1 //===--- TokenLexer.cpp - Lex from a token stream -------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements the TokenLexer interface.
12 //===----------------------------------------------------------------------===//
14 #include "clang/Lex/MacroArgs.h"
15 #include "clang/Lex/LexDiagnostic.h"
16 #include "clang/Lex/MacroInfo.h"
17 #include "clang/Lex/Preprocessor.h"
18 #include "llvm/ADT/SmallString.h"
19 #include "llvm/Support/SaveAndRestore.h"
22 using namespace clang;
24 /// MacroArgs ctor function - This destroys the vector passed in.
25 MacroArgs *MacroArgs::create(const MacroInfo *MI,
26 ArrayRef<Token> UnexpArgTokens,
27 bool VarargsElided, Preprocessor &PP) {
28 assert(MI->isFunctionLike() &&
29 "Can't have args for an object-like macro!");
30 MacroArgs **ResultEnt = 0;
31 unsigned ClosestMatch = ~0U;
33 // See if we have an entry with a big enough argument list to reuse on the
34 // free list. If so, reuse it.
35 for (MacroArgs **Entry = &PP.MacroArgCache; *Entry;
36 Entry = &(*Entry)->ArgCache)
37 if ((*Entry)->NumUnexpArgTokens >= UnexpArgTokens.size() &&
38 (*Entry)->NumUnexpArgTokens < ClosestMatch) {
41 // If we have an exact match, use it.
42 if ((*Entry)->NumUnexpArgTokens == UnexpArgTokens.size())
44 // Otherwise, use the best fit.
45 ClosestMatch = (*Entry)->NumUnexpArgTokens;
50 // Allocate memory for a MacroArgs object with the lexer tokens at the end.
51 Result = (MacroArgs*)malloc(sizeof(MacroArgs) +
52 UnexpArgTokens.size() * sizeof(Token));
53 // Construct the MacroArgs object.
54 new (Result) MacroArgs(UnexpArgTokens.size(), VarargsElided);
57 // Unlink this node from the preprocessors singly linked list.
58 *ResultEnt = Result->ArgCache;
59 Result->NumUnexpArgTokens = UnexpArgTokens.size();
60 Result->VarargsElided = VarargsElided;
63 // Copy the actual unexpanded tokens to immediately after the result ptr.
64 if (!UnexpArgTokens.empty())
65 std::copy(UnexpArgTokens.begin(), UnexpArgTokens.end(),
66 const_cast<Token*>(Result->getUnexpArgument(0)));
71 /// destroy - Destroy and deallocate the memory for this object.
73 void MacroArgs::destroy(Preprocessor &PP) {
74 StringifiedArgs.clear();
76 // Don't clear PreExpArgTokens, just clear the entries. Clearing the entries
77 // would deallocate the element vectors.
78 for (unsigned i = 0, e = PreExpArgTokens.size(); i != e; ++i)
79 PreExpArgTokens[i].clear();
81 // Add this to the preprocessor's free list.
82 ArgCache = PP.MacroArgCache;
83 PP.MacroArgCache = this;
86 /// deallocate - This should only be called by the Preprocessor when managing
88 MacroArgs *MacroArgs::deallocate() {
89 MacroArgs *Next = ArgCache;
91 // Run the dtor to deallocate the vectors.
93 // Release the memory for the object.
100 /// getArgLength - Given a pointer to an expanded or unexpanded argument,
101 /// return the number of tokens, not counting the EOF, that make up the
103 unsigned MacroArgs::getArgLength(const Token *ArgPtr) {
104 unsigned NumArgTokens = 0;
105 for (; ArgPtr->isNot(tok::eof); ++ArgPtr)
111 /// getUnexpArgument - Return the unexpanded tokens for the specified formal.
113 const Token *MacroArgs::getUnexpArgument(unsigned Arg) const {
114 // The unexpanded argument tokens start immediately after the MacroArgs object
116 const Token *Start = (const Token *)(this+1);
117 const Token *Result = Start;
119 for (; Arg; ++Result) {
120 assert(Result < Start+NumUnexpArgTokens && "Invalid arg #");
121 if (Result->is(tok::eof))
124 assert(Result < Start+NumUnexpArgTokens && "Invalid arg #");
129 /// ArgNeedsPreexpansion - If we can prove that the argument won't be affected
130 /// by pre-expansion, return false. Otherwise, conservatively return true.
131 bool MacroArgs::ArgNeedsPreexpansion(const Token *ArgTok,
132 Preprocessor &PP) const {
133 // If there are no identifiers in the argument list, or if the identifiers are
134 // known to not be macros, pre-expansion won't modify it.
135 for (; ArgTok->isNot(tok::eof); ++ArgTok)
136 if (IdentifierInfo *II = ArgTok->getIdentifierInfo()) {
137 if (II->hasMacroDefinition() && PP.getMacroInfo(II)->isEnabled())
138 // Return true even though the macro could be a function-like macro
139 // without a following '(' token.
145 /// getPreExpArgument - Return the pre-expanded form of the specified
147 const std::vector<Token> &
148 MacroArgs::getPreExpArgument(unsigned Arg, const MacroInfo *MI,
150 assert(Arg < MI->getNumArgs() && "Invalid argument number!");
152 // If we have already computed this, return it.
153 if (PreExpArgTokens.size() < MI->getNumArgs())
154 PreExpArgTokens.resize(MI->getNumArgs());
156 std::vector<Token> &Result = PreExpArgTokens[Arg];
157 if (!Result.empty()) return Result;
159 SaveAndRestore<bool> PreExpandingMacroArgs(PP.InMacroArgPreExpansion, true);
161 const Token *AT = getUnexpArgument(Arg);
162 unsigned NumToks = getArgLength(AT)+1; // Include the EOF.
164 // Otherwise, we have to pre-expand this argument, populating Result. To do
165 // this, we set up a fake TokenLexer to lex from the unexpanded argument
166 // list. With this installed, we lex expanded tokens until we hit the EOF
167 // token at the end of the unexp list.
168 PP.EnterTokenStream(AT, NumToks, false /*disable expand*/,
169 false /*owns tokens*/);
171 // Lex all of the macro-expanded tokens into Result.
173 Result.push_back(Token());
174 Token &Tok = Result.back();
176 } while (Result.back().isNot(tok::eof));
178 // Pop the token stream off the top of the stack. We know that the internal
179 // pointer inside of it is to the "end" of the token stream, but the stack
180 // will not otherwise be popped until the next token is lexed. The problem is
181 // that the token may be lexed sometime after the vector of tokens itself is
182 // destroyed, which would be badness.
183 if (PP.InCachingLexMode())
184 PP.ExitCachingLexMode();
185 PP.RemoveTopOfLexerStack();
190 /// StringifyArgument - Implement C99 6.10.3.2p2, converting a sequence of
191 /// tokens into the literal string token that should be produced by the C #
192 /// preprocessor operator. If Charify is true, then it should be turned into
193 /// a character literal for the Microsoft charize (#@) extension.
195 Token MacroArgs::StringifyArgument(const Token *ArgToks,
196 Preprocessor &PP, bool Charify,
197 SourceLocation ExpansionLocStart,
198 SourceLocation ExpansionLocEnd) {
201 Tok.setKind(Charify ? tok::char_constant : tok::string_literal);
203 const Token *ArgTokStart = ArgToks;
205 // Stringify all the tokens.
206 SmallString<128> Result;
210 for (; ArgToks->isNot(tok::eof); ++ArgToks) {
211 const Token &Tok = *ArgToks;
212 if (!isFirst && (Tok.hasLeadingSpace() || Tok.isAtStartOfLine()))
216 // If this is a string or character constant, escape the token as specified
218 if (tok::isStringLiteral(Tok.getKind()) || // "foo", u8R"x(foo)x"_bar, etc.
219 Tok.is(tok::char_constant) || // 'x'
220 Tok.is(tok::wide_char_constant) || // L'x'.
221 Tok.is(tok::utf16_char_constant) || // u'x'.
222 Tok.is(tok::utf32_char_constant)) { // U'x'.
223 bool Invalid = false;
224 std::string TokStr = PP.getSpelling(Tok, &Invalid);
226 std::string Str = Lexer::Stringify(TokStr);
227 Result.append(Str.begin(), Str.end());
229 } else if (Tok.is(tok::code_completion)) {
230 PP.CodeCompleteNaturalLanguage();
232 // Otherwise, just append the token. Do some gymnastics to get the token
233 // in place and avoid copies where possible.
234 unsigned CurStrLen = Result.size();
235 Result.resize(CurStrLen+Tok.getLength());
236 const char *BufPtr = &Result[CurStrLen];
237 bool Invalid = false;
238 unsigned ActualTokLen = PP.getSpelling(Tok, BufPtr, &Invalid);
241 // If getSpelling returned a pointer to an already uniqued version of
242 // the string instead of filling in BufPtr, memcpy it onto our string.
243 if (BufPtr != &Result[CurStrLen])
244 memcpy(&Result[CurStrLen], BufPtr, ActualTokLen);
246 // If the token was dirty, the spelling may be shorter than the token.
247 if (ActualTokLen != Tok.getLength())
248 Result.resize(CurStrLen+ActualTokLen);
253 // If the last character of the string is a \, and if it isn't escaped, this
254 // is an invalid string literal, diagnose it as specified in C99.
255 if (Result.back() == '\\') {
256 // Count the number of consequtive \ characters. If even, then they are
257 // just escaped backslashes, otherwise it's an error.
258 unsigned FirstNonSlash = Result.size()-2;
259 // Guaranteed to find the starting " if nothing else.
260 while (Result[FirstNonSlash] == '\\')
262 if ((Result.size()-1-FirstNonSlash) & 1) {
263 // Diagnose errors for things like: #define F(X) #X / F(\)
264 PP.Diag(ArgToks[-1], diag::pp_invalid_string_literal);
265 Result.pop_back(); // remove one of the \'s.
270 // If this is the charify operation and the result is not a legal character
271 // constant, diagnose it.
273 // First step, turn double quotes into single quotes:
275 Result[Result.size()-1] = '\'';
277 // Check for bogus character.
279 if (Result.size() == 3)
280 isBad = Result[1] == '\''; // ''' is not legal. '\' already fixed above.
282 isBad = (Result.size() != 4 || Result[1] != '\\'); // Not '\x'
285 PP.Diag(ArgTokStart[0], diag::err_invalid_character_to_charify);
286 Result = "' '"; // Use something arbitrary, but legal.
290 PP.CreateString(Result, Tok,
291 ExpansionLocStart, ExpansionLocEnd);
295 /// getStringifiedArgument - Compute, cache, and return the specified argument
296 /// that has been 'stringified' as required by the # operator.
297 const Token &MacroArgs::getStringifiedArgument(unsigned ArgNo,
299 SourceLocation ExpansionLocStart,
300 SourceLocation ExpansionLocEnd) {
301 assert(ArgNo < NumUnexpArgTokens && "Invalid argument number!");
302 if (StringifiedArgs.empty()) {
303 StringifiedArgs.resize(getNumArguments());
304 memset((void*)&StringifiedArgs[0], 0,
305 sizeof(StringifiedArgs[0])*getNumArguments());
307 if (StringifiedArgs[ArgNo].isNot(tok::string_literal))
308 StringifiedArgs[ArgNo] = StringifyArgument(getUnexpArgument(ArgNo), PP,
312 return StringifiedArgs[ArgNo];