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 
15 #include "NamespaceEndCommentsFixer.h"
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     // For `namespace [[foo]] A::B::inline C {` or
48     // `namespace MACRO1 MACRO2 A::B::inline C {`, returns "A::B::inline C".
49     // Peek for the first '::' (or '{') and then return all tokens from one
50     // token before that up until the '{'.
51     const FormatToken *FirstNSTok = Tok;
52     while (Tok && !Tok->is(tok::l_brace) && !Tok->is(tok::coloncolon)) {
53       FirstNSTok = Tok;
54       Tok = Tok->getNextNonComment();
55     }
56 
57     Tok = FirstNSTok;
58     while (Tok && !Tok->is(tok::l_brace)) {
59       name += Tok->TokenText;
60       if (Tok->is(tok::kw_inline))
61         name += " ";
62       Tok = Tok->getNextNonComment();
63     }
64   }
65   return name;
66 }
67 
68 std::string computeEndCommentText(StringRef NamespaceName, bool AddNewline,
69                                   const FormatToken *NamespaceTok) {
70   std::string text = "// ";
71   text += NamespaceTok->TokenText;
72   if (NamespaceTok->is(TT_NamespaceMacro))
73     text += "(";
74   else if (!NamespaceName.empty())
75     text += ' ';
76   text += NamespaceName;
77   if (NamespaceTok->is(TT_NamespaceMacro))
78     text += ")";
79   if (AddNewline)
80     text += '\n';
81   return text;
82 }
83 
84 bool hasEndComment(const FormatToken *RBraceTok) {
85   return RBraceTok->Next && RBraceTok->Next->is(tok::comment);
86 }
87 
88 bool validEndComment(const FormatToken *RBraceTok, StringRef NamespaceName,
89                      const FormatToken *NamespaceTok) {
90   assert(hasEndComment(RBraceTok));
91   const FormatToken *Comment = RBraceTok->Next;
92 
93   // Matches a valid namespace end comment.
94   // Valid namespace end comments don't need to be edited.
95   static const llvm::Regex NamespaceCommentPattern =
96       llvm::Regex("^/[/*] *(end (of )?)? *(anonymous|unnamed)? *"
97                   "namespace( +([a-zA-Z0-9:_]+))?\\.? *(\\*/)?$",
98                   llvm::Regex::IgnoreCase);
99   static const llvm::Regex NamespaceMacroCommentPattern =
100       llvm::Regex("^/[/*] *(end (of )?)? *(anonymous|unnamed)? *"
101                   "([a-zA-Z0-9_]+)\\(([a-zA-Z0-9:_]*)\\)\\.? *(\\*/)?$",
102                   llvm::Regex::IgnoreCase);
103 
104   SmallVector<StringRef, 8> Groups;
105   if (NamespaceTok->is(TT_NamespaceMacro) &&
106       NamespaceMacroCommentPattern.match(Comment->TokenText, &Groups)) {
107     StringRef NamespaceTokenText = Groups.size() > 4 ? Groups[4] : "";
108     // The name of the macro must be used.
109     if (NamespaceTokenText != NamespaceTok->TokenText)
110       return false;
111   } else if (NamespaceTok->isNot(tok::kw_namespace) ||
112              !NamespaceCommentPattern.match(Comment->TokenText, &Groups)) {
113     // Comment does not match regex.
114     return false;
115   }
116   StringRef NamespaceNameInComment = Groups.size() > 5 ? Groups[5] : "";
117   // Anonymous namespace comments must not mention a namespace name.
118   if (NamespaceName.empty() && !NamespaceNameInComment.empty())
119     return false;
120   StringRef AnonymousInComment = Groups.size() > 3 ? Groups[3] : "";
121   // Named namespace comments must not mention anonymous namespace.
122   if (!NamespaceName.empty() && !AnonymousInComment.empty())
123     return false;
124   if (NamespaceNameInComment == NamespaceName)
125     return true;
126 
127   // Has namespace comment flowed onto the next line.
128   // } // namespace
129   //   // verylongnamespacenamethatdidnotfitonthepreviouscommentline
130   if (!(Comment->Next && Comment->Next->is(TT_LineComment)))
131     return false;
132 
133   static const llvm::Regex CommentPattern = llvm::Regex(
134       "^/[/*] *( +([a-zA-Z0-9:_]+))?\\.? *(\\*/)?$", llvm::Regex::IgnoreCase);
135 
136   // Pull out just the comment text.
137   if (!CommentPattern.match(Comment->Next->TokenText, &Groups)) {
138     return false;
139   }
140   NamespaceNameInComment = Groups.size() > 2 ? Groups[2] : "";
141 
142   return (NamespaceNameInComment == NamespaceName);
143 }
144 
145 void addEndComment(const FormatToken *RBraceTok, StringRef EndCommentText,
146                    const SourceManager &SourceMgr,
147                    tooling::Replacements *Fixes) {
148   auto EndLoc = RBraceTok->Tok.getEndLoc();
149   auto Range = CharSourceRange::getCharRange(EndLoc, EndLoc);
150   auto Err = Fixes->add(tooling::Replacement(SourceMgr, Range, EndCommentText));
151   if (Err) {
152     llvm::errs() << "Error while adding namespace end comment: "
153                  << llvm::toString(std::move(Err)) << "\n";
154   }
155 }
156 
157 void updateEndComment(const FormatToken *RBraceTok, StringRef EndCommentText,
158                       const SourceManager &SourceMgr,
159                       tooling::Replacements *Fixes) {
160   assert(hasEndComment(RBraceTok));
161   const FormatToken *Comment = RBraceTok->Next;
162   auto Range = CharSourceRange::getCharRange(Comment->getStartOfNonWhitespace(),
163                                              Comment->Tok.getEndLoc());
164   auto Err = Fixes->add(tooling::Replacement(SourceMgr, Range, EndCommentText));
165   if (Err) {
166     llvm::errs() << "Error while updating namespace end comment: "
167                  << llvm::toString(std::move(Err)) << "\n";
168   }
169 }
170 } // namespace
171 
172 const FormatToken *
173 getNamespaceToken(const AnnotatedLine *Line,
174                   const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
175   if (!Line->Affected || Line->InPPDirective || !Line->startsWith(tok::r_brace))
176     return nullptr;
177   size_t StartLineIndex = Line->MatchingOpeningBlockLineIndex;
178   if (StartLineIndex == UnwrappedLine::kInvalidIndex)
179     return nullptr;
180   assert(StartLineIndex < AnnotatedLines.size());
181   const FormatToken *NamespaceTok = AnnotatedLines[StartLineIndex]->First;
182   if (NamespaceTok->is(tok::l_brace)) {
183     // "namespace" keyword can be on the line preceding '{', e.g. in styles
184     // where BraceWrapping.AfterNamespace is true.
185     if (StartLineIndex > 0)
186       NamespaceTok = AnnotatedLines[StartLineIndex - 1]->First;
187   }
188   return NamespaceTok->getNamespaceToken();
189 }
190 
191 StringRef
192 getNamespaceTokenText(const AnnotatedLine *Line,
193                       const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
194   const FormatToken *NamespaceTok = getNamespaceToken(Line, AnnotatedLines);
195   return NamespaceTok ? NamespaceTok->TokenText : StringRef();
196 }
197 
198 NamespaceEndCommentsFixer::NamespaceEndCommentsFixer(const Environment &Env,
199                                                      const FormatStyle &Style)
200     : TokenAnalyzer(Env, Style) {}
201 
202 std::pair<tooling::Replacements, unsigned> NamespaceEndCommentsFixer::analyze(
203     TokenAnnotator &Annotator, SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
204     FormatTokenLexer &Tokens) {
205   const SourceManager &SourceMgr = Env.getSourceManager();
206   AffectedRangeMgr.computeAffectedLines(AnnotatedLines);
207   tooling::Replacements Fixes;
208   std::string AllNamespaceNames = "";
209   size_t StartLineIndex = SIZE_MAX;
210   StringRef NamespaceTokenText;
211   unsigned int CompactedNamespacesCount = 0;
212   for (size_t I = 0, E = AnnotatedLines.size(); I != E; ++I) {
213     const AnnotatedLine *EndLine = AnnotatedLines[I];
214     const FormatToken *NamespaceTok =
215         getNamespaceToken(EndLine, AnnotatedLines);
216     if (!NamespaceTok)
217       continue;
218     FormatToken *RBraceTok = EndLine->First;
219     if (RBraceTok->Finalized)
220       continue;
221     RBraceTok->Finalized = true;
222     const FormatToken *EndCommentPrevTok = RBraceTok;
223     // Namespaces often end with '};'. In that case, attach namespace end
224     // comments to the semicolon tokens.
225     if (RBraceTok->Next && RBraceTok->Next->is(tok::semi)) {
226       EndCommentPrevTok = RBraceTok->Next;
227     }
228     if (StartLineIndex == SIZE_MAX)
229       StartLineIndex = EndLine->MatchingOpeningBlockLineIndex;
230     std::string NamespaceName = computeName(NamespaceTok);
231     if (Style.CompactNamespaces) {
232       if (CompactedNamespacesCount == 0)
233         NamespaceTokenText = NamespaceTok->TokenText;
234       if ((I + 1 < E) &&
235           NamespaceTokenText ==
236               getNamespaceTokenText(AnnotatedLines[I + 1], AnnotatedLines) &&
237           StartLineIndex - CompactedNamespacesCount - 1 ==
238               AnnotatedLines[I + 1]->MatchingOpeningBlockLineIndex &&
239           !AnnotatedLines[I + 1]->First->Finalized) {
240         if (hasEndComment(EndCommentPrevTok)) {
241           // remove end comment, it will be merged in next one
242           updateEndComment(EndCommentPrevTok, std::string(), SourceMgr, &Fixes);
243         }
244         CompactedNamespacesCount++;
245         AllNamespaceNames = "::" + NamespaceName + AllNamespaceNames;
246         continue;
247       }
248       NamespaceName += AllNamespaceNames;
249       CompactedNamespacesCount = 0;
250       AllNamespaceNames = std::string();
251     }
252     // The next token in the token stream after the place where the end comment
253     // token must be. This is either the next token on the current line or the
254     // first token on the next line.
255     const FormatToken *EndCommentNextTok = EndCommentPrevTok->Next;
256     if (EndCommentNextTok && EndCommentNextTok->is(tok::comment))
257       EndCommentNextTok = EndCommentNextTok->Next;
258     if (!EndCommentNextTok && I + 1 < E)
259       EndCommentNextTok = AnnotatedLines[I + 1]->First;
260     bool AddNewline = EndCommentNextTok &&
261                       EndCommentNextTok->NewlinesBefore == 0 &&
262                       EndCommentNextTok->isNot(tok::eof);
263     const std::string EndCommentText =
264         computeEndCommentText(NamespaceName, AddNewline, NamespaceTok);
265     if (!hasEndComment(EndCommentPrevTok)) {
266       bool isShort = I - StartLineIndex <= kShortNamespaceMaxLines + 1;
267       if (!isShort)
268         addEndComment(EndCommentPrevTok, EndCommentText, SourceMgr, &Fixes);
269     } else if (!validEndComment(EndCommentPrevTok, NamespaceName,
270                                 NamespaceTok)) {
271       updateEndComment(EndCommentPrevTok, EndCommentText, SourceMgr, &Fixes);
272     }
273     StartLineIndex = SIZE_MAX;
274   }
275   return {Fixes, 0};
276 }
277 
278 } // namespace format
279 } // namespace clang
280