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"
26 TGLexer::TGLexer(MemoryBuffer *StartBuf) : CurLineNo(1), CurBuf(StartBuf) {
27 CurPtr = CurBuf->getBufferStart();
32 while (!IncludeStack.empty()) {
33 delete IncludeStack.back().Buffer;
34 IncludeStack.pop_back();
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) {
46 void TGLexer::PrintIncludeStack(std::ostream &OS) const {
47 for (unsigned i = 0, e = IncludeStack.size(); i != e; ++i)
48 OS << "Included from " << IncludeStack[i].Buffer->getBufferIdentifier()
49 << ":" << IncludeStack[i].LineNo << ":\n";
50 OS << "Parsing " << CurBuf->getBufferIdentifier() << ":"
54 /// PrintError - Print the error at the specified location.
55 void TGLexer::PrintError(const char *ErrorLoc, const std::string &Msg) const {
56 PrintIncludeStack(*cerr.stream());
58 assert(ErrorLoc && "Location not specified!");
60 // Scan backward to find the start of the line.
61 const char *LineStart = ErrorLoc;
62 while (LineStart != CurBuf->getBufferStart() &&
63 LineStart[-1] != '\n' && LineStart[-1] != '\r')
65 // Get the end of the line.
66 const char *LineEnd = ErrorLoc;
67 while (LineEnd != CurBuf->getBufferEnd() &&
68 LineEnd[0] != '\n' && LineEnd[0] != '\r')
70 // Print out the line.
71 cerr << std::string(LineStart, LineEnd) << "\n";
72 // Print out spaces before the carat.
73 for (const char *Pos = LineStart; Pos != ErrorLoc; ++Pos)
74 cerr << (*Pos == '\t' ? '\t' : ' ');
78 int TGLexer::getNextChar() {
79 char CurChar = *CurPtr++;
82 return (unsigned char)CurChar;
84 // A nul character in the stream is either the end of the current buffer or
85 // a random nul in the file. Disambiguate that here.
86 if (CurPtr-1 != CurBuf->getBufferEnd())
87 return 0; // Just whitespace.
89 // If this is the end of an included file, pop the parent file off the
91 if (!IncludeStack.empty()) {
93 CurBuf = IncludeStack.back().Buffer;
94 CurLineNo = IncludeStack.back().LineNo;
95 CurPtr = IncludeStack.back().CurPtr;
96 IncludeStack.pop_back();
100 // Otherwise, return end of file.
101 --CurPtr; // Another call to lex will return EOF again.
105 // Handle the newline character by ignoring it and incrementing the line
106 // count. However, be careful about 'dos style' files with \n\r in them.
107 // Only treat a \n\r or \r\n as a single line.
108 if ((*CurPtr == '\n' || (*CurPtr == '\r')) &&
110 ++CurPtr; // Eat the two char newline sequence.
117 tgtok::TokKind TGLexer::LexToken() {
119 // This always consumes at least one character.
120 int CurChar = getNextChar();
124 // Handle letters: [a-zA-Z_]
125 if (isalpha(CurChar) || CurChar == '_')
126 return LexIdentifier();
128 // Unknown character, emit an error.
129 return ReturnError(TokStart, "Unexpected character");
130 case EOF: return tgtok::Eof;
131 case ':': return tgtok::colon;
132 case ';': return tgtok::semi;
133 case '.': return tgtok::period;
134 case ',': return tgtok::comma;
135 case '<': return tgtok::less;
136 case '>': return tgtok::greater;
137 case ']': return tgtok::r_square;
138 case '{': return tgtok::l_brace;
139 case '}': return tgtok::r_brace;
140 case '(': return tgtok::l_paren;
141 case ')': return tgtok::r_paren;
142 case '=': return tgtok::equal;
143 case '?': return tgtok::question;
150 // Ignore whitespace.
153 // If this is the start of a // comment, skip until the end of the line or
154 // the end of the buffer.
157 else if (*CurPtr == '*') {
160 } else // Otherwise, this is an error.
161 return ReturnError(TokStart, "Unexpected character");
164 case '0': case '1': case '2': case '3': case '4': case '5': case '6':
165 case '7': case '8': case '9':
167 case '"': return LexString();
168 case '$': return LexVarName();
169 case '[': return LexBracket();
170 case '!': return LexExclaim();
174 /// LexString - Lex "[^"]*"
175 tgtok::TokKind TGLexer::LexString() {
176 const char *StrStart = CurPtr;
178 while (*CurPtr != '"') {
179 // If we hit the end of the buffer, report an error.
180 if (*CurPtr == 0 && CurPtr == CurBuf->getBufferEnd())
181 return ReturnError(StrStart, "End of file in string literal");
183 if (*CurPtr == '\n' || *CurPtr == '\r')
184 return ReturnError(StrStart, "End of line in string literal");
189 CurStrVal.assign(StrStart, CurPtr);
191 return tgtok::StrVal;
194 tgtok::TokKind TGLexer::LexVarName() {
195 if (!isalpha(CurPtr[0]) && CurPtr[0] != '_')
196 return ReturnError(TokStart, "Invalid variable name");
198 // Otherwise, we're ok, consume the rest of the characters.
199 const char *VarNameStart = CurPtr++;
201 while (isalpha(*CurPtr) || isdigit(*CurPtr) || *CurPtr == '_')
204 CurStrVal.assign(VarNameStart, CurPtr);
205 return tgtok::VarName;
209 tgtok::TokKind TGLexer::LexIdentifier() {
210 // The first letter is [a-zA-Z_].
211 const char *IdentStart = TokStart;
213 // Match the rest of the identifier regex: [0-9a-zA-Z_]*
214 while (isalpha(*CurPtr) || isdigit(*CurPtr) || *CurPtr == '_')
217 // Check to see if this identifier is a keyword.
218 unsigned Len = CurPtr-IdentStart;
220 if (Len == 3 && !memcmp(IdentStart, "int", 3)) return tgtok::Int;
221 if (Len == 3 && !memcmp(IdentStart, "bit", 3)) return tgtok::Bit;
222 if (Len == 4 && !memcmp(IdentStart, "bits", 4)) return tgtok::Bits;
223 if (Len == 6 && !memcmp(IdentStart, "string", 6)) return tgtok::String;
224 if (Len == 4 && !memcmp(IdentStart, "list", 4)) return tgtok::List;
225 if (Len == 4 && !memcmp(IdentStart, "code", 4)) return tgtok::Code;
226 if (Len == 3 && !memcmp(IdentStart, "dag", 3)) return tgtok::Dag;
228 if (Len == 5 && !memcmp(IdentStart, "class", 5)) return tgtok::Class;
229 if (Len == 3 && !memcmp(IdentStart, "def", 3)) return tgtok::Def;
230 if (Len == 4 && !memcmp(IdentStart, "defm", 4)) return tgtok::Defm;
231 if (Len == 10 && !memcmp(IdentStart, "multiclass", 10))
232 return tgtok::MultiClass;
233 if (Len == 5 && !memcmp(IdentStart, "field", 5)) return tgtok::Field;
234 if (Len == 3 && !memcmp(IdentStart, "let", 3)) return tgtok::Let;
235 if (Len == 2 && !memcmp(IdentStart, "in", 2)) return tgtok::In;
237 if (Len == 7 && !memcmp(IdentStart, "include", 7)) {
238 if (LexInclude()) return tgtok::Error;
242 CurStrVal.assign(IdentStart, CurPtr);
246 /// LexInclude - We just read the "include" token. Get the string token that
247 /// comes next and enter the include.
248 bool TGLexer::LexInclude() {
249 // The token after the include must be a string.
250 tgtok::TokKind Tok = LexToken();
251 if (Tok == tgtok::Error) return true;
252 if (Tok != tgtok::StrVal) {
253 PrintError(getLoc(), "Expected filename after include");
258 std::string Filename = CurStrVal;
260 // Try to find the file.
261 MemoryBuffer *NewBuf = MemoryBuffer::getFile(Filename.c_str());
263 // If the file didn't exist directly, see if it's in an include path.
264 for (unsigned i = 0, e = IncludeDirectories.size(); i != e && !NewBuf; ++i) {
265 std::string IncFile = IncludeDirectories[i] + "/" + Filename;
266 NewBuf = MemoryBuffer::getFile(IncFile.c_str());
270 PrintError(getLoc(), "Could not find include file '" + Filename + "'");
274 // Save the line number and lex buffer of the includer.
275 IncludeStack.push_back(IncludeRec(CurBuf, CurPtr, CurLineNo));
277 CurLineNo = 1; // Reset line numbering.
279 CurPtr = CurBuf->getBufferStart();
283 void TGLexer::SkipBCPLComment() {
284 ++CurPtr; // skip the second slash.
289 return; // Newline is end of comment.
291 // If this is the end of the buffer, end the comment.
292 if (CurPtr == CurBuf->getBufferEnd())
296 // Otherwise, skip the character.
301 /// SkipCComment - This skips C-style /**/ comments. The only difference from C
302 /// is that we allow nesting.
303 bool TGLexer::SkipCComment() {
304 ++CurPtr; // skip the star.
305 unsigned CommentDepth = 1;
308 int CurChar = getNextChar();
311 PrintError(TokStart, "Unterminated comment!");
314 // End of the comment?
315 if (CurPtr[0] != '/') break;
317 ++CurPtr; // End the */.
318 if (--CommentDepth == 0)
322 // Start of a nested comment?
323 if (CurPtr[0] != '*') break;
335 tgtok::TokKind TGLexer::LexNumber() {
336 if (CurPtr[-1] == '0') {
337 if (CurPtr[0] == 'x') {
339 const char *NumStart = CurPtr;
340 while (isxdigit(CurPtr[0]))
343 // Requires at least one hex digit.
344 if (CurPtr == NumStart)
345 return ReturnError(CurPtr-2, "Invalid hexadecimal number");
348 CurIntVal = strtoll(NumStart, 0, 16);
350 return ReturnError(CurPtr-2, "Invalid hexadecimal number");
351 if (errno == ERANGE) {
353 CurIntVal = (int64_t)strtoull(NumStart, 0, 16);
355 return ReturnError(CurPtr-2, "Invalid hexadecimal number");
357 return ReturnError(CurPtr-2, "Hexadecimal number out of range");
359 return tgtok::IntVal;
360 } else if (CurPtr[0] == 'b') {
362 const char *NumStart = CurPtr;
363 while (CurPtr[0] == '0' || CurPtr[0] == '1')
366 // Requires at least one binary digit.
367 if (CurPtr == NumStart)
368 return ReturnError(CurPtr-2, "Invalid binary number");
369 CurIntVal = strtoll(NumStart, 0, 2);
370 return tgtok::IntVal;
374 // Check for a sign without a digit.
375 if (!isdigit(CurPtr[0])) {
376 if (CurPtr[-1] == '-')
378 else if (CurPtr[-1] == '+')
382 while (isdigit(CurPtr[0]))
384 CurIntVal = strtoll(TokStart, 0, 10);
385 return tgtok::IntVal;
388 /// LexBracket - We just read '['. If this is a code block, return it,
389 /// otherwise return the bracket. Match: '[' and '[{ ( [^}]+ | }[^]] )* }]'
390 tgtok::TokKind TGLexer::LexBracket() {
391 if (CurPtr[0] != '{')
392 return tgtok::l_square;
394 const char *CodeStart = CurPtr;
396 int Char = getNextChar();
397 if (Char == EOF) break;
399 if (Char != '}') continue;
401 Char = getNextChar();
402 if (Char == EOF) break;
404 CurStrVal.assign(CodeStart, CurPtr-2);
405 return tgtok::CodeFragment;
409 return ReturnError(CodeStart-2, "Unterminated Code Block");
412 /// LexExclaim - Lex '!' and '![a-zA-Z]+'.
413 tgtok::TokKind TGLexer::LexExclaim() {
414 if (!isalpha(*CurPtr))
415 return ReturnError(CurPtr-1, "Invalid \"!operator\"");
417 const char *Start = CurPtr++;
418 while (isalpha(*CurPtr))
421 // Check to see which operator this is.
422 unsigned Len = CurPtr-Start;
424 if (Len == 3 && !memcmp(Start, "con", 3)) return tgtok::XConcat;
425 if (Len == 3 && !memcmp(Start, "sra", 3)) return tgtok::XSRA;
426 if (Len == 3 && !memcmp(Start, "srl", 3)) return tgtok::XSRL;
427 if (Len == 3 && !memcmp(Start, "shl", 3)) return tgtok::XSHL;
428 if (Len == 9 && !memcmp(Start, "strconcat", 9)) return tgtok::XStrConcat;
430 return ReturnError(Start-1, "Unknown operator");