utils/TableGen/TGLexer.cpp

   1 //===- TGLexer.cpp - Lexer for TableGen -----------------------------------===//
   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 // Implement the Lexer for TableGen.
  11 //
  12 //===----------------------------------------------------------------------===//
  13
  14 #include "TGLexer.h"
  15 #include "TGSourceMgr.h"
  16 #include "llvm/Support/Streams.h"
  17 #include "llvm/Support/MemoryBuffer.h"
  18 #include <ostream>
  19 #include "llvm/Config/config.h"
  20 #include <cctype>
  21 #include <cstdio>
  22 #include <cstdlib>
  23 #include <cstring>
  24 #include <cerrno>
  25 using namespace llvm;
  26
  27 TGLexer::TGLexer(TGSourceMgr &SM) : SrcMgr(SM) {
  28   CurBuffer = 0;
  29   CurBuf = SrcMgr.getMemoryBuffer(CurBuffer);
  30   CurPtr = CurBuf->getBufferStart();
  31   TokStart = 0;
  32 }
  33
  34 TGLoc TGLexer::getLoc() const {
  35   return TGLoc::getFromPointer(TokStart);
  36 }
  37
  38
  39 /// ReturnError - Set the error to the specified string at the specified
  40 /// location.  This is defined to always return tgtok::Error.
  41 tgtok::TokKind TGLexer::ReturnError(const char *Loc, const std::string &Msg) {
  42   PrintError(Loc, Msg);
  43   return tgtok::Error;
  44 }
  45
  46
  47 void TGLexer::PrintError(const char *Loc, const std::string &Msg) const {
  48   SrcMgr.PrintError(TGLoc::getFromPointer(Loc), Msg);
  49 }
  50
  51 void TGLexer::PrintError(TGLoc Loc, const std::string &Msg) const {
  52   SrcMgr.PrintError(Loc, Msg);
  53 }
  54
  55
  56 int TGLexer::getNextChar() {
  57   char CurChar = *CurPtr++;
  58   switch (CurChar) {
  59   default:
  60     return (unsigned char)CurChar;
  61   case 0: {
  62     // A nul character in the stream is either the end of the current buffer or
  63     // a random nul in the file.  Disambiguate that here.
  64     if (CurPtr-1 != CurBuf->getBufferEnd())
  65       return 0;  // Just whitespace.
  66
  67     // If this is the end of an included file, pop the parent file off the
  68     // include stack.
  69     TGLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
  70     if (ParentIncludeLoc != TGLoc()) {
  71       CurBuffer = SrcMgr.FindBufferContainingLoc(ParentIncludeLoc);
  72       CurBuf = SrcMgr.getMemoryBuffer(CurBuffer);
  73       CurPtr = ParentIncludeLoc.getPointer();
  74       return getNextChar();
  75     }
  76
  77     // Otherwise, return end of file.
  78     --CurPtr;  // Another call to lex will return EOF again.
  79     return EOF;
  80   }
  81   case '\n':
  82   case '\r':
  83     // Handle the newline character by ignoring it and incrementing the line
  84     // count.  However, be careful about 'dos style' files with \n\r in them.
  85     // Only treat a \n\r or \r\n as a single line.
  86     if ((*CurPtr == '\n' || (*CurPtr == '\r')) &&
  87         *CurPtr != CurChar)
  88       ++CurPtr;  // Eat the two char newline sequence.
  89     return '\n';
  90   }
  91 }
  92
  93 tgtok::TokKind TGLexer::LexToken() {
  94   TokStart = CurPtr;
  95   // This always consumes at least one character.
  96   int CurChar = getNextChar();
  97
  98   switch (CurChar) {
  99   default:
 100     // Handle letters: [a-zA-Z_]
 101     if (isalpha(CurChar) || CurChar == '_')
 102       return LexIdentifier();
 103
 104     // Unknown character, emit an error.
 105     return ReturnError(TokStart, "Unexpected character");
 106   case EOF: return tgtok::Eof;
 107   case ':': return tgtok::colon;
 108   case ';': return tgtok::semi;
 109   case '.': return tgtok::period;
 110   case ',': return tgtok::comma;
 111   case '<': return tgtok::less;
 112   case '>': return tgtok::greater;
 113   case ']': return tgtok::r_square;
 114   case '{': return tgtok::l_brace;
 115   case '}': return tgtok::r_brace;
 116   case '(': return tgtok::l_paren;
 117   case ')': return tgtok::r_paren;
 118   case '=': return tgtok::equal;
 119   case '?': return tgtok::question;
 120
 121   case 0:
 122   case ' ':
 123   case '\t':
 124   case '\n':
 125   case '\r':
 126     // Ignore whitespace.
 127     return LexToken();
 128   case '/':
 129     // If this is the start of a // comment, skip until the end of the line or
 130     // the end of the buffer.
 131     if (*CurPtr == '/')
 132       SkipBCPLComment();
 133     else if (*CurPtr == '*') {
 134       if (SkipCComment())
 135         return tgtok::Error;
 136     } else // Otherwise, this is an error.
 137       return ReturnError(TokStart, "Unexpected character");
 138     return LexToken();
 139   case '-': case '+':
 140   case '0': case '1': case '2': case '3': case '4': case '5': case '6':
 141   case '7': case '8': case '9':
 142     return LexNumber();
 143   case '"': return LexString();
 144   case '$': return LexVarName();
 145   case '[': return LexBracket();
 146   case '!': return LexExclaim();
 147   }
 148 }
 149
 150 /// LexString - Lex "[^"]*"
 151 tgtok::TokKind TGLexer::LexString() {
 152   const char *StrStart = CurPtr;
 153
 154   CurStrVal = "";
 155
 156   while (*CurPtr != '"') {
 157     // If we hit the end of the buffer, report an error.
 158     if (*CurPtr == 0 && CurPtr == CurBuf->getBufferEnd())
 159       return ReturnError(StrStart, "End of file in string literal");
 160
 161     if (*CurPtr == '\n' || *CurPtr == '\r')
 162       return ReturnError(StrStart, "End of line in string literal");
 163
 164     if (*CurPtr != '\\') {
 165       CurStrVal += *CurPtr++;
 166       continue;
 167     }
 168
 169     ++CurPtr;
 170
 171     switch (*CurPtr) {
 172     case '\\': case '\'': case '"':
 173       // These turn into their literal character.
 174       CurStrVal += *CurPtr++;
 175       break;
 176     case 't':
 177       CurStrVal += '\t';
 178       ++CurPtr;
 179       break;
 180     case 'n':
 181       CurStrVal += '\n';
 182       ++CurPtr;
 183       break;
 184
 185     case '\n':
 186     case '\r':
 187       return ReturnError(CurPtr, "escaped newlines not supported in tblgen");
 188
 189     // If we hit the end of the buffer, report an error.
 190     case '\0':
 191       if (CurPtr == CurBuf->getBufferEnd())
 192         return ReturnError(StrStart, "End of file in string literal");
 193       // FALL THROUGH
 194     default:
 195       return ReturnError(CurPtr, "invalid escape in string literal");
 196     }
 197   }
 198
 199   ++CurPtr;
 200   return tgtok::StrVal;
 201 }
 202
 203 tgtok::TokKind TGLexer::LexVarName() {
 204   if (!isalpha(CurPtr[0]) && CurPtr[0] != '_')
 205     return ReturnError(TokStart, "Invalid variable name");
 206
 207   // Otherwise, we're ok, consume the rest of the characters.
 208   const char *VarNameStart = CurPtr++;
 209
 210   while (isalpha(*CurPtr) || isdigit(*CurPtr) || *CurPtr == '_')
 211     ++CurPtr;
 212
 213   CurStrVal.assign(VarNameStart, CurPtr);
 214   return tgtok::VarName;
 215 }
 216
 217
 218 tgtok::TokKind TGLexer::LexIdentifier() {
 219   // The first letter is [a-zA-Z_].
 220   const char *IdentStart = TokStart;
 221
 222   // Match the rest of the identifier regex: [0-9a-zA-Z_]*
 223   while (isalpha(*CurPtr) || isdigit(*CurPtr) || *CurPtr == '_')
 224     ++CurPtr;
 225
 226   // Check to see if this identifier is a keyword.
 227   unsigned Len = CurPtr-IdentStart;
 228
 229   if (Len == 3 && !memcmp(IdentStart, "int", 3)) return tgtok::Int;
 230   if (Len == 3 && !memcmp(IdentStart, "bit", 3)) return tgtok::Bit;
 231   if (Len == 4 && !memcmp(IdentStart, "bits", 4)) return tgtok::Bits;
 232   if (Len == 6 && !memcmp(IdentStart, "string", 6)) return tgtok::String;
 233   if (Len == 4 && !memcmp(IdentStart, "list", 4)) return tgtok::List;
 234   if (Len == 4 && !memcmp(IdentStart, "code", 4)) return tgtok::Code;
 235   if (Len == 3 && !memcmp(IdentStart, "dag", 3)) return tgtok::Dag;
 236
 237   if (Len == 5 && !memcmp(IdentStart, "class", 5)) return tgtok::Class;
 238   if (Len == 3 && !memcmp(IdentStart, "def", 3)) return tgtok::Def;
 239   if (Len == 4 && !memcmp(IdentStart, "defm", 4)) return tgtok::Defm;
 240   if (Len == 10 && !memcmp(IdentStart, "multiclass", 10))
 241     return tgtok::MultiClass;
 242   if (Len == 5 && !memcmp(IdentStart, "field", 5)) return tgtok::Field;
 243   if (Len == 3 && !memcmp(IdentStart, "let", 3)) return tgtok::Let;
 244   if (Len == 2 && !memcmp(IdentStart, "in", 2)) return tgtok::In;
 245
 246   if (Len == 7 && !memcmp(IdentStart, "include", 7)) {
 247     if (LexInclude()) return tgtok::Error;
 248     return Lex();
 249   }
 250
 251   CurStrVal.assign(IdentStart, CurPtr);
 252   return tgtok::Id;
 253 }
 254
 255 /// LexInclude - We just read the "include" token.  Get the string token that
 256 /// comes next and enter the include.
 257 bool TGLexer::LexInclude() {
 258   // The token after the include must be a string.
 259   tgtok::TokKind Tok = LexToken();
 260   if (Tok == tgtok::Error) return true;
 261   if (Tok != tgtok::StrVal) {
 262     PrintError(getLoc(), "Expected filename after include");
 263     return true;
 264   }
 265
 266   // Get the string.
 267   std::string Filename = CurStrVal;
 268
 269   // Try to find the file.
 270   MemoryBuffer *NewBuf = MemoryBuffer::getFile(Filename.c_str());
 271
 272   // If the file didn't exist directly, see if it's in an include path.
 273   for (unsigned i = 0, e = IncludeDirectories.size(); i != e && !NewBuf; ++i) {
 274     std::string IncFile = IncludeDirectories[i] + "/" + Filename;
 275     NewBuf = MemoryBuffer::getFile(IncFile.c_str());
 276   }
 277
 278   if (NewBuf == 0) {
 279     PrintError(getLoc(), "Could not find include file '" + Filename + "'");
 280     return true;
 281   }
 282
 283   // Save the line number and lex buffer of the includer.
 284   CurBuffer = SrcMgr.AddNewSourceBuffer(NewBuf, TGLoc::getFromPointer(CurPtr));
 285
 286   CurBuf = NewBuf;
 287   CurPtr = CurBuf->getBufferStart();
 288   return false;
 289 }
 290
 291 void TGLexer::SkipBCPLComment() {
 292   ++CurPtr;  // skip the second slash.
 293   while (1) {
 294     switch (*CurPtr) {
 295     case '\n':
 296     case '\r':
 297       return;  // Newline is end of comment.
 298     case 0:
 299       // If this is the end of the buffer, end the comment.
 300       if (CurPtr == CurBuf->getBufferEnd())
 301         return;
 302       break;
 303     }
 304     // Otherwise, skip the character.
 305     ++CurPtr;
 306   }
 307 }
 308
 309 /// SkipCComment - This skips C-style /**/ comments.  The only difference from C
 310 /// is that we allow nesting.
 311 bool TGLexer::SkipCComment() {
 312   ++CurPtr;  // skip the star.
 313   unsigned CommentDepth = 1;
 314
 315   while (1) {
 316     int CurChar = getNextChar();
 317     switch (CurChar) {
 318     case EOF:
 319       PrintError(TokStart, "Unterminated comment!");
 320       return true;
 321     case '*':
 322       // End of the comment?
 323       if (CurPtr[0] != '/') break;
 324
 325       ++CurPtr;   // End the */.
 326       if (--CommentDepth == 0)
 327         return false;
 328       break;
 329     case '/':
 330       // Start of a nested comment?
 331       if (CurPtr[0] != '*') break;
 332       ++CurPtr;
 333       ++CommentDepth;
 334       break;
 335     }
 336   }
 337 }
 338
 339 /// LexNumber - Lex:
 340 ///    [-+]?[0-9]+
 341 ///    0x[0-9a-fA-F]+
 342 ///    0b[01]+
 343 tgtok::TokKind TGLexer::LexNumber() {
 344   if (CurPtr[-1] == '0') {
 345     if (CurPtr[0] == 'x') {
 346       ++CurPtr;
 347       const char *NumStart = CurPtr;
 348       while (isxdigit(CurPtr[0]))
 349         ++CurPtr;
 350
 351       // Requires at least one hex digit.
 352       if (CurPtr == NumStart)
 353         return ReturnError(CurPtr-2, "Invalid hexadecimal number");
 354
 355       errno = 0;
 356       CurIntVal = strtoll(NumStart, 0, 16);
 357       if (errno == EINVAL)
 358         return ReturnError(CurPtr-2, "Invalid hexadecimal number");
 359       if (errno == ERANGE) {
 360         errno = 0;
 361         CurIntVal = (int64_t)strtoull(NumStart, 0, 16);
 362         if (errno == EINVAL)
 363           return ReturnError(CurPtr-2, "Invalid hexadecimal number");
 364         if (errno == ERANGE)
 365           return ReturnError(CurPtr-2, "Hexadecimal number out of range");
 366       }
 367       return tgtok::IntVal;
 368     } else if (CurPtr[0] == 'b') {
 369       ++CurPtr;
 370       const char *NumStart = CurPtr;
 371       while (CurPtr[0] == '0' || CurPtr[0] == '1')
 372         ++CurPtr;
 373
 374       // Requires at least one binary digit.
 375       if (CurPtr == NumStart)
 376         return ReturnError(CurPtr-2, "Invalid binary number");
 377       CurIntVal = strtoll(NumStart, 0, 2);
 378       return tgtok::IntVal;
 379     }
 380   }
 381
 382   // Check for a sign without a digit.
 383   if (!isdigit(CurPtr[0])) {
 384     if (CurPtr[-1] == '-')
 385       return tgtok::minus;
 386     else if (CurPtr[-1] == '+')
 387       return tgtok::plus;
 388   }
 389
 390   while (isdigit(CurPtr[0]))
 391     ++CurPtr;
 392   CurIntVal = strtoll(TokStart, 0, 10);
 393   return tgtok::IntVal;
 394 }
 395
 396 /// LexBracket - We just read '['.  If this is a code block, return it,
 397 /// otherwise return the bracket.  Match: '[' and '[{ ( [^}]+ | }[^]] )* }]'
 398 tgtok::TokKind TGLexer::LexBracket() {
 399   if (CurPtr[0] != '{')
 400     return tgtok::l_square;
 401   ++CurPtr;
 402   const char *CodeStart = CurPtr;
 403   while (1) {
 404     int Char = getNextChar();
 405     if (Char == EOF) break;
 406
 407     if (Char != '}') continue;
 408
 409     Char = getNextChar();
 410     if (Char == EOF) break;
 411     if (Char == ']') {
 412       CurStrVal.assign(CodeStart, CurPtr-2);
 413       return tgtok::CodeFragment;
 414     }
 415   }
 416
 417   return ReturnError(CodeStart-2, "Unterminated Code Block");
 418 }
 419
 420 /// LexExclaim - Lex '!' and '![a-zA-Z]+'.
 421 tgtok::TokKind TGLexer::LexExclaim() {
 422   if (!isalpha(*CurPtr))
 423     return ReturnError(CurPtr-1, "Invalid \"!operator\"");
 424
 425   const char *Start = CurPtr++;
 426   while (isalpha(*CurPtr))
 427     ++CurPtr;
 428
 429   // Check to see which operator this is.
 430   unsigned Len = CurPtr-Start;
 431
 432   if (Len == 3  && !memcmp(Start, "con", 3)) return tgtok::XConcat;
 433   if (Len == 3  && !memcmp(Start, "sra", 3)) return tgtok::XSRA;
 434   if (Len == 3  && !memcmp(Start, "srl", 3)) return tgtok::XSRL;
 435   if (Len == 3  && !memcmp(Start, "shl", 3)) return tgtok::XSHL;
 436   if (Len == 9  && !memcmp(Start, "strconcat", 9))   return tgtok::XStrConcat;
 437   if (Len == 10 && !memcmp(Start, "nameconcat", 10)) return tgtok::XNameConcat;
 438
 439   return ReturnError(Start-1, "Unknown operator");
 440 }
 441