Fall 2014-2015 Compiler Principles Lecture 5: Parsing part 4 Roman Manevich Ben-Gurion University
Tentative syllabus Front End Intermediate Representation Optimizations Code Generation Scanning Lowering Local Optimizations Register Allocation Top-down Parsing (LL) Dataflow Analysis Instruction Selection Bottom-up Parsing (LR) Loop Optimizations Attribute Grammars mid-term exam 2
Previously LR(0) parsing Running the parser Constructing transition diagram Constructing parser table Detecting conflicts SLR(0) Eliminating SHIFT-REDUCE via FOLLOW sets 3
SLR parsing 4
SLR parsing A handle should not be reduced to a nonterminal N if the lookahead is a token that cannot follow N A reduce item N α is applicable only when the lookahead is in FOLLOW(N) If b is not in FOLLOW(N) we just proved there is no terminating derivation S =>* βnb and thus it is safe to remove the reduce item from the conflicted state Differs from LR(0) only on the ACTION table Now a row in the parsing table may contain both shift actions and reduce actions and we need to consult the current token to decide which one to take 5
SLR action table Lookahead token from the input State id + ( ) [ ] $ state action 0 shift shift q0 shift 1 shift shift q1 shift 2 accept q2 3 shift shift q3 shift 4 E E+T E E+T E E+T q4 E E+T 5 T id T id r5, s6 T id vs. q5 T id 6 E T E T E T q6 E T 7 shift shift q7 shift 8 shift shift q8 shift 9 T (E) T (E) T (E) q9 T E as before T id T id[e] SLR use 1 token look-ahead [ is not in FOLLOW(T) LR(0) no look-ahead 6
Agenda LR(1) LALR(1) Dealing with ambiguity CUP parser generator 7
Going beyond SLR(0) Some common language constructs introduce conflicts even for SLR (0) S S (1) S L = R (2) S R (3) L * R (4) L id (5) R L 8
q0 S S S L = R S R L * R L id R L R S L q3 q1 q2 S R S S S L = R R L = q9 S L = R q6 R * * q4 id id q5 L id id S L = R R L L * R L id L * R R L L * R L id * L q8 L R L R L * R q7 9
shift/reduce conflict (0) S S (1) S L = R (2) S R (3) L * R (4) L id (5) R L q2 S L = R R L = q6 S L = R vs. R L FOLLOW(R) contains = S L = R * R = R SLR cannot resolve conflict S L = R R L L * R L id 10
Inputs requiring shift/reduce (0) S S (1) S L = R (2) S R (3) L * R (4) L id (5) R L q2 S L = R R L = q6 S L = R R L L * R L id For the input id the rightmost derivation S => S => R => L => id requires reducing in q2 For the input id = id S => S => L = R => L = L => L = id => id = id requires shifting 11
LR(1) grammars In SLR: a reduce item N α is applicable only when the lookahead is in FOLLOW(N) But FOLLOW(N) merges lookahead for all alternatives for N Insensitive to the context of a given production LR(1) keeps lookahead with each LR item Idea: a more refined notion of FOLLOW computed per item 12
LR(1) item Already matched To be matched Input N α β, t Hypothesis about αβ being a possible handle, so far we ve matched α, expecting to see β and after reducing N we expect to see the token t 13
LR(1) items LR(1) item is a pair LR(0) item Lookahead token Meaning We matched the part left of the dot, looking to match the part on the right of the dot, followed by the lookahead token Example The production L id yields the following LR(1) items LR(1) items (0) S S (1) S L = R (2) S R (3) L * R (4) L id (5) R L LR(0) items [L id] [L id ] [L id, *] [L id, =] [L id, id] [L id, $] [L id, *] [L id, =] [L id, id] [L id, $] 14
Computing Closure for LR(1) For every [A α Bβ, c] in S for every production B δ and every token b in the grammar such that b FIRST(βc) Add [B δ, b] to S 15
* q0 (S S, $) (S L = R, $) (S R, $) (L * R, = ) (L id, = ) (R L, $ ) (L id, $ ) (L * R, $ ) q4 * (L * R, =) (R L, =) (L * R, =) (L id, =) (L * R, $) (R L, $) (L * R, $) (L id, $) id R id q5 R L S q3 (S R, $) q1 (S S, $) q2 (S L = R, $) (R L, $) (L id, $) (L id, =) q7 L (L * R, =) (L * R, $) q6 q9 (S L = R, $) = R (S L = R, $) (R L, $) (L * R, $) (L id, $) q8 (R L, =) (R L, $) L * q11 (L id, $) id q12 (R L, $) q13 q10 (L * R, $) (R L, $) (L * R, $) (L id, $) R (L * R, $) id 16
Back to the conflict q2 (S L = R, $) (R L, $) = q6 (S L = R, $) (R L, $) (L * R, $) (L id, $) Is there a conflict now? 17
LALR(1) LR(1) tables have huge number of entries Often don t need such refined observation (and cost) Idea: find states with the same LR(0) component and merge their lookaheads component as long as there are no conflicts LALR(1) not as powerful as LR(1) in theory but works quite well in practice Merging may not introduce new shift-reduce conflicts, only reduce-reduce, which is unlikely in practice 18
* q0 (S S, $) (S L = R, $) (S R, $) (L * R, = ) (L id, = ) (R L, $ ) (L id, $ ) (L * R, $ ) q4 * (L * R, =) (R L, =) (L * R, =) (L id, =) (L * R, $) (R L, $) (L * R, $) (L id, $) id R id q5 R L S q3 (S R, $) q1 (S S, $) q2 (S L = R, $) (R L, $) (L id, $) (L id, =) q7 L (L * R, =) (L * R, $) q6 q9 (S L = R, $) = R (S L = R, $) (R L, $) (L * R, $) (L id, $) q8 (R L, =) (R L, $) L * q11 (L id, $) id q12 (R L, $) q13 q10 (L * R, $) (R L, $) (L * R, $) (L id, $) R (L * R, $) id 19
* q0 (S S, $) (S L = R, $) (S R, $) (L * R, = ) (L id, = ) (R L, $ ) (L id, $ ) (L * R, $ ) q4 * (L * R, =) (R L, =) (L * R, =) (L id, =) (L * R, $) (R L, $) (L * R, $) (L id, $) id R id q5 R L S q3 (S R, $) q1 (S S, $) q2 (S L = R, $) (R L, $) (L id, $) (L id, =) q7 L (L * R, =) (L * R, $) q6 q9 (S L = R, $) = R (S L = R, $) (R L, $) (L * R, $) (L id, $) q8 (R L, =) (R L, $) L * q11 (L id, $) id q12 (R L, $) q13 q10 (L * R, $) (R L, $) (L * R, $) (L id, $) R (L * R, $) id 20
* q0 (S S, $) (S L = R, $) (S R, $) (L * R, = ) (L id, = ) (R L, $ ) (L id, $ ) (L * R, $ ) q4 * (L * R, =) (R L, =) (L * R, =) (L id, =) (L * R, $) (R L, $) (L * R, $) (L id, $) id R id q5 R L S q3 (S R, $) q1 (S S, $) q2 (S L = R, $) (R L, $) (L id, $) (L id, =) q7 L id (L * R, =) (L * R, $) q6 q9 (S L = R, $) = R (S L = R, $) (R L, $) (L * R, $) (L id, $) q8 (R L, =) (R L, $) L * q12 (R L, $) q10 (L * R, $) (R L, $) (L * R, $) (L id, $) R id 21
Left/Right- recursion At home: create a simple grammar with leftrecursion and one with right-recursion Construct corresponding LR(0) parser Any conflicts? Run on simple input and observe behavior Attempt to generalize observation for long inputs 22
Automated Parser Generation (via CUP) 23
High-level structure text Lexer spec JFlex.java javac Lexical analyzer ADTL.lex Lexer.java tokens (Token.java) Parser spec CUP.java javac Parser ADTL.cup sym.java Parser.java AST 24
Expression calculator expr expr + expr expr - expr expr * expr expr / expr - expr ( expr ) number Goals of expression calculator parser: Is 2+3+4+5 a valid expression? What is the meaning (value) of this expression? 25
Syntax analysis with CUP CUP parser generator Generates an LALR(1) Parser Input: spec file Output: a syntax analyzer tokens Parser spec CUP.java javac Parser AST 26
CUP spec file Package and import specifications User code components Symbol (terminal and non-terminal) lists Terminals go to sym.java Types of AST nodes Precedence declarations The grammar Semantic actions to construct AST 27
Expression Calculator 1 st Attempt terminal Integer NUMBER; terminal PLUS, MINUS, MULT, DIV; terminal LPAREN, RPAREN; non terminal Integer expr; Symbol type explained later expr ::= expr PLUS expr expr MINUS expr expr MULT expr expr DIV expr MINUS expr LPAREN expr RPAREN NUMBER ; 28
Ambiguities + * * a b c + a + b * c + a b c + + a b c a + b + c + a b c 29
Ambiguities as conflicts for LR(1) + * + * a b c a + b * c + a b c + + a b c a + b + c + a b c 30
terminal Integer NUMBER; terminal PLUS,MINUS,MULT,DIV; terminal LPAREN, RPAREN; terminal UMINUS; non terminal Integer expr; Expression Calculator 2 nd Attempt precedence left PLUS, MINUS; precedence left DIV, MULT; precedence left UMINUS; Increasing precedence expr ::= expr PLUS expr expr MINUS expr expr MULT expr expr DIV expr MINUS expr %prec UMINUS LPAREN expr RPAREN NUMBER ; Contextual precedence 31
Parsing ambiguous grammars using precedence declarations Each terminal assigned with precedence By default all terminals have lowest precedence User can assign his own precedence CUP assigns each production a precedence Precedence of rightmost terminal in production or user-specified contextual precedence On shift/reduce conflict resolve ambiguity by comparing precedence of terminal and production and decides whether to shift or reduce In case of equal precedences left/right help resolve conflicts left means reduce right means shift More information on precedence declarations in CUP s manual 32
Resolving ambiguity precedence left PLUS + + + a b c + a b c a + b + c 33
Resolving ambiguity precedence left PLUS precedence left MULT * + + a b c * a b c a * b + c 34
Resolving ambiguity precedence left MULT MINUS expr %prec UMINUS * - - a b a * b - a * b 35
Resolving ambiguity terminal Integer NUMBER; terminal PLUS,MINUS,MULT,DIV; terminal LPAREN, RPAREN; terminal UMINUS; UMINUS never returned by scanner (used only to define precedence) precedence left PLUS, MINUS; precedence left DIV, MULT; precedence left UMINUS; expr ::= expr PLUS expr expr MINUS expr expr MULT expr expr DIV expr MINUS expr %prec UMINUS LPAREN expr RPAREN NUMBER ; Rule has precedence of UMINUS 36
More CUP directives precedence nonassoc NEQ Non-associative operators: < > ==!= etc. 1<2<3 identified as an error (semantic error?) start non-terminal Specifies start non-terminal other than first non-terminal Can change to test parts of grammar Getting internal representation Command line options: -dump_grammar -dump_states -dump_tables -dump 37
Scanner integration import java_cup.runtime.*; %% %cup Generated from token %eofval{ declarations in.cup file return new Symbol(sym.EOF); %eofval} NUMBER=[0-9]+ %% <YYINITIAL> + { return new Symbol(sym.PLUS); } <YYINITIAL> - { return new Symbol(sym.MINUS); } <YYINITIAL> * { return new Symbol(sym.MULT); } <YYINITIAL> / { return new Symbol(sym.DIV); } <YYINITIAL> ( { return new Symbol(sym.LPAREN); } <YYINITIAL> ) { return new Symbol(sym.RPAREN); } <YYINITIAL>{NUMBER} { return new Symbol(sym.NUMBER, new Integer(yytext())); } <YYINITIAL>\n { } <YYINITIAL>. { } Parser gets terminals from the scanner 38
Recap Package and import specifications and user code components Symbol (terminal and non-terminal) lists Define building-blocks of the grammar Precedence declarations May help resolve conflicts The grammar May introduce conflicts that have to be resolved 39
Assigning meaning expr ::= expr PLUS expr expr MINUS expr expr MULT expr expr DIV expr MINUS expr %prec UMINUS LPAREN expr RPAREN NUMBER ; So far, only validation Add Java code implementing semantic actions 40
Assigning meaning expr ::= expr:e1 PLUS expr:e2 {: RESULT = new Integer(e1.intValue() + e2.intvalue()); :} expr:e1 MINUS expr:e2 {: RESULT = new Integer(e1.intValue() - e2.intvalue()); :} expr:e1 MULT expr:e2 {: RESULT = new Integer(e1.intValue() * e2.intvalue()); :} expr:e1 DIV expr:e2 {: RESULT = new Integer(e1.intValue() / e2.intvalue()); :} MINUS expr:e1 {: RESULT = new Integer(0 - e1.intvalue(); :} %prec UMINUS LPAREN expr:e1 RPAREN {: RESULT = e1; :} NUMBER:n {: RESULT = n; :} ; Symbol labels used to name variables RESULT names the left-hand side symbol 41
Building an AST More useful representation of syntax tree Less clutter Actual level of detail depends on your design Basis for semantic analysis Later annotated with various information Type information Computed values Technically a class hierarchy of abstract syntax tree nodes 42
Parse tree vs. AST expr + expr + expr expr expr 1 + ( 2 ) + ( 3 ) 1 2 3 43
AST hierarchy example expr int_const plus minus times divide 44
AST construction AST Nodes constructed during parsing Stored in push-down stack Bottom-up parser Grammar rules annotated with actions for AST construction When node is constructed all children available (already constructed) Node (RESULT) pushed on stack Top-down parser More complicated 45
AST construction 1 + (2) + (3) expr + (2) + (3) expr + (expr) + (3) expr + (3) expr + (expr) expr expr expr ::= expr:e1 PLUS expr:e2 {: RESULT = new plus(e1,e2); :} LPAREN expr:e RPAREN {: RESULT = e; :} INT_CONST:i {: RESULT = new int_const(, i); :} plus e1 e2 expr plus e1 e2 expr expr expr 1 + ( 2 ) + ( 3 ) int_const val = 1 int_const val = 2 int_const val = 3 46
Example of lists terminal Integer NUMBER; terminal PLUS,MINUS,MULT,DIV,LPAREN,RPAREN,SEMI; terminal UMINUS; non terminal Integer expr; non terminal expr_list, expr_part; precedence left PLUS, MINUS; precedence left DIV, MULT; precedence left UMINUS; expr_list ::= expr_list expr_part expr_part ; expr_part ::= expr:e {: System.out.println("= " + e); :} SEMI ; expr ::= expr PLUS expr expr MINUS expr expr MULT expr expr DIV expr MINUS expr %prec UMINUS LPAREN expr RPAREN NUMBER ; 47
Next lecture: Intermediate Representation