1 //===--- NamespaceEndCommentsFixer.cpp --------------------------*- C++ -*-===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 ///
10 /// \file
11 /// \brief This file implements NamespaceEndCommentsFixer, a TokenAnalyzer that
12 /// fixes namespace end comments.
13 ///
14 //===----------------------------------------------------------------------===//
15 
16 #include "NamespaceEndCommentsFixer.h"
17 #include "llvm/Support/Debug.h"
18 #include "llvm/Support/Regex.h"
19 
20 #define DEBUG_TYPE "namespace-end-comments-fixer"
21 
22 namespace clang {
23 namespace format {
24 
25 namespace {
26 // The maximal number of unwrapped lines that a short namespace spans.
27 // Short namespaces don't need an end comment.
28 static const int kShortNamespaceMaxLines = 1;
29 
30 // Matches a valid namespace end comment.
31 // Valid namespace end comments don't need to be edited.
32 static llvm::Regex kNamespaceCommentPattern =
33     llvm::Regex("^/[/*] *(end (of )?)? *(anonymous|unnamed)? *"
34                 "namespace( +([a-zA-Z0-9:_]+))?\\.? *(\\*/)?$",
35                 llvm::Regex::IgnoreCase);
36 
37 // Computes the name of a namespace given the namespace token.
38 // Returns "" for anonymous namespace.
39 std::string computeName(const FormatToken *NamespaceTok) {
40   assert(NamespaceTok && NamespaceTok->is(tok::kw_namespace) &&
41          "expecting a namespace token");
42   std::string name = "";
43   // Collects all the non-comment tokens between 'namespace' and '{'.
44   const FormatToken *Tok = NamespaceTok->getNextNonComment();
45   while (Tok && !Tok->is(tok::l_brace)) {
46     name += Tok->TokenText;
47     Tok = Tok->getNextNonComment();
48   }
49   return name;
50 }
51 
52 std::string computeEndCommentText(StringRef NamespaceName, bool AddNewline) {
53   std::string text = "// namespace";
54   if (!NamespaceName.empty()) {
55     text += ' ';
56     text += NamespaceName;
57   }
58   if (AddNewline)
59     text += '\n';
60   return text;
61 }
62 
63 bool hasEndComment(const FormatToken *RBraceTok) {
64   return RBraceTok->Next && RBraceTok->Next->is(tok::comment);
65 }
66 
67 bool validEndComment(const FormatToken *RBraceTok, StringRef NamespaceName) {
68   assert(hasEndComment(RBraceTok));
69   const FormatToken *Comment = RBraceTok->Next;
70   SmallVector<StringRef, 7> Groups;
71   if (kNamespaceCommentPattern.match(Comment->TokenText, &Groups)) {
72     StringRef NamespaceNameInComment = Groups.size() > 5 ? Groups[5] : "";
73     // Anonymous namespace comments must not mention a namespace name.
74     if (NamespaceName.empty() && !NamespaceNameInComment.empty())
75       return false;
76     StringRef AnonymousInComment = Groups.size() > 3 ? Groups[3] : "";
77     // Named namespace comments must not mention anonymous namespace.
78     if (!NamespaceName.empty() && !AnonymousInComment.empty())
79       return false;
80     return NamespaceNameInComment == NamespaceName;
81   }
82   return false;
83 }
84 
85 void addEndComment(const FormatToken *RBraceTok, StringRef EndCommentText,
86                    const SourceManager &SourceMgr,
87                    tooling::Replacements *Fixes) {
88   auto EndLoc = RBraceTok->Tok.getEndLoc();
89   auto Range = CharSourceRange::getCharRange(EndLoc, EndLoc);
90   auto Err = Fixes->add(tooling::Replacement(SourceMgr, Range, EndCommentText));
91   if (Err) {
92     llvm::errs() << "Error while adding namespace end comment: "
93                  << llvm::toString(std::move(Err)) << "\n";
94   }
95 }
96 
97 void updateEndComment(const FormatToken *RBraceTok, StringRef EndCommentText,
98                       const SourceManager &SourceMgr,
99                       tooling::Replacements *Fixes) {
100   assert(hasEndComment(RBraceTok));
101   const FormatToken *Comment = RBraceTok->Next;
102   auto Range = CharSourceRange::getCharRange(Comment->getStartOfNonWhitespace(),
103                                              Comment->Tok.getEndLoc());
104   auto Err = Fixes->add(tooling::Replacement(SourceMgr, Range, EndCommentText));
105   if (Err) {
106     llvm::errs() << "Error while updating namespace end comment: "
107                  << llvm::toString(std::move(Err)) << "\n";
108   }
109 }
110 
111 const FormatToken *
112 getNamespaceToken(const AnnotatedLine *line,
113                   const SmallVectorImpl<AnnotatedLine *> &AnnotatedLines) {
114   if (!line->Affected || line->InPPDirective || !line->startsWith(tok::r_brace))
115     return nullptr;
116   size_t StartLineIndex = line->MatchingOpeningBlockLineIndex;
117   if (StartLineIndex == UnwrappedLine::kInvalidIndex)
118     return nullptr;
119   assert(StartLineIndex < AnnotatedLines.size());
120   const FormatToken *NamespaceTok = AnnotatedLines[StartLineIndex]->First;
121   if (NamespaceTok->is(tok::l_brace)) {
122     // "namespace" keyword can be on the line preceding '{', e.g. in styles
123     // where BraceWrapping.AfterNamespace is true.
124     if (StartLineIndex > 0)
125       NamespaceTok = AnnotatedLines[StartLineIndex - 1]->First;
126   }
127   // Detect "(inline)? namespace" in the beginning of a line.
128   if (NamespaceTok->is(tok::kw_inline))
129     NamespaceTok = NamespaceTok->getNextNonComment();
130   if (!NamespaceTok || NamespaceTok->isNot(tok::kw_namespace))
131     return nullptr;
132   return NamespaceTok;
133 }
134 } // namespace
135 
136 NamespaceEndCommentsFixer::NamespaceEndCommentsFixer(const Environment &Env,
137                                                      const FormatStyle &Style)
138     : TokenAnalyzer(Env, Style) {}
139 
140 std::pair<tooling::Replacements, unsigned> NamespaceEndCommentsFixer::analyze(
141     TokenAnnotator &Annotator, SmallVectorImpl<AnnotatedLine *> &AnnotatedLines,
142     FormatTokenLexer &Tokens) {
143   const SourceManager &SourceMgr = Env.getSourceManager();
144   AffectedRangeMgr.computeAffectedLines(AnnotatedLines.begin(),
145                                         AnnotatedLines.end());
146   tooling::Replacements Fixes;
147   std::string AllNamespaceNames = "";
148   size_t StartLineIndex = SIZE_MAX;
149   unsigned int CompactedNamespacesCount = 0;
150   for (size_t I = 0, E = AnnotatedLines.size(); I != E; ++I) {
151     const AnnotatedLine *EndLine = AnnotatedLines[I];
152     const FormatToken *NamespaceTok =
153         getNamespaceToken(EndLine, AnnotatedLines);
154     if (!NamespaceTok)
155       continue;
156     FormatToken *RBraceTok = EndLine->First;
157     if (RBraceTok->Finalized)
158       continue;
159     RBraceTok->Finalized = true;
160     const FormatToken *EndCommentPrevTok = RBraceTok;
161     // Namespaces often end with '};'. In that case, attach namespace end
162     // comments to the semicolon tokens.
163     if (RBraceTok->Next && RBraceTok->Next->is(tok::semi)) {
164       EndCommentPrevTok = RBraceTok->Next;
165     }
166     if (StartLineIndex == SIZE_MAX)
167       StartLineIndex = EndLine->MatchingOpeningBlockLineIndex;
168     std::string NamespaceName = computeName(NamespaceTok);
169     if (Style.CompactNamespaces) {
170       if ((I + 1 < E) &&
171           getNamespaceToken(AnnotatedLines[I + 1], AnnotatedLines) &&
172           StartLineIndex - CompactedNamespacesCount - 1 ==
173               AnnotatedLines[I + 1]->MatchingOpeningBlockLineIndex &&
174           !AnnotatedLines[I + 1]->First->Finalized) {
175         if (hasEndComment(EndCommentPrevTok)) {
176           // remove end comment, it will be merged in next one
177           updateEndComment(EndCommentPrevTok, std::string(), SourceMgr, &Fixes);
178         }
179         CompactedNamespacesCount++;
180         AllNamespaceNames = "::" + NamespaceName + AllNamespaceNames;
181         continue;
182       }
183       NamespaceName += AllNamespaceNames;
184       CompactedNamespacesCount = 0;
185       AllNamespaceNames = std::string();
186     }
187     // The next token in the token stream after the place where the end comment
188     // token must be. This is either the next token on the current line or the
189     // first token on the next line.
190     const FormatToken *EndCommentNextTok = EndCommentPrevTok->Next;
191     if (EndCommentNextTok && EndCommentNextTok->is(tok::comment))
192       EndCommentNextTok = EndCommentNextTok->Next;
193     if (!EndCommentNextTok && I + 1 < E)
194       EndCommentNextTok = AnnotatedLines[I + 1]->First;
195     bool AddNewline = EndCommentNextTok &&
196                       EndCommentNextTok->NewlinesBefore == 0 &&
197                       EndCommentNextTok->isNot(tok::eof);
198     const std::string EndCommentText =
199         computeEndCommentText(NamespaceName, AddNewline);
200     if (!hasEndComment(EndCommentPrevTok)) {
201       bool isShort = I - StartLineIndex <= kShortNamespaceMaxLines + 1;
202       if (!isShort)
203         addEndComment(EndCommentPrevTok, EndCommentText, SourceMgr, &Fixes);
204     } else if (!validEndComment(EndCommentPrevTok, NamespaceName)) {
205       updateEndComment(EndCommentPrevTok, EndCommentText, SourceMgr, &Fixes);
206     }
207     StartLineIndex = SIZE_MAX;
208   }
209   return {Fixes, 0};
210 }
211 
212 } // namespace format
213 } // namespace clang
214