Compilers: Yacc/7 1 Compiler Structures Objective – –describe yacc (actually bison) – –give simple examples of its use , Semester 1, Yacc
Compilers: Yacc/7 2 Overview 1. What is Yacc? 2. Format of a yacc/bison File 3. Expressions Compiler 4. Bottom-up Parsing Reminder 5. Expression Conflicts 6.Precedence/Associativity in yacc continued
Compilers: Yacc/7 3 7.Dangling Else Conflict 8.Left and Right Recursion 9.Error Recovery 10.Embedded Actions 11.More Information
Compilers: Yacc/ What is Yacc? Yacc (Yet Another Compiler Compiler) is a tool for translating a context free grammar into a bottom-up LALR parser – –it creates a parse table like that described in the last chapter Yacc is used with lex to create compilers. continued
Compilers: Yacc/7 5 Most people use bison, a much improved version of yacc – –on most modern Unixes, when you call yacc, you're really using bison bison works with flex (the fast version of lex).
Compilers: Yacc/7 6 Bison and Flex $ flex foo.l $ bison foo.y $ gcc foo.tab.c -o foo foo.l, a flex file foo.y, a bison file bison flexlex.yy.c foo.tab.c C compiler foo, c executable #include foo, c executable source program parsed output $./foo < program.txt
Compilers: Yacc/7 7 Compiler Components (in foo) lex.yy.c, Lexical Analyzer (using chars) foo.tab.c, Syntax Analyzer (using tokens) Source Program 3. Token, token value, token type 1. Get next token by calling yylex() lexical errors syntax errors 2. Get chars to make a token parsed output
Compilers: Yacc/7 8 actionsgotos Inside foo.tab.c$ anananan… aiaiaiai… a2a2a2a2 a1a1a1a1 LALR Parser X o s 0 X o s 0 … X m-1 s m-1 X m s m X m s m parsed output stack input tokens X is terminals or non-terminals, S = state Parse table (bison creates this based on your grammar)
Compilers: Yacc/ Format of a yacc/bison File declarations: C data and yacc definitions (or nothing) % Grammar rules (with actions) % #include "lex.yy.c" C functions, including main()
Compilers: Yacc/7 10 Declarations C data is put between %{ and %} The yacc definitions list the tokens (terminals) used in the grammar %token terminal1 terminal2... Other yacc definitions: – –%left and %right for associativity – –%prec for precedence
Compilers: Yacc/7 11 v v Precedence example: * 5 – –does it mean (2 + 3) * 5 or 2 + (3 * 5) ? v v Associativity example: 1 – 1 – 1 – –does it mean (1 – 1) – 1// left or 1 – (1 – 1) ?// right
Compilers: Yacc/7 12 Rules Rule format: nonterminal : body 1 {action 1} | body 2 {action 2}... | body n {action n) ; Actions are optional; they are C code. Actions are usually at the end of a body, but can be placed anywhere. grammar part is the same as: nonterminal body1 | body2 |... | bodyN
Compilers: Yacc/ Expressions Compiler $ flex expr.l $ bison expr.y $ gcc expr.tab.c -o exprEval expr.l, a flex file expr.y, a bison file bison flexlex.yy.c expr.tab.c gcc exprEval, c executable #include
Compilers: Yacc/7 14 Usage $./exprEval Value = (5 * 2) Value = -8 1 / 3 Value = 0 $ I typed these lines. I typed ctrl-D
Compilers: Yacc/7 15 expr.l % [-+*/()\n]{ return *yytext; } [0-9]* { yylval = atoi(yytext); return(NUMBER); } [ \t] ; /* skip whitespace */ % int yywrap(void) { return 1; } No main() function RE actions usually end with a return. The token is passed to the syntax analyser.
Compilers: Yacc/7 16 Lex File Format Reminder A lex program has three sections: REs and/or C code % RE/action rules % C functions
Compilers: Yacc/7 17 expr.y %token NUMBER % exprs: expr '\n' { printf("Value = %d\n", $1); } | exprs expr '\n' { printf("Value = %d\n", $2); } ; expr: expr '+' term { $$ = $1 + $3; } | expr '-' term { $$ = $1 - $3; } | term { $$ = $1; } ; continued declarations rules attributes
Compilers: Yacc/7 18 term: term '*' factor { $$ = $1 * $3; } | term '/' factor{ $$ = $1 / $3; } /* integer division */ | factor ; factor: '(' expr ')' { $$ = $2; } | NUMBER ; continued more rules
Compilers: Yacc/7 19 $$ #include "lex.yy.c" int yyerror(char *s) { fprintf(stderr, "%s\n", s); return 0; } int main(void) { yyparse(); // the syntax analyzer return 0; } c code
Compilers: Yacc/7 20 Yacc Actions yacc actions (the C code) can use attributes (variables). Each body terminal/non-terminal has an attribute, which contains it's return value.
Compilers: Yacc/7 21 Attributes An attribute is $n, where n is the position of the terminal/non-terminal in the body starting at 1 – –$1 = first terminal/non-terminal of the body – –$2 = second one – –etc. – –$$ = return value for the rule the default value for $$ is the $1 value
Compilers: Yacc/7 22 Evaluation in yacc Stack $ $ 3 $ F $ T $ T * $ T * 5 $ T * F $ T $ E $ E + $ E + 4 $ E + F $ E + T $ E $ E \n $ Es Input 3*5+4\n$ *5+4\n$ *5+4\n$ *5+4\n$ 5+4\n$ +4\n$ +4\n$ +4\n$ +4\n$ 4\n$ \n$ \n$ \n$ \n$ $ $ Action shift reduce F num reduce T F shift shift reduce F num reduce T T * F reduce E T shift shift reduce F num reduce T F reduce E E + T shift reduce Es E \n accept val _ Rule $$ = $1 (implicit) $$ = $1 (implicit) $$ = $1 (implicit) $$ = $1 * $3 $$ = $1 (implicit) $$ = $1 (implicit) $$ = $1 (implicit) $$ = $1 + $3 printf $1 Input: 3 * 5 + 4\n
Compilers: Yacc/ Bottom-up Parsing Reminder Simple expressions grammar: E => E '+' E// rule r1 E => E '*' E// rule r2 E => id// rule r3
Compilers: Yacc/7 24 Parsing "x + y * z" x + y * z // shift x. + y * z // reduce(r3) E. + y * z // shift E +. y * z // shift E + y. * z // reduce(r3) E + E. * z // shift E + E *. z // shift E + E * z. // reduce(r3) E + E * E. // reduce(r2) E + E. // reduce(r1) E. // accept
Compilers: Yacc/7 25 Shift/Reduce Conflict At step 6, a shift or a reduce is possible. 6. E + E. * z // reduce (r1) 7. E. * z : What should be done? – –by default, yacc (bison) shifts
Compilers: Yacc/7 26 Reduce/Reduce Conflict Modify the grammar to include: E => T// new rule r3 E => id// rule r4 T => id// rule r5 continued
Compilers: Yacc/7 27 Consider step 2: x. + y * z There are two ways to reduce: E. + y * z // reduce (r4) or T. + y * z // reduce (r5) What should be done? – –by default, yacc (bison) reduces using the first possible rule (i.e. rule r4)
Compilers: Yacc/7 28 Common Conflicts The two most common shift/reduce problems in prog. languages are: – –expression precedence – –dangling else yacc has features for fixing both of these Reduce/reduce problems are usually due to errors in your grammar.
Compilers: Yacc/7 29 Debugging Conflicts bison can generate extra conflict information, which can help you debug your grammar. – –use the -v option
Compilers: Yacc/ Expression Conflicts %token NUMBER % expr: expr '+' expr | expr '*' expr | '(' expr ')' | NUMBER ; in shiftE.y continued shift/reduce here, as in previous example
Compilers: Yacc/7 31 % #include "lex.yy.c" int yyerror(char *s) { fprintf(stderr, "%s\n", s); return 0; } int main(void) { yyparse(); return 0; }
Compilers: Yacc/7 32 Example When the parsing state is: expr '+' expr. '*' z should bison shift: expr '+' expr '*'. z or reduce?: expr. '*' z // using rule 1
Compilers: Yacc/7 33 Using -v $ bison shiftE.y shiftE.y: conflicts: 4 shift/reduce $ bison -v shiftE.y shiftE.y: conflicts: 4 shift/reduce – –creates a shiftE.output file with extra conflict information
Compilers: Yacc/7 34 Inside shiftE.output State 9 conflicts: 2 shift/reduce State 10 conflicts: 2 shift/reduce Grammar 0 $accept: expr $end 1 expr: expr '+' expr 2 | expr '*' expr 3 | '(' expr ')' 4 | NUMBER : // many state blocks states 9 and 10 are the problems the rules are numbered continued
Compilers: Yacc/7 35 state 9 1 expr: expr. '+' expr 1 | expr '+' expr. 2 | expr. '*' expr '+' shift, and go to state 6 '*' shift, and go to state 7 '+' [reduce using rule 1 (expr)] '*' [reduce using rule 1 (expr)] $default reduce using rule 1 (expr) bison does this but it could do this when bison is looking at these kinds of parsing states continued
Compilers: Yacc/7 36 state 10 1 expr: expr. '+' expr 2 | expr. '*' expr 2 | expr '*' expr. '+' shift, and go to state 6 '*' shift, and go to state 7 '+' [reduce using rule 2 (expr)] '*' [reduce using rule 2 (expr)] $default reduce using rule 2 (expr) bison does this but it could do this when bison is looking at these kinds of parsing states
Compilers: Yacc/7 37 What causes Expression Conflicts? The problems are the precedence and associativity of the operators: – –does * 5 mean (2 + 3) * 5 or 2 + (3 * 5) ? // should be 2nd – –does mean (1 - 1) - 1 or 1 - (1 - 1) ? // should be 1st * should have higher precedence than +, and – should be left associative.
Compilers: Yacc/ Precedence/Associativity in yacc The declarations section can contain associativity and precedence settings for tokens: – –%left, %right, %nonassoc – –precedence is given by the order of the lines Example: %left '+' '-' %left '*' '/' All left associative, with '*' and '/' higher precedence than '+' and '-'.
Compilers: Yacc/7 39 Expressions Variables Compiler $ flex exprVars.l $ bison exprVars.y $ gcc exprVars.tab.c -o exprVarsEval exprVars.l, a flex file exprVars.y, a bison file bison flexlex.yy.c exprVars.tab.c gcc exprVarsEval, c executable #include
Compilers: Yacc/7 40 Usage $./exprVarsEval * 3 Value = Value = -1 a = 3 * 4 a Value = 12 b = (3 - 6) * a b Value = -36 $ I typed these lines. I typed ctrl-D
Compilers: Yacc/7 41 exprVars.l /* Added: RE vars, token names, VAR token, assignment, error msgs */ digits [0-9]+ letter [a-z] % \n return('\n'); \= return(ASSIGN); \+ return(PLUS); \-return(MINUS); \* return(TIMES); \/return(DIV); \( return(LPAREN); \) return(RPAREN); continued the token names are defined in the yacc file
Compilers: Yacc/7 42 {letter} { yylval = *yytext - 'a'; return(VAR); } {digits} { yylval = atoi(yytext); return(NUMBER); } [ \t] ; /* skip whitespace */. yyerror("Invalid char"); /* reject everything else */ % int yywrap(void) { return 1; }
Compilers: Yacc/7 43 exprVars.y /* Added: token names, assoc/precedence ops, changed grammar rules, vars and assignment. */ %token VAR NUMBER ASSIGN PLUS MINUS TIMES DIV LPAREN RPAREN %left PLUS MINUS %left TIMES DIV %{ int symbol[26]; // stores var's values %} % continued
Compilers: Yacc/7 44 program: program statement '\n' | ; statement: expr { printf("Value = %d\n", $1); } | VAR ASSIGN expr { symbol[$1] = $3; } expr: NUMBER | VAR { $$ = symbol[$1]; } | expr PLUS expr { $$ = $1 + $3; } | expr MINUS expr { $$ = $1 - $3; } | expr TIMES expr { $$ = $1 * $3; } | expr DIV expr { $$ = $1 / $3; } /* integer division */ | LPAREN expr RPAREN { $$ = $2; } ; % continued
Compilers: Yacc/7 45 #include "lex.yy.c" int yyerror(char *s) { fprintf(stderr, "%s\n", s); return 0; } int main(void) { yyparse(); return 0; }
Compilers: Yacc/ Dangling Else Conflict %token IF ELSE variable % stmt: expr | if_stmt ; if_stmt: IF expr stmt | IF expr stmt ELSE stmt ; expr: variable ; in iffy.y $ bison -v iffy.y iffy.y: conflicts: 1 shift/reduce
Compilers: Yacc/7 47 Shift or Reduce? Current state: – –IF expr IF expr stmt. ELSE stmt Shift choice: – –IF expr IF expr stmt. ELSE stmt – –IF expr IF expr stmt ELSE. stmt – –IF expr IF expr stmt ELSE stmt. – –IF expr stmt. the second ELSE is paired with the second IF continued if (x < 5) if (x < 3) y = a – b; else y = b – a;
Compilers: Yacc/7 48 Reduce option: – –IF expr IF expr stmt. ELSE stmt – –IF expr stmt. ELSE stmt – –IF expr stmt ELSE. stmt – –IF expr stmt ELSE stmt. the second ELSE is paired with the first IF if (x < 5) if (x < 3) y = a – b; else y = b – a;
Compilers: Yacc/7 49 Inside iffy.output State 8 conflicts: 1 shift/reduce Grammar 0 $accept: stmt $end 1 stmt: expr 2 | if_stmt 3 if_stmt: IF expr stmt 4 | IF expr stmt ELSE stmt 5 expr: variable : // many state blocks continued
Compilers: Yacc/7 50 state 8 3 if_stmt: IF expr stmt. 4 | IF expr stmt. ELSE stmt ELSE shift, and go to state 9 ELSE [reduce using rule 3 (if_stmt)] $default reduce using rule 3 (if_stmt) bison does this but it could do this when bison is looking at these kinds of parsing states
Compilers: Yacc/ Left and Right Recursion A left recursive rule: list: item | list ',' item ; A right recursion rule: list: item | item ',' list Left recusion keeps the parse table stack smaller, so may be a better choice this is the opposite of top-down
Compilers: Yacc/ Error Recovery When an error occurs, yacc/bison calls yyerror() and then terminates. A better approach is to call yyerror(), then try to continue – –this can be done by using the keyword error in the grammar rules
Compilers: Yacc/7 53 Example If there's an error in the stmt rule, then skip the rest of the input tokens until ';'" or '}' is seen, then continue as before: stmt: ';' | expr ';' | VAR '=' expr ';' | '{' stmt_list '}' | error ';' | error '}' ;
Compilers: Yacc/ Embedded Actions Actions can be placed anywhere in a rule, not just at the end: listPair: item1 { do_item1($1); } item2 { do_item2($3); } – –the action variable in the second action block is $3 since the first action block is counted as part of the rule
Compilers: Yacc/ More Information Lex and Yacc by Levine, Mason, and Brown O'Reilly; 2nd edition On UNIX: – –man yacc – –info yacc continued in our library
Compilers: Yacc/7 56 A Compact Guide to Lex & Yacc by Tom Niemann – –with several yacc calculator examples, which I'll be discussing in the next few chapters The Lex & Yacc Page – –documentation and tools continued
Compilers: Yacc/7 57 Compiler Construction using Flex and Bison by Anthony A. Aaby Compiler Construction using Flex and Bison by Anthony A. Aaby –in the "Useful Info" subdirectory of the course website