clang  10.0.0svn
NamespaceEndCommentsFixer.cpp
Go to the documentation of this file.
1 //===--- NamespaceEndCommentsFixer.cpp --------------------------*- 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 NamespaceEndCommentsFixer, a TokenAnalyzer that
11 /// fixes namespace end comments.
12 ///
13 //===----------------------------------------------------------------------===//
14 
16 #include "llvm/Support/Debug.h"
17 #include "llvm/Support/Regex.h"
18 
19 #define DEBUG_TYPE "namespace-end-comments-fixer"
20 
21 namespace clang {
22 namespace format {
23 
24 namespace {
25 // The maximal number of unwrapped lines that a short namespace spans.
26 // Short namespaces don't need an end comment.
27 static const int kShortNamespaceMaxLines = 1;
28 
29 // Computes the name of a namespace given the namespace token.
30 // Returns "" for anonymous namespace.
31 std::string computeName(const FormatToken *NamespaceTok) {
32  assert(NamespaceTok &&
33  NamespaceTok->isOneOf(tok::kw_namespace, TT_NamespaceMacro) &&
34  "expecting a namespace token");
35  std::string name = "";
36  const FormatToken *Tok = NamespaceTok->getNextNonComment();
37  if (NamespaceTok->is(TT_NamespaceMacro)) {
38  // Collects all the non-comment tokens between opening parenthesis
39  // and closing parenthesis or comma
40  assert(Tok && Tok->is(tok::l_paren) && "expected an opening parenthesis");
41  Tok = Tok->getNextNonComment();
42  while (Tok && !Tok->isOneOf(tok::r_paren, tok::comma)) {
43  name += Tok->TokenText;
44  Tok = Tok->getNextNonComment();
45  }
46  } else {
47  // Collects all the non-comment tokens between 'namespace' and '{'.
48  while (Tok && !Tok->is(tok::l_brace)) {
49  name += Tok->TokenText;
50  Tok = Tok->getNextNonComment();
51  }
52  }
53  return name;
54 }
55 
56 std::string computeEndCommentText(StringRef NamespaceName, bool AddNewline,
57  const FormatToken *NamespaceTok) {
58  std::string text = "// ";
59  text += NamespaceTok->TokenText;
60  if (NamespaceTok->is(TT_NamespaceMacro))
61  text += "(";
62  else if (!NamespaceName.empty())
63  text += ' ';
64  text += NamespaceName;
65  if (NamespaceTok->is(TT_NamespaceMacro))
66  text += ")";
67  if (AddNewline)
68  text += '\n';
69  return text;
70 }
71 
72 bool hasEndComment(const FormatToken *RBraceTok) {
73  return RBraceTok->Next && RBraceTok->Next->is(tok::comment);
74 }
75 
76 bool validEndComment(const FormatToken *RBraceTok, StringRef NamespaceName,
77  const FormatToken *NamespaceTok) {
78  assert(hasEndComment(RBraceTok));
79  const FormatToken *Comment = RBraceTok->Next;
80 
81  // Matches a valid namespace end comment.
82  // Valid namespace end comments don't need to be edited.
83  static llvm::Regex *const NamespaceCommentPattern =
84  new llvm::Regex("^/[/*] *(end (of )?)? *(anonymous|unnamed)? *"
85  "namespace( +([a-zA-Z0-9:_]+))?\\.? *(\\*/)?$",
86  llvm::Regex::IgnoreCase);
87  static llvm::Regex *const NamespaceMacroCommentPattern =
88  new llvm::Regex("^/[/*] *(end (of )?)? *(anonymous|unnamed)? *"
89  "([a-zA-Z0-9_]+)\\(([a-zA-Z0-9:_]*)\\)\\.? *(\\*/)?$",
90  llvm::Regex::IgnoreCase);
91 
92  SmallVector<StringRef, 8> Groups;
93  if (NamespaceTok->is(TT_NamespaceMacro) &&
94  NamespaceMacroCommentPattern->match(Comment->TokenText, &Groups)) {
95  StringRef NamespaceTokenText = Groups.size() > 4 ? Groups[4] : "";
96  // The name of the macro must be used.
97  if (NamespaceTokenText != NamespaceTok->TokenText)
98  return false;
99  } else if (NamespaceTok->isNot(tok::kw_namespace) ||
100  !NamespaceCommentPattern->match(Comment->TokenText, &Groups)) {
101  // Comment does not match regex.
102  return false;
103  }
104  StringRef NamespaceNameInComment = Groups.size() > 5 ? Groups[5] : "";
105  // Anonymous namespace comments must not mention a namespace name.
106  if (NamespaceName.empty() && !NamespaceNameInComment.empty())
107  return false;
108  StringRef AnonymousInComment = Groups.size() > 3 ? Groups[3] : "";
109  // Named namespace comments must not mention anonymous namespace.
110  if (!NamespaceName.empty() && !AnonymousInComment.empty())
111  return false;
112  return NamespaceNameInComment == NamespaceName;
113 }
114 
115 void addEndComment(const FormatToken *RBraceTok, StringRef EndCommentText,
116  const SourceManager &SourceMgr,
117  tooling::Replacements *Fixes) {
118  auto EndLoc = RBraceTok->Tok.getEndLoc();
119  auto Range = CharSourceRange::getCharRange(EndLoc, EndLoc);
120  auto Err = Fixes->add(tooling::Replacement(SourceMgr, Range, EndCommentText));
121  if (Err) {
122  llvm::errs() << "Error while adding namespace end comment: "
123  << llvm::toString(std::move(Err)) << "\n";
124  }
125 }
126 
127 void updateEndComment(const FormatToken *RBraceTok, StringRef EndCommentText,
128  const SourceManager &SourceMgr,
129  tooling::Replacements *Fixes) {
130  assert(hasEndComment(RBraceTok));
131  const FormatToken *Comment = RBraceTok->Next;
132  auto Range = CharSourceRange::getCharRange(Comment->getStartOfNonWhitespace(),
133  Comment->Tok.getEndLoc());
134  auto Err = Fixes->add(tooling::Replacement(SourceMgr, Range, EndCommentText));
135  if (Err) {
136  llvm::errs() << "Error while updating namespace end comment: "
137  << llvm::toString(std::move(Err)) << "\n";
138  }
139 }
140 } // namespace
141 
142 const FormatToken *
144  const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
145  if (!Line->Affected || Line->InPPDirective || !Line->startsWith(tok::r_brace))
146  return nullptr;
147  size_t StartLineIndex = Line->MatchingOpeningBlockLineIndex;
148  if (StartLineIndex == UnwrappedLine::kInvalidIndex)
149  return nullptr;
150  assert(StartLineIndex < AnnotatedLines.size());
151  const FormatToken *NamespaceTok = AnnotatedLines[StartLineIndex]->First;
152  if (NamespaceTok->is(tok::l_brace)) {
153  // "namespace" keyword can be on the line preceding '{', e.g. in styles
154  // where BraceWrapping.AfterNamespace is true.
155  if (StartLineIndex > 0)
156  NamespaceTok = AnnotatedLines[StartLineIndex - 1]->First;
157  }
158  return NamespaceTok->getNamespaceToken();
159 }
160 
161 StringRef
163  const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
164  const FormatToken *NamespaceTok = getNamespaceToken(Line, AnnotatedLines);
165  return NamespaceTok ? NamespaceTok->TokenText : StringRef();
166 }
167 
169  const FormatStyle &Style)
170  : TokenAnalyzer(Env, Style) {}
171 
172 std::pair<tooling::Replacements, unsigned> NamespaceEndCommentsFixer::analyze(
173  TokenAnnotator &Annotator, SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
174  FormatTokenLexer &Tokens) {
175  const SourceManager &SourceMgr = Env.getSourceManager();
176  AffectedRangeMgr.computeAffectedLines(AnnotatedLines);
177  tooling::Replacements Fixes;
178  std::string AllNamespaceNames = "";
179  size_t StartLineIndex = SIZE_MAX;
180  StringRef NamespaceTokenText;
181  unsigned int CompactedNamespacesCount = 0;
182  for (size_t I = 0, E = AnnotatedLines.size(); I != E; ++I) {
183  const AnnotatedLine *EndLine = AnnotatedLines[I];
184  const FormatToken *NamespaceTok =
185  getNamespaceToken(EndLine, AnnotatedLines);
186  if (!NamespaceTok)
187  continue;
188  FormatToken *RBraceTok = EndLine->First;
189  if (RBraceTok->Finalized)
190  continue;
191  RBraceTok->Finalized = true;
192  const FormatToken *EndCommentPrevTok = RBraceTok;
193  // Namespaces often end with '};'. In that case, attach namespace end
194  // comments to the semicolon tokens.
195  if (RBraceTok->Next && RBraceTok->Next->is(tok::semi)) {
196  EndCommentPrevTok = RBraceTok->Next;
197  }
198  if (StartLineIndex == SIZE_MAX)
199  StartLineIndex = EndLine->MatchingOpeningBlockLineIndex;
200  std::string NamespaceName = computeName(NamespaceTok);
201  if (Style.CompactNamespaces) {
202  if (CompactedNamespacesCount == 0)
203  NamespaceTokenText = NamespaceTok->TokenText;
204  if ((I + 1 < E) &&
205  NamespaceTokenText ==
206  getNamespaceTokenText(AnnotatedLines[I + 1], AnnotatedLines) &&
207  StartLineIndex - CompactedNamespacesCount - 1 ==
208  AnnotatedLines[I + 1]->MatchingOpeningBlockLineIndex &&
209  !AnnotatedLines[I + 1]->First->Finalized) {
210  if (hasEndComment(EndCommentPrevTok)) {
211  // remove end comment, it will be merged in next one
212  updateEndComment(EndCommentPrevTok, std::string(), SourceMgr, &Fixes);
213  }
214  CompactedNamespacesCount++;
215  AllNamespaceNames = "::" + NamespaceName + AllNamespaceNames;
216  continue;
217  }
218  NamespaceName += AllNamespaceNames;
219  CompactedNamespacesCount = 0;
220  AllNamespaceNames = std::string();
221  }
222  // The next token in the token stream after the place where the end comment
223  // token must be. This is either the next token on the current line or the
224  // first token on the next line.
225  const FormatToken *EndCommentNextTok = EndCommentPrevTok->Next;
226  if (EndCommentNextTok && EndCommentNextTok->is(tok::comment))
227  EndCommentNextTok = EndCommentNextTok->Next;
228  if (!EndCommentNextTok && I + 1 < E)
229  EndCommentNextTok = AnnotatedLines[I + 1]->First;
230  bool AddNewline = EndCommentNextTok &&
231  EndCommentNextTok->NewlinesBefore == 0 &&
232  EndCommentNextTok->isNot(tok::eof);
233  const std::string EndCommentText =
234  computeEndCommentText(NamespaceName, AddNewline, NamespaceTok);
235  if (!hasEndComment(EndCommentPrevTok)) {
236  bool isShort = I - StartLineIndex <= kShortNamespaceMaxLines + 1;
237  if (!isShort)
238  addEndComment(EndCommentPrevTok, EndCommentText, SourceMgr, &Fixes);
239  } else if (!validEndComment(EndCommentPrevTok, NamespaceName,
240  NamespaceTok)) {
241  updateEndComment(EndCommentPrevTok, EndCommentText, SourceMgr, &Fixes);
242  }
243  StartLineIndex = SIZE_MAX;
244  }
245  return {Fixes, 0};
246 }
247 
248 } // namespace format
249 } // namespace clang
if(T->getSizeExpr()) TRY_TO(TraverseStmt(T -> getSizeExpr()))
StringRef getNamespaceTokenText(const AnnotatedLine *Line, const SmallVectorImpl< AnnotatedLine *> &AnnotatedLines)
NamespaceEndCommentsFixer(const Environment &Env, const FormatStyle &Style)
const FormatToken * getNextNonComment() const
Returns the next token ignoring comments.
Definition: FormatToken.h:486
RangeSelector name(std::string ID)
Given a node with a "name", (like NamedDecl, DeclRefExpr or CxxCtorInitializer) selects the name&#39;s to...
Maintains a set of replacements that are conflict-free.
Definition: Replacement.h:209
AffectedRangeManager AffectedRangeMgr
unsigned NewlinesBefore
The number of newlines immediately before the Token.
Definition: FormatToken.h:139
FormatToken * Next
The next token in the unwrapped line.
Definition: FormatToken.h:298
const Environment & Env
Definition: TokenAnalyzer.h:98
static std::string toString(const clang::SanitizerSet &Sanitizers)
Produce a string containing comma-separated names of sanitizers in Sanitizers set.
const FormatToken * getNamespaceToken(const AnnotatedLine *Line, const SmallVectorImpl< AnnotatedLine *> &AnnotatedLines)
bool isNot(T Kind) const
Definition: FormatToken.h:328
const FormatToken & Tok
const SourceManager & getSourceManager() const
Definition: TokenAnalyzer.h:49
Determines extra information about the tokens comprising an UnwrappedLine.
const AnnotatedLine * Line
static const size_t kInvalidIndex
A wrapper around a Token storing information about the whitespace characters preceding it...
Definition: FormatToken.h:129
bool Affected
True if this line should be formatted, i.e.
static CharSourceRange getCharRange(SourceRange R)
std::pair< tooling::Replacements, unsigned > analyze(TokenAnnotator &Annotator, SmallVectorImpl< AnnotatedLine *> &AnnotatedLines, FormatTokenLexer &Tokens) override
bool is(tok::TokenKind Kind) const
Definition: FormatToken.h:312
bool startsWith(Ts... Tokens) const
true if this line starts with the given tokens in order, ignoring comments.
StringRef TokenText
The raw text of the token.
Definition: FormatToken.h:177
#define SIZE_MAX
Definition: stdint.h:653
The FormatStyle is used to configure the formatting to follow specific guidelines.
Definition: Format.h:49
bool computeAffectedLines(SmallVectorImpl< AnnotatedLine *> &Lines)
Dataflow Directional Tag Classes.
bool Finalized
If true, this token has been fully formatted (indented and potentially re-formatted inside)...
Definition: FormatToken.h:310
This file declares NamespaceEndCommentsFixer, a TokenAnalyzer that fixes namespace end comments...
This class handles loading and caching of source files into memory.
const FormatStyle & Style