1 //===- TGLexer.cpp - Lexer for TableGen -----------------------------------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // Implement the Lexer for TableGen.
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "TGLexer.h"
15 #include "llvm/ADT/StringSwitch.h"
16 #include "llvm/ADT/Twine.h"
17 #include "llvm/Config/config.h" // for strtoull()/strtoll() define
18 #include "llvm/Support/Compiler.h"
19 #include "llvm/Support/MemoryBuffer.h"
20 #include "llvm/Support/SourceMgr.h"
21 #include "llvm/TableGen/Error.h"
22 #include <cctype>
23 #include <cerrno>
24 #include <cstdint>
25 #include <cstdio>
26 #include <cstdlib>
27 #include <cstring>
28 
29 using namespace llvm;
30 
31 TGLexer::TGLexer(SourceMgr &SM) : SrcMgr(SM) {
32   CurBuffer = SrcMgr.getMainFileID();
33   CurBuf = SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer();
34   CurPtr = CurBuf.begin();
35   TokStart = nullptr;
36 }
37 
38 SMLoc TGLexer::getLoc() const {
39   return SMLoc::getFromPointer(TokStart);
40 }
41 
42 /// ReturnError - Set the error to the specified string at the specified
43 /// location.  This is defined to always return tgtok::Error.
44 tgtok::TokKind TGLexer::ReturnError(const char *Loc, const Twine &Msg) {
45   PrintError(Loc, Msg);
46   return tgtok::Error;
47 }
48 
49 int TGLexer::getNextChar() {
50   char CurChar = *CurPtr++;
51   switch (CurChar) {
52   default:
53     return (unsigned char)CurChar;
54   case 0: {
55     // A nul character in the stream is either the end of the current buffer or
56     // a random nul in the file.  Disambiguate that here.
57     if (CurPtr-1 != CurBuf.end())
58       return 0;  // Just whitespace.
59 
60     // If this is the end of an included file, pop the parent file off the
61     // include stack.
62     SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
63     if (ParentIncludeLoc != SMLoc()) {
64       CurBuffer = SrcMgr.FindBufferContainingLoc(ParentIncludeLoc);
65       CurBuf = SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer();
66       CurPtr = ParentIncludeLoc.getPointer();
67       return getNextChar();
68     }
69 
70     // Otherwise, return end of file.
71     --CurPtr;  // Another call to lex will return EOF again.
72     return EOF;
73   }
74   case '\n':
75   case '\r':
76     // Handle the newline character by ignoring it and incrementing the line
77     // count.  However, be careful about 'dos style' files with \n\r in them.
78     // Only treat a \n\r or \r\n as a single line.
79     if ((*CurPtr == '\n' || (*CurPtr == '\r')) &&
80         *CurPtr != CurChar)
81       ++CurPtr;  // Eat the two char newline sequence.
82     return '\n';
83   }
84 }
85 
86 int TGLexer::peekNextChar(int Index) {
87   return *(CurPtr + Index);
88 }
89 
90 tgtok::TokKind TGLexer::LexToken() {
91   TokStart = CurPtr;
92   // This always consumes at least one character.
93   int CurChar = getNextChar();
94 
95   switch (CurChar) {
96   default:
97     // Handle letters: [a-zA-Z_]
98     if (isalpha(CurChar) || CurChar == '_')
99       return LexIdentifier();
100 
101     // Unknown character, emit an error.
102     return ReturnError(TokStart, "Unexpected character");
103   case EOF: return tgtok::Eof;
104   case ':': return tgtok::colon;
105   case ';': return tgtok::semi;
106   case '.': return tgtok::period;
107   case ',': return tgtok::comma;
108   case '<': return tgtok::less;
109   case '>': return tgtok::greater;
110   case ']': return tgtok::r_square;
111   case '{': return tgtok::l_brace;
112   case '}': return tgtok::r_brace;
113   case '(': return tgtok::l_paren;
114   case ')': return tgtok::r_paren;
115   case '=': return tgtok::equal;
116   case '?': return tgtok::question;
117   case '#': return tgtok::paste;
118 
119   case 0:
120   case ' ':
121   case '\t':
122   case '\n':
123   case '\r':
124     // Ignore whitespace.
125     return LexToken();
126   case '/':
127     // If this is the start of a // comment, skip until the end of the line or
128     // the end of the buffer.
129     if (*CurPtr == '/')
130       SkipBCPLComment();
131     else if (*CurPtr == '*') {
132       if (SkipCComment())
133         return tgtok::Error;
134     } else // Otherwise, this is an error.
135       return ReturnError(TokStart, "Unexpected character");
136     return LexToken();
137   case '-': case '+':
138   case '0': case '1': case '2': case '3': case '4': case '5': case '6':
139   case '7': case '8': case '9': {
140     int NextChar = 0;
141     if (isdigit(CurChar)) {
142       // Allow identifiers to start with a number if it is followed by
143       // an identifier.  This can happen with paste operations like
144       // foo#8i.
145       int i = 0;
146       do {
147         NextChar = peekNextChar(i++);
148       } while (isdigit(NextChar));
149 
150       if (NextChar == 'x' || NextChar == 'b') {
151         // If this is [0-9]b[01] or [0-9]x[0-9A-fa-f] this is most
152         // likely a number.
153         int NextNextChar = peekNextChar(i);
154         switch (NextNextChar) {
155         default:
156           break;
157         case '0': case '1':
158           if (NextChar == 'b')
159             return LexNumber();
160           LLVM_FALLTHROUGH;
161         case '2': case '3': case '4': case '5':
162         case '6': case '7': case '8': case '9':
163         case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
164         case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
165           if (NextChar == 'x')
166             return LexNumber();
167           break;
168         }
169       }
170     }
171 
172     if (isalpha(NextChar) || NextChar == '_')
173       return LexIdentifier();
174 
175     return LexNumber();
176   }
177   case '"': return LexString();
178   case '$': return LexVarName();
179   case '[': return LexBracket();
180   case '!': return LexExclaim();
181   }
182 }
183 
184 /// LexString - Lex "[^"]*"
185 tgtok::TokKind TGLexer::LexString() {
186   const char *StrStart = CurPtr;
187 
188   CurStrVal = "";
189 
190   while (*CurPtr != '"') {
191     // If we hit the end of the buffer, report an error.
192     if (*CurPtr == 0 && CurPtr == CurBuf.end())
193       return ReturnError(StrStart, "End of file in string literal");
194 
195     if (*CurPtr == '\n' || *CurPtr == '\r')
196       return ReturnError(StrStart, "End of line in string literal");
197 
198     if (*CurPtr != '\\') {
199       CurStrVal += *CurPtr++;
200       continue;
201     }
202 
203     ++CurPtr;
204 
205     switch (*CurPtr) {
206     case '\\': case '\'': case '"':
207       // These turn into their literal character.
208       CurStrVal += *CurPtr++;
209       break;
210     case 't':
211       CurStrVal += '\t';
212       ++CurPtr;
213       break;
214     case 'n':
215       CurStrVal += '\n';
216       ++CurPtr;
217       break;
218 
219     case '\n':
220     case '\r':
221       return ReturnError(CurPtr, "escaped newlines not supported in tblgen");
222 
223     // If we hit the end of the buffer, report an error.
224     case '\0':
225       if (CurPtr == CurBuf.end())
226         return ReturnError(StrStart, "End of file in string literal");
227       LLVM_FALLTHROUGH;
228     default:
229       return ReturnError(CurPtr, "invalid escape in string literal");
230     }
231   }
232 
233   ++CurPtr;
234   return tgtok::StrVal;
235 }
236 
237 tgtok::TokKind TGLexer::LexVarName() {
238   if (!isalpha(CurPtr[0]) && CurPtr[0] != '_')
239     return ReturnError(TokStart, "Invalid variable name");
240 
241   // Otherwise, we're ok, consume the rest of the characters.
242   const char *VarNameStart = CurPtr++;
243 
244   while (isalpha(*CurPtr) || isdigit(*CurPtr) || *CurPtr == '_')
245     ++CurPtr;
246 
247   CurStrVal.assign(VarNameStart, CurPtr);
248   return tgtok::VarName;
249 }
250 
251 tgtok::TokKind TGLexer::LexIdentifier() {
252   // The first letter is [a-zA-Z_#].
253   const char *IdentStart = TokStart;
254 
255   // Match the rest of the identifier regex: [0-9a-zA-Z_#]*
256   while (isalpha(*CurPtr) || isdigit(*CurPtr) || *CurPtr == '_')
257     ++CurPtr;
258 
259   // Check to see if this identifier is a keyword.
260   StringRef Str(IdentStart, CurPtr-IdentStart);
261 
262   if (Str == "include") {
263     if (LexInclude()) return tgtok::Error;
264     return Lex();
265   }
266 
267   tgtok::TokKind Kind = StringSwitch<tgtok::TokKind>(Str)
268     .Case("int", tgtok::Int)
269     .Case("bit", tgtok::Bit)
270     .Case("bits", tgtok::Bits)
271     .Case("string", tgtok::String)
272     .Case("list", tgtok::List)
273     .Case("code", tgtok::Code)
274     .Case("dag", tgtok::Dag)
275     .Case("class", tgtok::Class)
276     .Case("def", tgtok::Def)
277     .Case("foreach", tgtok::Foreach)
278     .Case("defm", tgtok::Defm)
279     .Case("multiclass", tgtok::MultiClass)
280     .Case("field", tgtok::Field)
281     .Case("let", tgtok::Let)
282     .Case("in", tgtok::In)
283     .Default(tgtok::Id);
284 
285   if (Kind == tgtok::Id)
286     CurStrVal.assign(Str.begin(), Str.end());
287   return Kind;
288 }
289 
290 /// LexInclude - We just read the "include" token.  Get the string token that
291 /// comes next and enter the include.
292 bool TGLexer::LexInclude() {
293   // The token after the include must be a string.
294   tgtok::TokKind Tok = LexToken();
295   if (Tok == tgtok::Error) return true;
296   if (Tok != tgtok::StrVal) {
297     PrintError(getLoc(), "Expected filename after include");
298     return true;
299   }
300 
301   // Get the string.
302   std::string Filename = CurStrVal;
303   std::string IncludedFile;
304 
305   CurBuffer = SrcMgr.AddIncludeFile(Filename, SMLoc::getFromPointer(CurPtr),
306                                     IncludedFile);
307   if (!CurBuffer) {
308     PrintError(getLoc(), "Could not find include file '" + Filename + "'");
309     return true;
310   }
311 
312   DependenciesMapTy::const_iterator Found = Dependencies.find(IncludedFile);
313   if (Found != Dependencies.end()) {
314     PrintError(getLoc(),
315                "File '" + IncludedFile + "' has already been included.");
316     SrcMgr.PrintMessage(Found->second, SourceMgr::DK_Note,
317                         "previously included here");
318     return true;
319   }
320   Dependencies.insert(std::make_pair(IncludedFile, getLoc()));
321   // Save the line number and lex buffer of the includer.
322   CurBuf = SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer();
323   CurPtr = CurBuf.begin();
324   return false;
325 }
326 
327 void TGLexer::SkipBCPLComment() {
328   ++CurPtr;  // skip the second slash.
329   while (true) {
330     switch (*CurPtr) {
331     case '\n':
332     case '\r':
333       return;  // Newline is end of comment.
334     case 0:
335       // If this is the end of the buffer, end the comment.
336       if (CurPtr == CurBuf.end())
337         return;
338       break;
339     }
340     // Otherwise, skip the character.
341     ++CurPtr;
342   }
343 }
344 
345 /// SkipCComment - This skips C-style /**/ comments.  The only difference from C
346 /// is that we allow nesting.
347 bool TGLexer::SkipCComment() {
348   ++CurPtr;  // skip the star.
349   unsigned CommentDepth = 1;
350 
351   while (true) {
352     int CurChar = getNextChar();
353     switch (CurChar) {
354     case EOF:
355       PrintError(TokStart, "Unterminated comment!");
356       return true;
357     case '*':
358       // End of the comment?
359       if (CurPtr[0] != '/') break;
360 
361       ++CurPtr;   // End the */.
362       if (--CommentDepth == 0)
363         return false;
364       break;
365     case '/':
366       // Start of a nested comment?
367       if (CurPtr[0] != '*') break;
368       ++CurPtr;
369       ++CommentDepth;
370       break;
371     }
372   }
373 }
374 
375 /// LexNumber - Lex:
376 ///    [-+]?[0-9]+
377 ///    0x[0-9a-fA-F]+
378 ///    0b[01]+
379 tgtok::TokKind TGLexer::LexNumber() {
380   if (CurPtr[-1] == '0') {
381     if (CurPtr[0] == 'x') {
382       ++CurPtr;
383       const char *NumStart = CurPtr;
384       while (isxdigit(CurPtr[0]))
385         ++CurPtr;
386 
387       // Requires at least one hex digit.
388       if (CurPtr == NumStart)
389         return ReturnError(TokStart, "Invalid hexadecimal number");
390 
391       errno = 0;
392       CurIntVal = strtoll(NumStart, nullptr, 16);
393       if (errno == EINVAL)
394         return ReturnError(TokStart, "Invalid hexadecimal number");
395       if (errno == ERANGE) {
396         errno = 0;
397         CurIntVal = (int64_t)strtoull(NumStart, nullptr, 16);
398         if (errno == EINVAL)
399           return ReturnError(TokStart, "Invalid hexadecimal number");
400         if (errno == ERANGE)
401           return ReturnError(TokStart, "Hexadecimal number out of range");
402       }
403       return tgtok::IntVal;
404     } else if (CurPtr[0] == 'b') {
405       ++CurPtr;
406       const char *NumStart = CurPtr;
407       while (CurPtr[0] == '0' || CurPtr[0] == '1')
408         ++CurPtr;
409 
410       // Requires at least one binary digit.
411       if (CurPtr == NumStart)
412         return ReturnError(CurPtr-2, "Invalid binary number");
413       CurIntVal = strtoll(NumStart, nullptr, 2);
414       return tgtok::BinaryIntVal;
415     }
416   }
417 
418   // Check for a sign without a digit.
419   if (!isdigit(CurPtr[0])) {
420     if (CurPtr[-1] == '-')
421       return tgtok::minus;
422     else if (CurPtr[-1] == '+')
423       return tgtok::plus;
424   }
425 
426   while (isdigit(CurPtr[0]))
427     ++CurPtr;
428   CurIntVal = strtoll(TokStart, nullptr, 10);
429   return tgtok::IntVal;
430 }
431 
432 /// LexBracket - We just read '['.  If this is a code block, return it,
433 /// otherwise return the bracket.  Match: '[' and '[{ ( [^}]+ | }[^]] )* }]'
434 tgtok::TokKind TGLexer::LexBracket() {
435   if (CurPtr[0] != '{')
436     return tgtok::l_square;
437   ++CurPtr;
438   const char *CodeStart = CurPtr;
439   while (true) {
440     int Char = getNextChar();
441     if (Char == EOF) break;
442 
443     if (Char != '}') continue;
444 
445     Char = getNextChar();
446     if (Char == EOF) break;
447     if (Char == ']') {
448       CurStrVal.assign(CodeStart, CurPtr-2);
449       return tgtok::CodeFragment;
450     }
451   }
452 
453   return ReturnError(CodeStart-2, "Unterminated Code Block");
454 }
455 
456 /// LexExclaim - Lex '!' and '![a-zA-Z]+'.
457 tgtok::TokKind TGLexer::LexExclaim() {
458   if (!isalpha(*CurPtr))
459     return ReturnError(CurPtr - 1, "Invalid \"!operator\"");
460 
461   const char *Start = CurPtr++;
462   while (isalpha(*CurPtr))
463     ++CurPtr;
464 
465   // Check to see which operator this is.
466   tgtok::TokKind Kind =
467     StringSwitch<tgtok::TokKind>(StringRef(Start, CurPtr - Start))
468     .Case("eq", tgtok::XEq)
469     .Case("if", tgtok::XIf)
470     .Case("head", tgtok::XHead)
471     .Case("tail", tgtok::XTail)
472     .Case("con", tgtok::XConcat)
473     .Case("add", tgtok::XADD)
474     .Case("and", tgtok::XAND)
475     .Case("shl", tgtok::XSHL)
476     .Case("sra", tgtok::XSRA)
477     .Case("srl", tgtok::XSRL)
478     .Case("cast", tgtok::XCast)
479     .Case("empty", tgtok::XEmpty)
480     .Case("subst", tgtok::XSubst)
481     .Case("foreach", tgtok::XForEach)
482     .Case("listconcat", tgtok::XListConcat)
483     .Case("strconcat", tgtok::XStrConcat)
484     .Default(tgtok::Error);
485 
486   return Kind != tgtok::Error ? Kind : ReturnError(Start-1, "Unknown operator");
487 }
488