1 //===- AsmLexer.cpp - Lexer for Assembly Files ----------------------------===//
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 class implements the lexer for assembly files.
12 //===----------------------------------------------------------------------===//
14 #include "llvm/MC/MCParser/AsmLexer.h"
15 #include "llvm/MC/MCAsmInfo.h"
16 #include "llvm/Support/MemoryBuffer.h"
17 #include "llvm/Support/SMLoc.h"
24 AsmLexer::AsmLexer(const MCAsmInfo &_MAI) : MAI(_MAI) {
27 isAtStartOfLine = true;
30 AsmLexer::~AsmLexer() {
33 void AsmLexer::setBuffer(const MemoryBuffer *buf, const char *ptr) {
39 CurPtr = CurBuf->getBufferStart();
44 /// ReturnError - Set the error to the specified string at the specified
45 /// location. This is defined to always return AsmToken::Error.
46 AsmToken AsmLexer::ReturnError(const char *Loc, const std::string &Msg) {
47 SetError(SMLoc::getFromPointer(Loc), Msg);
49 return AsmToken(AsmToken::Error, StringRef(Loc, 0));
52 int AsmLexer::getNextChar() {
53 char CurChar = *CurPtr++;
56 return (unsigned char)CurChar;
58 // A nul character in the stream is either the end of the current buffer or
59 // a random nul in the file. Disambiguate that here.
60 if (CurPtr-1 != CurBuf->getBufferEnd())
61 return 0; // Just whitespace.
63 // Otherwise, return end of file.
64 --CurPtr; // Another call to lex will return EOF again.
69 /// LexFloatLiteral: [0-9]*[.][0-9]*([eE][+-]?[0-9]*)?
71 /// The leading integral digit sequence and dot should have already been
72 /// consumed, some or all of the fractional digit sequence *can* have been
74 AsmToken AsmLexer::LexFloatLiteral() {
75 // Skip the fractional digit sequence.
76 while (isdigit(*CurPtr))
79 // Check for exponent; we intentionally accept a slighlty wider set of
80 // literals here and rely on the upstream client to reject invalid ones (e.g.,
82 if (*CurPtr == 'e' || *CurPtr == 'E') {
84 if (*CurPtr == '-' || *CurPtr == '+')
86 while (isdigit(*CurPtr))
90 return AsmToken(AsmToken::Real,
91 StringRef(TokStart, CurPtr - TokStart));
94 /// LexIdentifier: [a-zA-Z_.][a-zA-Z0-9_$.@]*
95 static bool IsIdentifierChar(char c) {
96 return isalnum(c) || c == '_' || c == '$' || c == '.' || c == '@';
98 AsmToken AsmLexer::LexIdentifier() {
99 // Check for floating point literals.
100 if (CurPtr[-1] == '.' && isdigit(*CurPtr)) {
101 // Disambiguate a .1243foo identifier from a floating literal.
102 while (isdigit(*CurPtr))
104 if (*CurPtr == 'e' || *CurPtr == 'E' || !IsIdentifierChar(*CurPtr))
105 return LexFloatLiteral();
108 while (IsIdentifierChar(*CurPtr))
111 // Handle . as a special case.
112 if (CurPtr == TokStart+1 && TokStart[0] == '.')
113 return AsmToken(AsmToken::Dot, StringRef(TokStart, 1));
115 return AsmToken(AsmToken::Identifier, StringRef(TokStart, CurPtr - TokStart));
118 /// LexSlash: Slash: /
119 /// C-Style Comment: /* ... */
120 AsmToken AsmLexer::LexSlash() {
122 case '*': break; // C style comment.
123 case '/': return ++CurPtr, LexLineComment();
124 default: return AsmToken(AsmToken::Slash, StringRef(CurPtr-1, 1));
128 ++CurPtr; // skip the star.
130 int CurChar = getNextChar();
133 return ReturnError(TokStart, "unterminated comment");
135 // End of the comment?
136 if (CurPtr[0] != '/') break;
138 ++CurPtr; // End the */.
144 /// LexLineComment: Comment: #[^\n]*
146 AsmToken AsmLexer::LexLineComment() {
147 // FIXME: This is broken if we happen to a comment at the end of a file, which
148 // was .included, and which doesn't end with a newline.
149 int CurChar = getNextChar();
150 while (CurChar != '\n' && CurChar != '\r' && CurChar != EOF)
151 CurChar = getNextChar();
154 return AsmToken(AsmToken::Eof, StringRef(CurPtr, 0));
155 return AsmToken(AsmToken::EndOfStatement, StringRef(CurPtr, 0));
158 static void SkipIgnoredIntegerSuffix(const char *&CurPtr) {
159 if (CurPtr[0] == 'L' && CurPtr[1] == 'L')
161 if (CurPtr[0] == 'U' && CurPtr[1] == 'L' && CurPtr[2] == 'L')
165 // Look ahead to search for first non-hex digit, if it's [hH], then we treat the
166 // integer as a hexadecimal, possibly with leading zeroes.
167 static unsigned doLookAhead(const char *&CurPtr, unsigned DefaultRadix) {
168 const char *FirstHex = 0;
169 const char *LookAhead = CurPtr;
171 if (isdigit(*LookAhead)) {
173 } else if (isxdigit(*LookAhead)) {
175 FirstHex = LookAhead;
181 bool isHex = *LookAhead == 'h' || *LookAhead == 'H';
182 bool isBinary = LookAhead[-1] == 'b' || LookAhead[-1] == 'B';
183 CurPtr = (isBinary || isHex || !FirstHex) ? LookAhead : FirstHex;
193 /// LexDigit: First character is [0-9].
194 /// Local Label: [0-9][:]
195 /// Forward/Backward Label: [0-9]+f or [0-9]b
196 /// Binary integer: 0b[01]+ or [01][bB]
197 /// Octal integer: 0[0-7]+
198 /// Hex integer: 0x[0-9a-fA-F]+ or [0x]?[0-9][0-9a-fA-F]*[hH]
199 /// Decimal integer: [1-9][0-9]*
200 AsmToken AsmLexer::LexDigit() {
202 // Backward Label: [0-9]b
203 if (*CurPtr == 'b') {
204 // See if we actually have "0b" as part of something like "jmp 0b\n"
205 if (!isdigit(CurPtr[1])) {
207 StringRef Result(TokStart, CurPtr - TokStart);
208 if (Result.getAsInteger(10, Value))
209 return ReturnError(TokStart, "invalid backward label");
211 return AsmToken(AsmToken::Integer, Result, Value);
215 // Binary integer: 1[01]*[bB]
216 // Decimal integer: [1-9][0-9]*
217 // Hexidecimal integer: [1-9][0-9a-fA-F]*[hH]
218 if (CurPtr[-1] != '0' || CurPtr[0] == '.') {
219 unsigned Radix = doLookAhead(CurPtr, 10);
220 bool isDecimal = Radix == 10;
222 // Check for floating point literals.
223 if (isDecimal && (*CurPtr == '.' || *CurPtr == 'e')) {
225 return LexFloatLiteral();
228 StringRef Result(TokStart, CurPtr - TokStart);
231 if (Result.getAsInteger(Radix, Value)) {
232 // Allow positive values that are too large to fit into a signed 64-bit
233 // integer, but that do fit in an unsigned one, we just convert them over.
234 unsigned long long UValue;
235 if (Result.getAsInteger(Radix, UValue))
236 return ReturnError(TokStart, isDecimal ? "invalid decimal number" :
237 "invalid hexdecimal number");
238 Value = (long long)UValue;
241 // Consume the [bB][hH].
242 if (Radix == 2 || Radix == 16)
245 // The darwin/x86 (and x86-64) assembler accepts and ignores ULL and LL
246 // suffixes on integer literals.
247 SkipIgnoredIntegerSuffix(CurPtr);
249 return AsmToken(AsmToken::Integer, Result, Value);
252 // Binary integer: 0b[01]+
253 if (*CurPtr == 'b') {
254 const char *NumStart = ++CurPtr;
255 while (CurPtr[0] == '0' || CurPtr[0] == '1')
258 // Requires at least one binary digit.
259 if (CurPtr == NumStart)
260 return ReturnError(TokStart, "invalid binary number");
262 StringRef Result(TokStart, CurPtr - TokStart);
265 if (Result.substr(2).getAsInteger(2, Value))
266 return ReturnError(TokStart, "invalid binary number");
268 // The darwin/x86 (and x86-64) assembler accepts and ignores ULL and LL
269 // suffixes on integer literals.
270 SkipIgnoredIntegerSuffix(CurPtr);
272 return AsmToken(AsmToken::Integer, Result, Value);
275 // Hex integer: 0x[0-9a-fA-F]+
276 if (*CurPtr == 'x') {
278 const char *NumStart = CurPtr;
279 while (isxdigit(CurPtr[0]))
282 // Requires at least one hex digit.
283 if (CurPtr == NumStart)
284 return ReturnError(CurPtr-2, "invalid hexadecimal number");
286 unsigned long long Result;
287 if (StringRef(TokStart, CurPtr - TokStart).getAsInteger(0, Result))
288 return ReturnError(TokStart, "invalid hexadecimal number");
290 // Consume the optional [hH].
291 if (*CurPtr == 'h' || *CurPtr == 'H')
294 // The darwin/x86 (and x86-64) assembler accepts and ignores ULL and LL
295 // suffixes on integer literals.
296 SkipIgnoredIntegerSuffix(CurPtr);
298 return AsmToken(AsmToken::Integer, StringRef(TokStart, CurPtr - TokStart),
302 // Binary: 0[01]*[Bb], but not 0b.
304 // Hexidecimal: [0][0-9a-fA-F]*[hH]
306 unsigned Radix = doLookAhead(CurPtr, 8);
307 bool isBinary = Radix == 2;
308 bool isOctal = Radix == 8;
309 StringRef Result(TokStart, CurPtr - TokStart);
310 if (Result.getAsInteger(Radix, Value))
311 return ReturnError(TokStart, isOctal ? "invalid octal number" :
312 isBinary ? "invalid binary number" :
313 "invalid hexdecimal number");
315 // Consume the [bB][hH].
316 if (Radix == 2 || Radix == 16)
319 // The darwin/x86 (and x86-64) assembler accepts and ignores ULL and LL
320 // suffixes on integer literals.
321 SkipIgnoredIntegerSuffix(CurPtr);
323 return AsmToken(AsmToken::Integer, Result, Value);
326 /// LexSingleQuote: Integer: 'b'
327 AsmToken AsmLexer::LexSingleQuote() {
328 int CurChar = getNextChar();
331 CurChar = getNextChar();
334 return ReturnError(TokStart, "unterminated single quote");
336 CurChar = getNextChar();
339 return ReturnError(TokStart, "single quote way too long");
341 // The idea here being that 'c' is basically just an integral
343 StringRef Res = StringRef(TokStart,CurPtr - TokStart);
346 if (Res.startswith("\'\\")) {
347 char theChar = Res[2];
349 default: Value = theChar; break;
350 case '\'': Value = '\''; break;
351 case 't': Value = '\t'; break;
352 case 'n': Value = '\n'; break;
353 case 'b': Value = '\b'; break;
358 return AsmToken(AsmToken::Integer, Res, Value);
362 /// LexQuote: String: "..."
363 AsmToken AsmLexer::LexQuote() {
364 int CurChar = getNextChar();
365 // TODO: does gas allow multiline string constants?
366 while (CurChar != '"') {
367 if (CurChar == '\\') {
369 CurChar = getNextChar();
373 return ReturnError(TokStart, "unterminated string constant");
375 CurChar = getNextChar();
378 return AsmToken(AsmToken::String, StringRef(TokStart, CurPtr - TokStart));
381 StringRef AsmLexer::LexUntilEndOfStatement() {
384 while (!isAtStartOfComment(*CurPtr) && // Start of line comment.
385 !isAtStatementSeparator(CurPtr) && // End of statement marker.
388 (*CurPtr != 0 || CurPtr != CurBuf->getBufferEnd())) {
391 return StringRef(TokStart, CurPtr-TokStart);
394 StringRef AsmLexer::LexUntilEndOfLine() {
397 while (*CurPtr != '\n' &&
399 (*CurPtr != 0 || CurPtr != CurBuf->getBufferEnd())) {
402 return StringRef(TokStart, CurPtr-TokStart);
405 bool AsmLexer::isAtStartOfComment(char Char) {
406 // FIXME: This won't work for multi-character comment indicators like "//".
407 return Char == *MAI.getCommentString();
410 bool AsmLexer::isAtStatementSeparator(const char *Ptr) {
411 return strncmp(Ptr, MAI.getSeparatorString(),
412 strlen(MAI.getSeparatorString())) == 0;
415 AsmToken AsmLexer::LexToken() {
417 // This always consumes at least one character.
418 int CurChar = getNextChar();
420 if (isAtStartOfComment(CurChar)) {
421 // If this comment starts with a '#', then return the Hash token and let
422 // the assembler parser see if it can be parsed as a cpp line filename
423 // comment. We do this only if we are at the start of a line.
424 if (CurChar == '#' && isAtStartOfLine)
425 return AsmToken(AsmToken::Hash, StringRef(TokStart, 1));
426 isAtStartOfLine = true;
427 return LexLineComment();
429 if (isAtStatementSeparator(TokStart)) {
430 CurPtr += strlen(MAI.getSeparatorString()) - 1;
431 return AsmToken(AsmToken::EndOfStatement,
432 StringRef(TokStart, strlen(MAI.getSeparatorString())));
435 // If we're missing a newline at EOF, make sure we still get an
436 // EndOfStatement token before the Eof token.
437 if (CurChar == EOF && !isAtStartOfLine) {
438 isAtStartOfLine = true;
439 return AsmToken(AsmToken::EndOfStatement, StringRef(TokStart, 1));
442 isAtStartOfLine = false;
445 // Handle identifier: [a-zA-Z_.][a-zA-Z0-9_$.@]*
446 if (isalpha(CurChar) || CurChar == '_' || CurChar == '.')
447 return LexIdentifier();
449 // Unknown character, emit an error.
450 return ReturnError(TokStart, "invalid character in input");
451 case EOF: return AsmToken(AsmToken::Eof, StringRef(TokStart, 0));
456 // Ignore whitespace.
460 while (*CurPtr==' ' || *CurPtr=='\t') {
464 return AsmToken(AsmToken::Space, StringRef(TokStart, len));
466 case '\n': // FALL THROUGH.
468 isAtStartOfLine = true;
469 return AsmToken(AsmToken::EndOfStatement, StringRef(TokStart, 1));
470 case ':': return AsmToken(AsmToken::Colon, StringRef(TokStart, 1));
471 case '+': return AsmToken(AsmToken::Plus, StringRef(TokStart, 1));
472 case '-': return AsmToken(AsmToken::Minus, StringRef(TokStart, 1));
473 case '~': return AsmToken(AsmToken::Tilde, StringRef(TokStart, 1));
474 case '(': return AsmToken(AsmToken::LParen, StringRef(TokStart, 1));
475 case ')': return AsmToken(AsmToken::RParen, StringRef(TokStart, 1));
476 case '[': return AsmToken(AsmToken::LBrac, StringRef(TokStart, 1));
477 case ']': return AsmToken(AsmToken::RBrac, StringRef(TokStart, 1));
478 case '{': return AsmToken(AsmToken::LCurly, StringRef(TokStart, 1));
479 case '}': return AsmToken(AsmToken::RCurly, StringRef(TokStart, 1));
480 case '*': return AsmToken(AsmToken::Star, StringRef(TokStart, 1));
481 case ',': return AsmToken(AsmToken::Comma, StringRef(TokStart, 1));
482 case '$': return AsmToken(AsmToken::Dollar, StringRef(TokStart, 1));
483 case '@': return AsmToken(AsmToken::At, StringRef(TokStart, 1));
484 case '\\': return AsmToken(AsmToken::BackSlash, StringRef(TokStart, 1));
487 return ++CurPtr, AsmToken(AsmToken::EqualEqual, StringRef(TokStart, 2));
488 return AsmToken(AsmToken::Equal, StringRef(TokStart, 1));
491 return ++CurPtr, AsmToken(AsmToken::PipePipe, StringRef(TokStart, 2));
492 return AsmToken(AsmToken::Pipe, StringRef(TokStart, 1));
493 case '^': return AsmToken(AsmToken::Caret, StringRef(TokStart, 1));
496 return ++CurPtr, AsmToken(AsmToken::AmpAmp, StringRef(TokStart, 2));
497 return AsmToken(AsmToken::Amp, StringRef(TokStart, 1));
500 return ++CurPtr, AsmToken(AsmToken::ExclaimEqual, StringRef(TokStart, 2));
501 return AsmToken(AsmToken::Exclaim, StringRef(TokStart, 1));
502 case '%': return AsmToken(AsmToken::Percent, StringRef(TokStart, 1));
503 case '/': return LexSlash();
504 case '#': return AsmToken(AsmToken::Hash, StringRef(TokStart, 1));
505 case '\'': return LexSingleQuote();
506 case '"': return LexQuote();
507 case '0': case '1': case '2': case '3': case '4':
508 case '5': case '6': case '7': case '8': case '9':
512 case '<': return ++CurPtr, AsmToken(AsmToken::LessLess,
513 StringRef(TokStart, 2));
514 case '=': return ++CurPtr, AsmToken(AsmToken::LessEqual,
515 StringRef(TokStart, 2));
516 case '>': return ++CurPtr, AsmToken(AsmToken::LessGreater,
517 StringRef(TokStart, 2));
518 default: return AsmToken(AsmToken::Less, StringRef(TokStart, 1));
522 case '>': return ++CurPtr, AsmToken(AsmToken::GreaterGreater,
523 StringRef(TokStart, 2));
524 case '=': return ++CurPtr, AsmToken(AsmToken::GreaterEqual,
525 StringRef(TokStart, 2));
526 default: return AsmToken(AsmToken::Greater, StringRef(TokStart, 1));
529 // TODO: Quoted identifiers (objc methods etc)
530 // local labels: [0-9][:]
531 // Forward/backward labels: [0-9][fb]
532 // Integers, fp constants, character constants.