Parsing shift-reduce conflict

Question

I am using JavaCUP tool in order to produce a parser for my language. I am trying to write a grammar that matches nested and multiple if_else statements.

Input file

// matches 
if () 
    if ()

    else

    if ()

    else
else

// no match -> modifying grammar leads to shift/reduce conflict
if ()

else

Grammar

expr ::=  if_then_else_statement;

if_then_else_statement ::= IF LPAREN RPAREN if_then_else_statement ELSE if_then_else_statement
                        | ;

This grammar matches nested if_else statements. However it only recognizes the first nested if_else statement of my input file.

I modified my grammar in order to match multiple statements like this:

expr ::=  expr if_then_else_statement;
      | ;

if_then_else_statement ::= IF LPAREN RPAREN if_then_else_statement ELSE if_then_else_statement
                        | ;

The result was a shift/reduce conflict caused by the empty rule (I guess). How can I modify it to support both nested and multiple if_else statements without the use of precedence?

Yes, after modifying it, expr could reduce to expr if_then_else_statement which could reduce to expr if_then_else_statement if_then_else_statement and finally to if_then_else_statement if_then_else_statement. But this works only in theory — pirox22, Mar 14 '16 at 17:22

rici · Answer 1 · 2016-03-15T05:09:57.250

0

The usual solution would be something like this:

expr_list ::= | expr_list expr ;
expr ::= if_then_else_statement ;
if_then_else_statement ::= IF LPAREN RPAREN expr ELSE expr ;

That doesn't allow empty else clauses (or empty then clauses) because an empty else clause is ambiguous: there is no way to tell whether the s2 in if () s1 else s2 is the else clause or a statement which follows a complete if statement with an empty else clause.

To resolve that ambiguity, you need statement terminators (eg. semicolon) or if statement terminators (fi and end are common choices) or something else.

edited Mar 15 '16 at 05:09

answered Mar 14 '16 at 17:34

rici

234,347
28
237
341

This doesn't work either .Warning : *** Shift/Reduce conflict found in state #10 between if_then_else_statement ::= IF LPAREN RPAREN expr_list ELSE expr_list (*) and if_then_else_statement ::= (*) IF LPAREN RPAREN expr_list ELSE expr_list under symbol IF Resolved in favor of shifting. – pirox22 Mar 14 '16 at 17:53
Replaced the stupid unworkable answer with an explanation of the ambiguity. Ambiguous grammars are hard to parse unless your computer has a mind-reading interface, which you s still have uncommon. Good luck. I'm afk until next week, probably. – rici Mar 15 '16 at 05:13

Parsing shift-reduce conflict

1 Answers1