1//===-- ClangHighlighter.cpp ----------------------------------------------===//
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 "ClangHighlighter.h"
10
11#include "lldb/Host/FileSystem.h"
12#include "lldb/Target/Language.h"
13#include "lldb/Utility/AnsiTerminal.h"
14#include "lldb/Utility/StreamString.h"
15
16#include "clang/Basic/FileManager.h"
17#include "clang/Basic/SourceManager.h"
18#include "clang/Lex/Lexer.h"
19#include "llvm/ADT/StringSet.h"
20#include "llvm/Support/MemoryBuffer.h"
21#include <optional>
22
23using namespace lldb_private;
24
25bool ClangHighlighter::isKeyword(llvm::StringRef token) const {
26  return keywords.contains(token);
27}
28
29ClangHighlighter::ClangHighlighter() {
30#define KEYWORD(X, N) keywords.insert(#X);
31#include "clang/Basic/TokenKinds.def"
32}
33
34/// Determines which style should be applied to the given token.
35/// \param highlighter
36///     The current highlighter that should use the style.
37/// \param token
38///     The current token.
39/// \param tok_str
40///     The string in the source code the token represents.
41/// \param options
42///     The style we use for coloring the source code.
43/// \param in_pp_directive
44///     If we are currently in a preprocessor directive. NOTE: This is
45///     passed by reference and will be updated if the current token starts
46///     or ends a preprocessor directive.
47/// \return
48///     The ColorStyle that should be applied to the token.
49static HighlightStyle::ColorStyle
50determineClangStyle(const ClangHighlighter &highlighter,
51                    const clang::Token &token, llvm::StringRef tok_str,
52                    const HighlightStyle &options, bool &in_pp_directive) {
53  using namespace clang;
54
55  if (token.is(tok::comment)) {
56    // If we were in a preprocessor directive before, we now left it.
57    in_pp_directive = false;
58    return options.comment;
59  } else if (in_pp_directive || token.getKind() == tok::hash) {
60    // Let's assume that the rest of the line is a PP directive.
61    in_pp_directive = true;
62    // Preprocessor directives are hard to match, so we have to hack this in.
63    return options.pp_directive;
64  } else if (tok::isStringLiteral(token.getKind()))
65    return options.string_literal;
66  else if (tok::isLiteral(token.getKind()))
67    return options.scalar_literal;
68  else if (highlighter.isKeyword(tok_str))
69    return options.keyword;
70  else
71    switch (token.getKind()) {
72    case tok::raw_identifier:
73    case tok::identifier:
74      return options.identifier;
75    case tok::l_brace:
76    case tok::r_brace:
77      return options.braces;
78    case tok::l_square:
79    case tok::r_square:
80      return options.square_brackets;
81    case tok::l_paren:
82    case tok::r_paren:
83      return options.parentheses;
84    case tok::comma:
85      return options.comma;
86    case tok::coloncolon:
87    case tok::colon:
88      return options.colon;
89
90    case tok::amp:
91    case tok::ampamp:
92    case tok::ampequal:
93    case tok::star:
94    case tok::starequal:
95    case tok::plus:
96    case tok::plusplus:
97    case tok::plusequal:
98    case tok::minus:
99    case tok::arrow:
100    case tok::minusminus:
101    case tok::minusequal:
102    case tok::tilde:
103    case tok::exclaim:
104    case tok::exclaimequal:
105    case tok::slash:
106    case tok::slashequal:
107    case tok::percent:
108    case tok::percentequal:
109    case tok::less:
110    case tok::lessless:
111    case tok::lessequal:
112    case tok::lesslessequal:
113    case tok::spaceship:
114    case tok::greater:
115    case tok::greatergreater:
116    case tok::greaterequal:
117    case tok::greatergreaterequal:
118    case tok::caret:
119    case tok::caretequal:
120    case tok::pipe:
121    case tok::pipepipe:
122    case tok::pipeequal:
123    case tok::question:
124    case tok::equal:
125    case tok::equalequal:
126      return options.operators;
127    default:
128      break;
129    }
130  return HighlightStyle::ColorStyle();
131}
132
133void ClangHighlighter::Highlight(const HighlightStyle &options,
134                                 llvm::StringRef line,
135                                 std::optional<size_t> cursor_pos,
136                                 llvm::StringRef previous_lines,
137                                 Stream &result) const {
138  using namespace clang;
139
140  FileSystemOptions file_opts;
141  FileManager file_mgr(file_opts,
142                       FileSystem::Instance().GetVirtualFileSystem());
143
144  // The line might end in a backslash which would cause Clang to drop the
145  // backslash and the terminating new line. This makes sense when parsing C++,
146  // but when highlighting we care about preserving the backslash/newline. To
147  // not lose this information we remove the new line here so that Clang knows
148  // this is just a single line we are highlighting. We add back the newline
149  // after tokenizing.
150  llvm::StringRef line_ending = "";
151  // There are a few legal line endings Clang recognizes and we need to
152  // temporarily remove from the string.
153  if (line.consume_back("\r\n"))
154    line_ending = "\r\n";
155  else if (line.consume_back("\n"))
156    line_ending = "\n";
157  else if (line.consume_back("\r"))
158    line_ending = "\r";
159
160  unsigned line_number = previous_lines.count('\n') + 1U;
161
162  // Let's build the actual source code Clang needs and setup some utility
163  // objects.
164  std::string full_source = previous_lines.str() + line.str();
165  llvm::IntrusiveRefCntPtr<DiagnosticIDs> diag_ids(new DiagnosticIDs());
166  llvm::IntrusiveRefCntPtr<DiagnosticOptions> diags_opts(
167      new DiagnosticOptions());
168  DiagnosticsEngine diags(diag_ids, diags_opts);
169  clang::SourceManager SM(diags, file_mgr);
170  auto buf = llvm::MemoryBuffer::getMemBuffer(full_source);
171
172  FileID FID = SM.createFileID(buf->getMemBufferRef());
173
174  // Let's just enable the latest ObjC and C++ which should get most tokens
175  // right.
176  LangOptions Opts;
177  Opts.ObjC = true;
178  // FIXME: This should probably set CPlusPlus, CPlusPlus11, ... too
179  Opts.CPlusPlus17 = true;
180  Opts.LineComment = true;
181
182  Lexer lex(FID, buf->getMemBufferRef(), SM, Opts);
183  // The lexer should keep whitespace around.
184  lex.SetKeepWhitespaceMode(true);
185
186  // Keeps track if we have entered a PP directive.
187  bool in_pp_directive = false;
188
189  // True once we actually lexed the user provided line.
190  bool found_user_line = false;
191
192  // True if we already highlighted the token under the cursor, false otherwise.
193  bool highlighted_cursor = false;
194  Token token;
195  bool exit = false;
196  while (!exit) {
197    // Returns true if this is the last token we get from the lexer.
198    exit = lex.LexFromRawLexer(token);
199
200    bool invalid = false;
201    unsigned current_line_number =
202        SM.getSpellingLineNumber(token.getLocation(), &invalid);
203    if (current_line_number != line_number)
204      continue;
205    found_user_line = true;
206
207    // We don't need to print any tokens without a spelling line number.
208    if (invalid)
209      continue;
210
211    // Same as above but with the column number.
212    invalid = false;
213    unsigned start = SM.getSpellingColumnNumber(token.getLocation(), &invalid);
214    if (invalid)
215      continue;
216    // Column numbers start at 1, but indexes in our string start at 0.
217    --start;
218
219    // Annotations don't have a length, so let's skip them.
220    if (token.isAnnotation())
221      continue;
222
223    // Extract the token string from our source code.
224    llvm::StringRef tok_str = line.substr(start, token.getLength());
225
226    // If the token is just an empty string, we can skip all the work below.
227    if (tok_str.empty())
228      continue;
229
230    // If the cursor is inside this token, we have to apply the 'selected'
231    // highlight style before applying the actual token color.
232    llvm::StringRef to_print = tok_str;
233    StreamString storage;
234    auto end = start + token.getLength();
235    if (cursor_pos && end > *cursor_pos && !highlighted_cursor) {
236      highlighted_cursor = true;
237      options.selected.Apply(storage, tok_str);
238      to_print = storage.GetString();
239    }
240
241    // See how we are supposed to highlight this token.
242    HighlightStyle::ColorStyle color =
243        determineClangStyle(*this, token, tok_str, options, in_pp_directive);
244
245    color.Apply(result, to_print);
246  }
247
248  // Add the line ending we trimmed before tokenizing.
249  result << line_ending;
250
251  // If we went over the whole file but couldn't find our own file, then
252  // somehow our setup was wrong. When we're in release mode we just give the
253  // user the normal line and pretend we don't know how to highlight it. In
254  // debug mode we bail out with an assert as this should never happen.
255  if (!found_user_line) {
256    result << line;
257    assert(false && "We couldn't find the user line in the input file?");
258  }
259}
260