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 CurPtr = isHex || !FirstHex ? LookAhead : FirstHex;
188 /// LexDigit: First character is [0-9].
189 /// Local Label: [0-9][:]
190 /// Forward/Backward Label: [0-9][fb]
191 /// Binary integer: 0b[01]+
192 /// Octal integer: 0[0-7]+
193 /// Hex integer: 0x[0-9a-fA-F]+ or [0x]?[0-9][0-9a-fA-F]*[hH]
194 /// Decimal integer: [1-9][0-9]*
195 AsmToken AsmLexer::LexDigit() {
196 // Decimal integer: [1-9][0-9]*
197 if (CurPtr[-1] != '0' || CurPtr[0] == '.') {
198 unsigned Radix = doLookAhead(CurPtr, 10);
199 bool isHex = Radix == 16;
200 // Check for floating point literals.
201 if (!isHex && (*CurPtr == '.' || *CurPtr == 'e')) {
203 return LexFloatLiteral();
206 StringRef Result(TokStart, CurPtr - TokStart);
209 if (Result.getAsInteger(Radix, Value)) {
210 // Allow positive values that are too large to fit into a signed 64-bit
211 // integer, but that do fit in an unsigned one, we just convert them over.
212 unsigned long long UValue;
213 if (Result.getAsInteger(Radix, UValue))
214 return ReturnError(TokStart, !isHex ? "invalid decimal number" :
215 "invalid hexdecimal number");
216 Value = (long long)UValue;
219 // Consume the [bB][hH].
220 if (Radix == 2 || Radix == 16)
223 // The darwin/x86 (and x86-64) assembler accepts and ignores ULL and LL
224 // suffixes on integer literals.
225 SkipIgnoredIntegerSuffix(CurPtr);
227 return AsmToken(AsmToken::Integer, Result, Value);
230 if (*CurPtr == 'b') {
232 // See if we actually have "0b" as part of something like "jmp 0b\n"
233 if (!isdigit(CurPtr[0])) {
235 StringRef Result(TokStart, CurPtr - TokStart);
236 return AsmToken(AsmToken::Integer, Result, 0);
238 const char *NumStart = CurPtr;
239 while (CurPtr[0] == '0' || CurPtr[0] == '1')
242 // Requires at least one binary digit.
243 if (CurPtr == NumStart)
244 return ReturnError(TokStart, "invalid binary number");
246 StringRef Result(TokStart, CurPtr - TokStart);
249 if (Result.substr(2).getAsInteger(2, Value))
250 return ReturnError(TokStart, "invalid binary number");
252 // The darwin/x86 (and x86-64) assembler accepts and ignores ULL and LL
253 // suffixes on integer literals.
254 SkipIgnoredIntegerSuffix(CurPtr);
256 return AsmToken(AsmToken::Integer, Result, Value);
259 if (*CurPtr == 'x') {
261 const char *NumStart = CurPtr;
262 while (isxdigit(CurPtr[0]))
265 // Requires at least one hex digit.
266 if (CurPtr == NumStart)
267 return ReturnError(CurPtr-2, "invalid hexadecimal number");
269 unsigned long long Result;
270 if (StringRef(TokStart, CurPtr - TokStart).getAsInteger(0, Result))
271 return ReturnError(TokStart, "invalid hexadecimal number");
273 // Consume the optional [hH].
274 if (*CurPtr == 'h' || *CurPtr == 'H')
277 // The darwin/x86 (and x86-64) assembler accepts and ignores ULL and LL
278 // suffixes on integer literals.
279 SkipIgnoredIntegerSuffix(CurPtr);
281 return AsmToken(AsmToken::Integer, StringRef(TokStart, CurPtr - TokStart),
285 // Either octal or hexidecimal.
287 unsigned Radix = doLookAhead(CurPtr, 8);
288 bool isHex = Radix == 16;
289 StringRef Result(TokStart, CurPtr - TokStart);
290 if (Result.getAsInteger(Radix, Value))
291 return ReturnError(TokStart, !isHex ? "invalid octal number" :
292 "invalid hexdecimal number");
298 // The darwin/x86 (and x86-64) assembler accepts and ignores ULL and LL
299 // suffixes on integer literals.
300 SkipIgnoredIntegerSuffix(CurPtr);
302 return AsmToken(AsmToken::Integer, Result, Value);
305 /// LexSingleQuote: Integer: 'b'
306 AsmToken AsmLexer::LexSingleQuote() {
307 int CurChar = getNextChar();
310 CurChar = getNextChar();
313 return ReturnError(TokStart, "unterminated single quote");
315 CurChar = getNextChar();
318 return ReturnError(TokStart, "single quote way too long");
320 // The idea here being that 'c' is basically just an integral
322 StringRef Res = StringRef(TokStart,CurPtr - TokStart);
325 if (Res.startswith("\'\\")) {
326 char theChar = Res[2];
328 default: Value = theChar; break;
329 case '\'': Value = '\''; break;
330 case 't': Value = '\t'; break;
331 case 'n': Value = '\n'; break;
332 case 'b': Value = '\b'; break;
337 return AsmToken(AsmToken::Integer, Res, Value);
341 /// LexQuote: String: "..."
342 AsmToken AsmLexer::LexQuote() {
343 int CurChar = getNextChar();
344 // TODO: does gas allow multiline string constants?
345 while (CurChar != '"') {
346 if (CurChar == '\\') {
348 CurChar = getNextChar();
352 return ReturnError(TokStart, "unterminated string constant");
354 CurChar = getNextChar();
357 return AsmToken(AsmToken::String, StringRef(TokStart, CurPtr - TokStart));
360 StringRef AsmLexer::LexUntilEndOfStatement() {
363 while (!isAtStartOfComment(*CurPtr) && // Start of line comment.
364 !isAtStatementSeparator(CurPtr) && // End of statement marker.
367 (*CurPtr != 0 || CurPtr != CurBuf->getBufferEnd())) {
370 return StringRef(TokStart, CurPtr-TokStart);
373 StringRef AsmLexer::LexUntilEndOfLine() {
376 while (*CurPtr != '\n' &&
378 (*CurPtr != 0 || CurPtr != CurBuf->getBufferEnd())) {
381 return StringRef(TokStart, CurPtr-TokStart);
384 bool AsmLexer::isAtStartOfComment(char Char) {
385 // FIXME: This won't work for multi-character comment indicators like "//".
386 return Char == *MAI.getCommentString();
389 bool AsmLexer::isAtStatementSeparator(const char *Ptr) {
390 return strncmp(Ptr, MAI.getSeparatorString(),
391 strlen(MAI.getSeparatorString())) == 0;
394 AsmToken AsmLexer::LexToken() {
396 // This always consumes at least one character.
397 int CurChar = getNextChar();
399 if (isAtStartOfComment(CurChar)) {
400 // If this comment starts with a '#', then return the Hash token and let
401 // the assembler parser see if it can be parsed as a cpp line filename
402 // comment. We do this only if we are at the start of a line.
403 if (CurChar == '#' && isAtStartOfLine)
404 return AsmToken(AsmToken::Hash, StringRef(TokStart, 1));
405 isAtStartOfLine = true;
406 return LexLineComment();
408 if (isAtStatementSeparator(TokStart)) {
409 CurPtr += strlen(MAI.getSeparatorString()) - 1;
410 return AsmToken(AsmToken::EndOfStatement,
411 StringRef(TokStart, strlen(MAI.getSeparatorString())));
414 // If we're missing a newline at EOF, make sure we still get an
415 // EndOfStatement token before the Eof token.
416 if (CurChar == EOF && !isAtStartOfLine) {
417 isAtStartOfLine = true;
418 return AsmToken(AsmToken::EndOfStatement, StringRef(TokStart, 1));
421 isAtStartOfLine = false;
424 // Handle identifier: [a-zA-Z_.][a-zA-Z0-9_$.@]*
425 if (isalpha(CurChar) || CurChar == '_' || CurChar == '.')
426 return LexIdentifier();
428 // Unknown character, emit an error.
429 return ReturnError(TokStart, "invalid character in input");
430 case EOF: return AsmToken(AsmToken::Eof, StringRef(TokStart, 0));
435 // Ignore whitespace.
439 while (*CurPtr==' ' || *CurPtr=='\t') {
443 return AsmToken(AsmToken::Space, StringRef(TokStart, len));
445 case '\n': // FALL THROUGH.
447 isAtStartOfLine = true;
448 return AsmToken(AsmToken::EndOfStatement, StringRef(TokStart, 1));
449 case ':': return AsmToken(AsmToken::Colon, StringRef(TokStart, 1));
450 case '+': return AsmToken(AsmToken::Plus, StringRef(TokStart, 1));
451 case '-': return AsmToken(AsmToken::Minus, StringRef(TokStart, 1));
452 case '~': return AsmToken(AsmToken::Tilde, StringRef(TokStart, 1));
453 case '(': return AsmToken(AsmToken::LParen, StringRef(TokStart, 1));
454 case ')': return AsmToken(AsmToken::RParen, StringRef(TokStart, 1));
455 case '[': return AsmToken(AsmToken::LBrac, StringRef(TokStart, 1));
456 case ']': return AsmToken(AsmToken::RBrac, StringRef(TokStart, 1));
457 case '{': return AsmToken(AsmToken::LCurly, StringRef(TokStart, 1));
458 case '}': return AsmToken(AsmToken::RCurly, StringRef(TokStart, 1));
459 case '*': return AsmToken(AsmToken::Star, StringRef(TokStart, 1));
460 case ',': return AsmToken(AsmToken::Comma, StringRef(TokStart, 1));
461 case '$': return AsmToken(AsmToken::Dollar, StringRef(TokStart, 1));
462 case '@': return AsmToken(AsmToken::At, StringRef(TokStart, 1));
463 case '\\': return AsmToken(AsmToken::BackSlash, StringRef(TokStart, 1));
466 return ++CurPtr, AsmToken(AsmToken::EqualEqual, StringRef(TokStart, 2));
467 return AsmToken(AsmToken::Equal, StringRef(TokStart, 1));
470 return ++CurPtr, AsmToken(AsmToken::PipePipe, StringRef(TokStart, 2));
471 return AsmToken(AsmToken::Pipe, StringRef(TokStart, 1));
472 case '^': return AsmToken(AsmToken::Caret, StringRef(TokStart, 1));
475 return ++CurPtr, AsmToken(AsmToken::AmpAmp, StringRef(TokStart, 2));
476 return AsmToken(AsmToken::Amp, StringRef(TokStart, 1));
479 return ++CurPtr, AsmToken(AsmToken::ExclaimEqual, StringRef(TokStart, 2));
480 return AsmToken(AsmToken::Exclaim, StringRef(TokStart, 1));
481 case '%': return AsmToken(AsmToken::Percent, StringRef(TokStart, 1));
482 case '/': return LexSlash();
483 case '#': return AsmToken(AsmToken::Hash, StringRef(TokStart, 1));
484 case '\'': return LexSingleQuote();
485 case '"': return LexQuote();
486 case '0': case '1': case '2': case '3': case '4':
487 case '5': case '6': case '7': case '8': case '9':
491 case '<': return ++CurPtr, AsmToken(AsmToken::LessLess,
492 StringRef(TokStart, 2));
493 case '=': return ++CurPtr, AsmToken(AsmToken::LessEqual,
494 StringRef(TokStart, 2));
495 case '>': return ++CurPtr, AsmToken(AsmToken::LessGreater,
496 StringRef(TokStart, 2));
497 default: return AsmToken(AsmToken::Less, StringRef(TokStart, 1));
501 case '>': return ++CurPtr, AsmToken(AsmToken::GreaterGreater,
502 StringRef(TokStart, 2));
503 case '=': return ++CurPtr, AsmToken(AsmToken::GreaterEqual,
504 StringRef(TokStart, 2));
505 default: return AsmToken(AsmToken::Greater, StringRef(TokStart, 1));
508 // TODO: Quoted identifiers (objc methods etc)
509 // local labels: [0-9][:]
510 // Forward/backward labels: [0-9][fb]
511 // Integers, fp constants, character constants.