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/Streams.h"
16 #include "llvm/Support/MemoryBuffer.h"
18 #include "llvm/Config/config.h"
25 TGLexer::TGLexer(MemoryBuffer *StartBuf) : CurLineNo(1), CurBuf(StartBuf) {
26 CurPtr = CurBuf->getBufferStart();
31 while (!IncludeStack.empty()) {
32 delete IncludeStack.back().Buffer;
33 IncludeStack.pop_back();
38 /// ReturnError - Set the error to the specified string at the specified
39 /// location. This is defined to always return tgtok::Error.
40 tgtok::TokKind TGLexer::ReturnError(const char *Loc, const std::string &Msg) {
45 void TGLexer::PrintIncludeStack(std::ostream &OS) const {
46 for (unsigned i = 0, e = IncludeStack.size(); i != e; ++i)
47 OS << "Included from " << IncludeStack[i].Buffer->getBufferIdentifier()
48 << ":" << IncludeStack[i].LineNo << ":\n";
49 OS << "Parsing " << CurBuf->getBufferIdentifier() << ":"
53 /// PrintError - Print the error at the specified location.
54 void TGLexer::PrintError(const char *ErrorLoc, const std::string &Msg) const {
55 PrintIncludeStack(*cerr.stream());
57 assert(ErrorLoc && "Location not specified!");
59 // Scan backward to find the start of the line.
60 const char *LineStart = ErrorLoc;
61 while (LineStart != CurBuf->getBufferStart() &&
62 LineStart[-1] != '\n' && LineStart[-1] != '\r')
64 // Get the end of the line.
65 const char *LineEnd = ErrorLoc;
66 while (LineEnd != CurBuf->getBufferEnd() &&
67 LineEnd[0] != '\n' && LineEnd[0] != '\r')
69 // Print out the line.
70 cerr << std::string(LineStart, LineEnd) << "\n";
71 // Print out spaces before the carat.
72 for (const char *Pos = LineStart; Pos != ErrorLoc; ++Pos)
73 cerr << (*Pos == '\t' ? '\t' : ' ');
77 int TGLexer::getNextChar() {
78 char CurChar = *CurPtr++;
81 return (unsigned char)CurChar;
83 // A nul character in the stream is either the end of the current buffer or
84 // a random nul in the file. Disambiguate that here.
85 if (CurPtr-1 != CurBuf->getBufferEnd())
86 return 0; // Just whitespace.
88 // If this is the end of an included file, pop the parent file off the
90 if (!IncludeStack.empty()) {
92 CurBuf = IncludeStack.back().Buffer;
93 CurLineNo = IncludeStack.back().LineNo;
94 CurPtr = IncludeStack.back().CurPtr;
95 IncludeStack.pop_back();
99 // Otherwise, return end of file.
100 --CurPtr; // Another call to lex will return EOF again.
104 // Handle the newline character by ignoring it and incrementing the line
105 // count. However, be careful about 'dos style' files with \n\r in them.
106 // Only treat a \n\r or \r\n as a single line.
107 if ((*CurPtr == '\n' || (*CurPtr == '\r')) &&
109 ++CurPtr; // Eat the two char newline sequence.
116 tgtok::TokKind TGLexer::LexToken() {
118 // This always consumes at least one character.
119 int CurChar = getNextChar();
123 // Handle letters: [a-zA-Z_]
124 if (isalpha(CurChar) || CurChar == '_')
125 return LexIdentifier();
127 // Unknown character, emit an error.
128 return ReturnError(TokStart, "Unexpected character");
129 case EOF: return tgtok::Eof;
130 case ':': return tgtok::colon;
131 case ';': return tgtok::semi;
132 case '.': return tgtok::period;
133 case ',': return tgtok::comma;
134 case '<': return tgtok::less;
135 case '>': return tgtok::greater;
136 case ']': return tgtok::r_square;
137 case '{': return tgtok::l_brace;
138 case '}': return tgtok::r_brace;
139 case '(': return tgtok::l_paren;
140 case ')': return tgtok::r_paren;
141 case '=': return tgtok::equal;
142 case '?': return tgtok::question;
149 // Ignore whitespace.
152 // If this is the start of a // comment, skip until the end of the line or
153 // the end of the buffer.
156 else if (*CurPtr == '*') {
159 } else // Otherwise, this is an error.
160 return ReturnError(TokStart, "Unexpected character");
163 case '0': case '1': case '2': case '3': case '4': case '5': case '6':
164 case '7': case '8': case '9':
166 case '"': return LexString();
167 case '$': return LexVarName();
168 case '[': return LexBracket();
169 case '!': return LexExclaim();
173 /// LexString - Lex "[^"]*"
174 tgtok::TokKind TGLexer::LexString() {
175 const char *StrStart = CurPtr;
177 while (*CurPtr != '"') {
178 // If we hit the end of the buffer, report an error.
179 if (*CurPtr == 0 && CurPtr == CurBuf->getBufferEnd())
180 return ReturnError(StrStart, "End of file in string literal");
182 if (*CurPtr == '\n' || *CurPtr == '\r')
183 return ReturnError(StrStart, "End of line in string literal");
188 CurStrVal.assign(StrStart, CurPtr);
190 return tgtok::StrVal;
193 tgtok::TokKind TGLexer::LexVarName() {
194 if (!isalpha(CurPtr[0]) && CurPtr[0] != '_')
195 return ReturnError(TokStart, "Invalid variable name");
197 // Otherwise, we're ok, consume the rest of the characters.
198 const char *VarNameStart = CurPtr++;
200 while (isalpha(*CurPtr) || isdigit(*CurPtr) || *CurPtr == '_')
203 CurStrVal.assign(VarNameStart, CurPtr);
204 return tgtok::VarName;
208 tgtok::TokKind TGLexer::LexIdentifier() {
209 // The first letter is [a-zA-Z_].
210 const char *IdentStart = TokStart;
212 // Match the rest of the identifier regex: [0-9a-zA-Z_]*
213 while (isalpha(*CurPtr) || isdigit(*CurPtr) || *CurPtr == '_')
216 // Check to see if this identifier is a keyword.
217 unsigned Len = CurPtr-IdentStart;
219 if (Len == 3 && !memcmp(IdentStart, "int", 3)) return tgtok::Int;
220 if (Len == 3 && !memcmp(IdentStart, "bit", 3)) return tgtok::Bit;
221 if (Len == 4 && !memcmp(IdentStart, "bits", 4)) return tgtok::Bits;
222 if (Len == 6 && !memcmp(IdentStart, "string", 6)) return tgtok::String;
223 if (Len == 4 && !memcmp(IdentStart, "list", 4)) return tgtok::List;
224 if (Len == 4 && !memcmp(IdentStart, "code", 4)) return tgtok::Code;
225 if (Len == 3 && !memcmp(IdentStart, "dag", 3)) return tgtok::Dag;
227 if (Len == 5 && !memcmp(IdentStart, "class", 5)) return tgtok::Class;
228 if (Len == 3 && !memcmp(IdentStart, "def", 3)) return tgtok::Def;
229 if (Len == 4 && !memcmp(IdentStart, "defm", 4)) return tgtok::Defm;
230 if (Len == 10 && !memcmp(IdentStart, "multiclass", 10))
231 return tgtok::MultiClass;
232 if (Len == 5 && !memcmp(IdentStart, "field", 5)) return tgtok::Field;
233 if (Len == 3 && !memcmp(IdentStart, "let", 3)) return tgtok::Let;
234 if (Len == 2 && !memcmp(IdentStart, "in", 2)) return tgtok::In;
236 if (Len == 7 && !memcmp(IdentStart, "include", 7)) {
237 if (LexInclude()) return tgtok::Error;
241 CurStrVal.assign(IdentStart, CurPtr);
245 /// LexInclude - We just read the "include" token. Get the string token that
246 /// comes next and enter the include.
247 bool TGLexer::LexInclude() {
248 // The token after the include must be a string.
249 tgtok::TokKind Tok = LexToken();
250 if (Tok == tgtok::Error) return true;
251 if (Tok != tgtok::StrVal) {
252 PrintError(getLoc(), "Expected filename after include");
257 std::string Filename = CurStrVal;
259 // Try to find the file.
260 MemoryBuffer *NewBuf = MemoryBuffer::getFile(Filename.c_str());
262 // If the file didn't exist directly, see if it's in an include path.
263 for (unsigned i = 0, e = IncludeDirectories.size(); i != e && !NewBuf; ++i) {
264 std::string IncFile = IncludeDirectories[i] + "/" + Filename;
265 NewBuf = MemoryBuffer::getFile(IncFile.c_str());
269 PrintError(getLoc(), "Could not find include file '" + Filename + "'");
273 // Save the line number and lex buffer of the includer.
274 IncludeStack.push_back(IncludeRec(CurBuf, CurPtr, CurLineNo));
276 CurLineNo = 1; // Reset line numbering.
278 CurPtr = CurBuf->getBufferStart();
282 void TGLexer::SkipBCPLComment() {
283 ++CurPtr; // skip the second slash.
288 return; // Newline is end of comment.
290 // If this is the end of the buffer, end the comment.
291 if (CurPtr == CurBuf->getBufferEnd())
295 // Otherwise, skip the character.
300 /// SkipCComment - This skips C-style /**/ comments. The only difference from C
301 /// is that we allow nesting.
302 bool TGLexer::SkipCComment() {
303 ++CurPtr; // skip the star.
304 unsigned CommentDepth = 1;
307 int CurChar = getNextChar();
310 PrintError(TokStart, "Unterminated comment!");
313 // End of the comment?
314 if (CurPtr[0] != '/') break;
316 ++CurPtr; // End the */.
317 if (--CommentDepth == 0)
321 // Start of a nested comment?
322 if (CurPtr[0] != '*') break;
334 tgtok::TokKind TGLexer::LexNumber() {
335 if (CurPtr[-1] == '0') {
336 if (CurPtr[0] == 'x') {
338 const char *NumStart = CurPtr;
339 while (isxdigit(CurPtr[0]))
342 // Requires at least one hex digit.
343 if (CurPtr == NumStart)
344 return ReturnError(CurPtr-2, "Invalid hexadecimal number");
346 CurIntVal = strtoll(NumStart, 0, 16);
347 return tgtok::IntVal;
348 } else if (CurPtr[0] == 'b') {
350 const char *NumStart = CurPtr;
351 while (CurPtr[0] == '0' || CurPtr[0] == '1')
354 // Requires at least one binary digit.
355 if (CurPtr == NumStart)
356 return ReturnError(CurPtr-2, "Invalid binary number");
357 CurIntVal = strtoll(NumStart, 0, 2);
358 return tgtok::IntVal;
362 // Check for a sign without a digit.
363 if (!isdigit(CurPtr[0])) {
364 if (CurPtr[-1] == '-')
366 else if (CurPtr[-1] == '+')
370 while (isdigit(CurPtr[0]))
372 CurIntVal = strtoll(TokStart, 0, 10);
373 return tgtok::IntVal;
376 /// LexBracket - We just read '['. If this is a code block, return it,
377 /// otherwise return the bracket. Match: '[' and '[{ ( [^}]+ | }[^]] )* }]'
378 tgtok::TokKind TGLexer::LexBracket() {
379 if (CurPtr[0] != '{')
380 return tgtok::l_square;
382 const char *CodeStart = CurPtr;
384 int Char = getNextChar();
385 if (Char == EOF) break;
387 if (Char != '}') continue;
389 Char = getNextChar();
390 if (Char == EOF) break;
392 CurStrVal.assign(CodeStart, CurPtr-2);
393 return tgtok::CodeFragment;
397 return ReturnError(CodeStart-2, "Unterminated Code Block");
400 /// LexExclaim - Lex '!' and '![a-zA-Z]+'.
401 tgtok::TokKind TGLexer::LexExclaim() {
402 if (!isalpha(*CurPtr))
403 return ReturnError(CurPtr-1, "Invalid \"!operator\"");
405 const char *Start = CurPtr++;
406 while (isalpha(*CurPtr))
409 // Check to see which operator this is.
410 unsigned Len = CurPtr-Start;
412 if (Len == 3 && !memcmp(Start, "con", 3)) return tgtok::XConcat;
413 if (Len == 3 && !memcmp(Start, "sra", 3)) return tgtok::XSRA;
414 if (Len == 3 && !memcmp(Start, "srl", 3)) return tgtok::XSRL;
415 if (Len == 3 && !memcmp(Start, "shl", 3)) return tgtok::XSHL;
416 if (Len == 9 && !memcmp(Start, "strconcat", 9)) return tgtok::XStrConcat;
418 return ReturnError(Start-1, "Unknown operator");