1 //===- TGLexer.cpp - Lexer for TableGen -----------------------------------===//
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 // Implement the Lexer for TableGen.
12 //===----------------------------------------------------------------------===//
15 #include "llvm/Support/SourceMgr.h"
16 #include "llvm/Support/Streams.h"
17 #include "llvm/Support/MemoryBuffer.h"
19 #include "llvm/Config/config.h"
27 TGLexer::TGLexer(SourceMgr &SM) : SrcMgr(SM) {
29 CurBuf = SrcMgr.getMemoryBuffer(CurBuffer);
30 CurPtr = CurBuf->getBufferStart();
34 SMLoc TGLexer::getLoc() const {
35 return SMLoc::getFromPointer(TokStart);
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) {
47 void TGLexer::PrintError(const char *Loc, const std::string &Msg) const {
48 SrcMgr.PrintError(SMLoc::getFromPointer(Loc), Msg);
51 void TGLexer::PrintError(SMLoc Loc, const std::string &Msg) const {
52 SrcMgr.PrintError(Loc, Msg);
56 int TGLexer::getNextChar() {
57 char CurChar = *CurPtr++;
60 return (unsigned char)CurChar;
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.
67 // If this is the end of an included file, pop the parent file off the
69 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
70 if (ParentIncludeLoc != SMLoc()) {
71 CurBuffer = SrcMgr.FindBufferContainingLoc(ParentIncludeLoc);
72 CurBuf = SrcMgr.getMemoryBuffer(CurBuffer);
73 CurPtr = ParentIncludeLoc.getPointer();
77 // Otherwise, return end of file.
78 --CurPtr; // Another call to lex will return EOF again.
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')) &&
88 ++CurPtr; // Eat the two char newline sequence.
93 tgtok::TokKind TGLexer::LexToken() {
95 // This always consumes at least one character.
96 int CurChar = getNextChar();
100 // Handle letters: [a-zA-Z_]
101 if (isalpha(CurChar) || CurChar == '_' || CurChar == '#')
102 return LexIdentifier();
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;
126 // Ignore whitespace.
129 // If this is the start of a // comment, skip until the end of the line or
130 // the end of the buffer.
133 else if (*CurPtr == '*') {
136 } else // Otherwise, this is an error.
137 return ReturnError(TokStart, "Unexpected character");
140 case '0': case '1': case '2': case '3': case '4': case '5': case '6':
141 case '7': case '8': case '9':
143 case '"': return LexString();
144 case '$': return LexVarName();
145 case '[': return LexBracket();
146 case '!': return LexExclaim();
150 /// LexString - Lex "[^"]*"
151 tgtok::TokKind TGLexer::LexString() {
152 const char *StrStart = CurPtr;
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");
161 if (*CurPtr == '\n' || *CurPtr == '\r')
162 return ReturnError(StrStart, "End of line in string literal");
164 if (*CurPtr != '\\') {
165 CurStrVal += *CurPtr++;
172 case '\\': case '\'': case '"':
173 // These turn into their literal character.
174 CurStrVal += *CurPtr++;
187 return ReturnError(CurPtr, "escaped newlines not supported in tblgen");
189 // If we hit the end of the buffer, report an error.
191 if (CurPtr == CurBuf->getBufferEnd())
192 return ReturnError(StrStart, "End of file in string literal");
195 return ReturnError(CurPtr, "invalid escape in string literal");
200 return tgtok::StrVal;
203 tgtok::TokKind TGLexer::LexVarName() {
204 if (!isalpha(CurPtr[0]) && CurPtr[0] != '_')
205 return ReturnError(TokStart, "Invalid variable name");
207 // Otherwise, we're ok, consume the rest of the characters.
208 const char *VarNameStart = CurPtr++;
210 while (isalpha(*CurPtr) || isdigit(*CurPtr) || *CurPtr == '_')
213 CurStrVal.assign(VarNameStart, CurPtr);
214 return tgtok::VarName;
218 tgtok::TokKind TGLexer::LexIdentifier() {
219 // The first letter is [a-zA-Z_].
220 const char *IdentStart = TokStart;
222 // Match the rest of the identifier regex: [0-9a-zA-Z_]*
223 while (isalpha(*CurPtr) || isdigit(*CurPtr) || *CurPtr == '_'
225 // If this contains a '#', make sure it's value
226 if (*CurPtr == '#') {
227 if (strncmp(CurPtr, "#NAME#", 6) != 0) {
238 // Check to see if this identifier is a keyword.
239 unsigned Len = CurPtr-IdentStart;
241 if (Len == 3 && !memcmp(IdentStart, "int", 3)) return tgtok::Int;
242 if (Len == 3 && !memcmp(IdentStart, "bit", 3)) return tgtok::Bit;
243 if (Len == 4 && !memcmp(IdentStart, "bits", 4)) return tgtok::Bits;
244 if (Len == 6 && !memcmp(IdentStart, "string", 6)) return tgtok::String;
245 if (Len == 4 && !memcmp(IdentStart, "list", 4)) return tgtok::List;
246 if (Len == 4 && !memcmp(IdentStart, "code", 4)) return tgtok::Code;
247 if (Len == 3 && !memcmp(IdentStart, "dag", 3)) return tgtok::Dag;
249 if (Len == 5 && !memcmp(IdentStart, "class", 5)) return tgtok::Class;
250 if (Len == 3 && !memcmp(IdentStart, "def", 3)) return tgtok::Def;
251 if (Len == 4 && !memcmp(IdentStart, "defm", 4)) return tgtok::Defm;
252 if (Len == 10 && !memcmp(IdentStart, "multiclass", 10))
253 return tgtok::MultiClass;
254 if (Len == 5 && !memcmp(IdentStart, "field", 5)) return tgtok::Field;
255 if (Len == 3 && !memcmp(IdentStart, "let", 3)) return tgtok::Let;
256 if (Len == 2 && !memcmp(IdentStart, "in", 2)) return tgtok::In;
258 if (Len == 7 && !memcmp(IdentStart, "include", 7)) {
259 if (LexInclude()) return tgtok::Error;
263 CurStrVal.assign(IdentStart, CurPtr);
267 /// LexInclude - We just read the "include" token. Get the string token that
268 /// comes next and enter the include.
269 bool TGLexer::LexInclude() {
270 // The token after the include must be a string.
271 tgtok::TokKind Tok = LexToken();
272 if (Tok == tgtok::Error) return true;
273 if (Tok != tgtok::StrVal) {
274 PrintError(getLoc(), "Expected filename after include");
279 std::string Filename = CurStrVal;
282 CurBuffer = SrcMgr.AddIncludeFile(Filename, SMLoc::getFromPointer(CurPtr));
283 if (CurBuffer == -1) {
284 PrintError(getLoc(), "Could not find include file '" + Filename + "'");
288 // Save the line number and lex buffer of the includer.
289 CurBuf = SrcMgr.getMemoryBuffer(CurBuffer);
290 CurPtr = CurBuf->getBufferStart();
294 void TGLexer::SkipBCPLComment() {
295 ++CurPtr; // skip the second slash.
300 return; // Newline is end of comment.
302 // If this is the end of the buffer, end the comment.
303 if (CurPtr == CurBuf->getBufferEnd())
307 // Otherwise, skip the character.
312 /// SkipCComment - This skips C-style /**/ comments. The only difference from C
313 /// is that we allow nesting.
314 bool TGLexer::SkipCComment() {
315 ++CurPtr; // skip the star.
316 unsigned CommentDepth = 1;
319 int CurChar = getNextChar();
322 PrintError(TokStart, "Unterminated comment!");
325 // End of the comment?
326 if (CurPtr[0] != '/') break;
328 ++CurPtr; // End the */.
329 if (--CommentDepth == 0)
333 // Start of a nested comment?
334 if (CurPtr[0] != '*') break;
346 tgtok::TokKind TGLexer::LexNumber() {
347 if (CurPtr[-1] == '0') {
348 if (CurPtr[0] == 'x') {
350 const char *NumStart = CurPtr;
351 while (isxdigit(CurPtr[0]))
354 // Requires at least one hex digit.
355 if (CurPtr == NumStart)
356 return ReturnError(CurPtr-2, "Invalid hexadecimal number");
359 CurIntVal = strtoll(NumStart, 0, 16);
361 return ReturnError(CurPtr-2, "Invalid hexadecimal number");
362 if (errno == ERANGE) {
364 CurIntVal = (int64_t)strtoull(NumStart, 0, 16);
366 return ReturnError(CurPtr-2, "Invalid hexadecimal number");
368 return ReturnError(CurPtr-2, "Hexadecimal number out of range");
370 return tgtok::IntVal;
371 } else if (CurPtr[0] == 'b') {
373 const char *NumStart = CurPtr;
374 while (CurPtr[0] == '0' || CurPtr[0] == '1')
377 // Requires at least one binary digit.
378 if (CurPtr == NumStart)
379 return ReturnError(CurPtr-2, "Invalid binary number");
380 CurIntVal = strtoll(NumStart, 0, 2);
381 return tgtok::IntVal;
385 // Check for a sign without a digit.
386 if (!isdigit(CurPtr[0])) {
387 if (CurPtr[-1] == '-')
389 else if (CurPtr[-1] == '+')
393 while (isdigit(CurPtr[0]))
395 CurIntVal = strtoll(TokStart, 0, 10);
396 return tgtok::IntVal;
399 /// LexBracket - We just read '['. If this is a code block, return it,
400 /// otherwise return the bracket. Match: '[' and '[{ ( [^}]+ | }[^]] )* }]'
401 tgtok::TokKind TGLexer::LexBracket() {
402 if (CurPtr[0] != '{')
403 return tgtok::l_square;
405 const char *CodeStart = CurPtr;
407 int Char = getNextChar();
408 if (Char == EOF) break;
410 if (Char != '}') continue;
412 Char = getNextChar();
413 if (Char == EOF) break;
415 CurStrVal.assign(CodeStart, CurPtr-2);
416 return tgtok::CodeFragment;
420 return ReturnError(CodeStart-2, "Unterminated Code Block");
423 /// LexExclaim - Lex '!' and '![a-zA-Z]+'.
424 tgtok::TokKind TGLexer::LexExclaim() {
425 if (!isalpha(*CurPtr))
426 return ReturnError(CurPtr-1, "Invalid \"!operator\"");
428 const char *Start = CurPtr++;
429 while (isalpha(*CurPtr))
432 // Check to see which operator this is.
433 unsigned Len = CurPtr-Start;
435 if (Len == 3 && !memcmp(Start, "con", 3)) return tgtok::XConcat;
436 if (Len == 3 && !memcmp(Start, "sra", 3)) return tgtok::XSRA;
437 if (Len == 3 && !memcmp(Start, "srl", 3)) return tgtok::XSRL;
438 if (Len == 3 && !memcmp(Start, "shl", 3)) return tgtok::XSHL;
439 if (Len == 9 && !memcmp(Start, "strconcat", 9)) return tgtok::XStrConcat;
440 if (Len == 10 && !memcmp(Start, "nameconcat", 10)) return tgtok::XNameConcat;
441 if (Len == 5 && !memcmp(Start, "subst", 5)) return tgtok::XSubst;
442 if (Len == 7 && !memcmp(Start, "foreach", 7)) return tgtok::XForEach;
443 if (Len == 4 && !memcmp(Start, "cast", 4)) return tgtok::XCast;
444 if (Len == 3 && !memcmp(Start, "car", 3)) return tgtok::XCar;
445 if (Len == 3 && !memcmp(Start, "cdr", 3)) return tgtok::XCdr;
446 if (Len == 4 && !memcmp(Start, "null", 4)) return tgtok::XNull;
447 if (Len == 2 && !memcmp(Start, "if", 2)) return tgtok::XIf;
449 return ReturnError(Start-1, "Unknown operator");