summaryrefslogtreecommitdiffstats
path: root/lib/Lex/TokenConcatenation.cpp
blob: 51d2e2326fcaadf45fbce37678b945030804c237 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
//===--- TokenConcatenation.cpp - Token Concatenation Avoidance -----------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file implements the TokenConcatenation class.
//
//===----------------------------------------------------------------------===//

#include "clang/Lex/TokenConcatenation.h"
#include "clang/Lex/Preprocessor.h"
using namespace clang;


/// StartsWithL - Return true if the spelling of this token starts with 'L'.
bool TokenConcatenation::StartsWithL(const Token &Tok) const {
  if (!Tok.needsCleaning()) {
    SourceManager &SM = PP.getSourceManager();
    return *SM.getCharacterData(SM.getSpellingLoc(Tok.getLocation())) == 'L';
  }

  if (Tok.getLength() < 256) {
    char Buffer[256];
    const char *TokPtr = Buffer;
    PP.getSpelling(Tok, TokPtr);
    return TokPtr[0] == 'L';
  }

  return PP.getSpelling(Tok)[0] == 'L';
}

/// IsIdentifierL - Return true if the spelling of this token is literally
/// 'L'.
bool TokenConcatenation::IsIdentifierL(const Token &Tok) const {
  if (!Tok.needsCleaning()) {
    if (Tok.getLength() != 1)
      return false;
    SourceManager &SM = PP.getSourceManager();
    return *SM.getCharacterData(SM.getSpellingLoc(Tok.getLocation())) == 'L';
  }

  if (Tok.getLength() < 256) {
    char Buffer[256];
    const char *TokPtr = Buffer;
    if (PP.getSpelling(Tok, TokPtr) != 1)
      return false;
    return TokPtr[0] == 'L';
  }

  return PP.getSpelling(Tok) == "L";
}

TokenConcatenation::TokenConcatenation(Preprocessor &pp) : PP(pp) {
  memset(TokenInfo, 0, sizeof(TokenInfo));

  // These tokens have custom code in AvoidConcat.
  TokenInfo[tok::identifier      ] |= aci_custom;
  TokenInfo[tok::numeric_constant] |= aci_custom_firstchar;
  TokenInfo[tok::period          ] |= aci_custom_firstchar;
  TokenInfo[tok::amp             ] |= aci_custom_firstchar;
  TokenInfo[tok::plus            ] |= aci_custom_firstchar;
  TokenInfo[tok::minus           ] |= aci_custom_firstchar;
  TokenInfo[tok::slash           ] |= aci_custom_firstchar;
  TokenInfo[tok::less            ] |= aci_custom_firstchar;
  TokenInfo[tok::greater         ] |= aci_custom_firstchar;
  TokenInfo[tok::pipe            ] |= aci_custom_firstchar;
  TokenInfo[tok::percent         ] |= aci_custom_firstchar;
  TokenInfo[tok::colon           ] |= aci_custom_firstchar;
  TokenInfo[tok::hash            ] |= aci_custom_firstchar;
  TokenInfo[tok::arrow           ] |= aci_custom_firstchar;

  // These tokens change behavior if followed by an '='.
  TokenInfo[tok::amp         ] |= aci_avoid_equal;           // &=
  TokenInfo[tok::plus        ] |= aci_avoid_equal;           // +=
  TokenInfo[tok::minus       ] |= aci_avoid_equal;           // -=
  TokenInfo[tok::slash       ] |= aci_avoid_equal;           // /=
  TokenInfo[tok::less        ] |= aci_avoid_equal;           // <=
  TokenInfo[tok::greater     ] |= aci_avoid_equal;           // >=
  TokenInfo[tok::pipe        ] |= aci_avoid_equal;           // |=
  TokenInfo[tok::percent     ] |= aci_avoid_equal;           // %=
  TokenInfo[tok::star        ] |= aci_avoid_equal;           // *=
  TokenInfo[tok::exclaim     ] |= aci_avoid_equal;           // !=
  TokenInfo[tok::lessless    ] |= aci_avoid_equal;           // <<=
  TokenInfo[tok::greatergreater] |= aci_avoid_equal;         // >>=
  TokenInfo[tok::caret       ] |= aci_avoid_equal;           // ^=
  TokenInfo[tok::equal       ] |= aci_avoid_equal;           // ==
}

/// GetFirstChar - Get the first character of the token \arg Tok,
/// avoiding calls to getSpelling where possible.
static char GetFirstChar(Preprocessor &PP, const Token &Tok) {
  if (IdentifierInfo *II = Tok.getIdentifierInfo()) {
    // Avoid spelling identifiers, the most common form of token.
    return II->getNameStart()[0];
  } else if (!Tok.needsCleaning()) {
    if (Tok.isLiteral() && Tok.getLiteralData()) {
      return *Tok.getLiteralData();
    } else {
      SourceManager &SM = PP.getSourceManager();
      return *SM.getCharacterData(SM.getSpellingLoc(Tok.getLocation()));
    }
  } else if (Tok.getLength() < 256) {
    char Buffer[256];
    const char *TokPtr = Buffer;
    PP.getSpelling(Tok, TokPtr);
    return TokPtr[0];
  } else {
    return PP.getSpelling(Tok)[0];
  }
}

/// AvoidConcat - If printing PrevTok immediately followed by Tok would cause
/// the two individual tokens to be lexed as a single token, return true
/// (which causes a space to be printed between them).  This allows the output
/// of -E mode to be lexed to the same token stream as lexing the input
/// directly would.
///
/// This code must conservatively return true if it doesn't want to be 100%
/// accurate.  This will cause the output to include extra space characters,
/// but the resulting output won't have incorrect concatenations going on.
/// Examples include "..", which we print with a space between, because we
/// don't want to track enough to tell "x.." from "...".
bool TokenConcatenation::AvoidConcat(const Token &PrevTok,
                                     const Token &Tok) const {
  // First, check to see if the tokens were directly adjacent in the original
  // source.  If they were, it must be okay to stick them together: if there
  // were an issue, the tokens would have been lexed differently.
  if (PrevTok.getLocation().isFileID() && Tok.getLocation().isFileID() &&
      PrevTok.getLocation().getFileLocWithOffset(PrevTok.getLength()) ==
        Tok.getLocation())
    return false;

  tok::TokenKind PrevKind = PrevTok.getKind();
  if (PrevTok.getIdentifierInfo())  // Language keyword or named operator.
    PrevKind = tok::identifier;

  // Look up information on when we should avoid concatenation with prevtok.
  unsigned ConcatInfo = TokenInfo[PrevKind];

  // If prevtok never causes a problem for anything after it, return quickly.
  if (ConcatInfo == 0) return false;

  if (ConcatInfo & aci_avoid_equal) {
    // If the next token is '=' or '==', avoid concatenation.
    if (Tok.is(tok::equal) || Tok.is(tok::equalequal))
      return true;
    ConcatInfo &= ~aci_avoid_equal;
  }

  if (ConcatInfo == 0) return false;

  // Basic algorithm: we look at the first character of the second token, and
  // determine whether it, if appended to the first token, would form (or
  // would contribute) to a larger token if concatenated.
  char FirstChar = 0;
  if (ConcatInfo & aci_custom) {
    // If the token does not need to know the first character, don't get it.
  } else {
    FirstChar = GetFirstChar(PP, Tok);
  }

  switch (PrevKind) {
  default: assert(0 && "InitAvoidConcatTokenInfo built wrong");
  case tok::identifier:   // id+id or id+number or id+L"foo".
    // id+'.'... will not append.
    if (Tok.is(tok::numeric_constant))
      return GetFirstChar(PP, Tok) != '.';

    if (Tok.getIdentifierInfo() || Tok.is(tok::wide_string_literal) /* ||
     Tok.is(tok::wide_char_literal)*/)
      return true;

    // If this isn't identifier + string, we're done.
    if (Tok.isNot(tok::char_constant) && Tok.isNot(tok::string_literal))
      return false;

    // FIXME: need a wide_char_constant!

    // If the string was a wide string L"foo" or wide char L'f', it would
    // concat with the previous identifier into fooL"bar".  Avoid this.
    if (StartsWithL(Tok))
      return true;

    // Otherwise, this is a narrow character or string.  If the *identifier*
    // is a literal 'L', avoid pasting L "foo" -> L"foo".
    return IsIdentifierL(PrevTok);
  case tok::numeric_constant:
    return isalnum(FirstChar) || Tok.is(tok::numeric_constant) ||
    FirstChar == '+' || FirstChar == '-' || FirstChar == '.';
  case tok::period:          // ..., .*, .1234
    return FirstChar == '.' || isdigit(FirstChar) ||
    (PP.getLangOptions().CPlusPlus && FirstChar == '*');
  case tok::amp:             // &&
    return FirstChar == '&';
  case tok::plus:            // ++
    return FirstChar == '+';
  case tok::minus:           // --, ->, ->*
    return FirstChar == '-' || FirstChar == '>';
  case tok::slash:           //, /*, //
    return FirstChar == '*' || FirstChar == '/';
  case tok::less:            // <<, <<=, <:, <%
    return FirstChar == '<' || FirstChar == ':' || FirstChar == '%';
  case tok::greater:         // >>, >>=
    return FirstChar == '>';
  case tok::pipe:            // ||
    return FirstChar == '|';
  case tok::percent:         // %>, %:
    return FirstChar == '>' || FirstChar == ':';
  case tok::colon:           // ::, :>
    return FirstChar == '>' ||
    (PP.getLangOptions().CPlusPlus && FirstChar == ':');
  case tok::hash:            // ##, #@, %:%:
    return FirstChar == '#' || FirstChar == '@' || FirstChar == '%';
  case tok::arrow:           // ->*
    return PP.getLangOptions().CPlusPlus && FirstChar == '*';
  }
}
OpenPOWER on IntegriCloud