrapped line. |
374 | /// |
375 | /// As the indenting tries different combinations this is copied by value. |
376 | struct LineState { |
377 | /// The number of used columns in the current line. |
378 | unsigned Column; |
379 | |
380 | /// The token that needs to be next formatted. |
381 | FormatToken *NextToken; |
382 | |
383 | /// \c true if this line contains a continued for-loop section. |
384 | bool LineContainsContinuedForLoopSection; |
385 | |
386 | /// \c true if \p NextToken should not continue this line. |
387 | bool NoContinuation; |
388 | |
389 | /// The \c NestingLevel at the start of this line. |
390 | unsigned StartOfLineLevel; |
391 | |
392 | /// The lowest \c NestingLevel on the current line. |
393 | unsigned LowestLevelOnLine; |
394 | |
395 | /// The start column of the string literal, if we're in a string |
396 | /// literal sequence, 0 otherwise. |
397 | unsigned StartOfStringLiteral; |
398 | |
399 | /// A stack keeping track of properties applying to parenthesis |
400 | /// levels. |
401 | std::vector<ParenState> Stack; |
402 | |
403 | /// Ignore the stack of \c ParenStates for state comparison. |
404 | /// |
405 | /// In long and deeply nested unwrapped lines, the current algorithm can |
406 | /// be insufficient for finding the best formatting with a reasonable amount |
407 | /// of time and memory. Setting this flag will effectively lead to the |
408 | /// algorithm not analyzing some combinations. However, these combinations |
409 | /// rarely contain the optimal solution: In short, accepting a higher |
410 | /// penalty early would need to lead to different values in the \c |
411 | /// ParenState stack (in an otherwise identical state) and these different |
412 | /// values would need to lead to a significant amount of avoided penalty |
413 | /// later. |
414 | /// |
415 | /// FIXME: Come up with a better algorithm instead. |
416 | bool IgnoreStackForComparison; |
417 | |
418 | /// The indent of the first token. |
419 | unsigned FirstIndent; |
420 | |
421 | /// The line that is being formatted. |
422 | /// |
423 | /// Does not need to be considered for memoization because it doesn't change. |
424 | const AnnotatedLine *Line; |
425 | |
426 | /// Comparison operator to be able to used \c LineState in \c map. |
427 | bool operator<(const LineState &Other) const { |
428 | if (NextToken != Other.NextToken) |
429 | return NextToken < Other.NextToken; |
430 | if (Column != Other.Column) |
431 | return Column < Other.Column; |
432 | if (LineContainsContinuedForLoopSection != |
433 | Other.LineContainsContinuedForLoopSection) |
434 | return LineContainsContinuedForLoopSection; |
435 | if (NoContinuation != Other.NoContinuation) |
436 | return NoContinuation; |
437 | if (StartOfLineLevel != Other.StartOfLineLevel) |
438 | return StartOfLineLevel < Other.StartOfLineLevel; |
439 | if (LowestLevelOnLine != Other.LowestLevelOnLine) |
440 | return LowestLevelOnLine < Other.LowestLevelOnLine; |
441 | if (StartOfStringLiteral != Other.StartOfStringLiteral) |
442 | return StartOfStringLiteral < Other.StartOfStringLiteral; |
443 | if (IgnoreStackForComparison || Other.IgnoreStackForComparison) |
444 | return false; |
445 | return Stack < Other.Stack; |
446 | } |
447 | }; |
448 | |
449 | } // end namespace format |
450 | } // end namespace clang |
451 | |
452 | #endif |
453 | |