Nested Boolean Expression Parser using ANTLR

Question

I'm trying to parse a Nested Boolean Expression and get the individual conditions within the expression separately. For e.g., if the input string is:

(A = a OR B = b OR C = c AND ((D = d AND E = e) OR (F = f AND G = g)))

I would like to get the conditions with the correct order. i.e.,

D =d AND E = e OR F = f AND G = g AND A = a OR B = b OR C = c

I'm using ANTLR 4 to parse the input text and here's my grammar:

grammar SimpleBoolean;

rule_set : nestedCondition* EOF;

AND : 'AND' ;
OR  : 'OR' ;
NOT : 'NOT';

TRUE  : 'TRUE' ;
FALSE : 'FALSE' ;

GT : '>' ;
GE : '>=' ;
LT : '<' ;
LE : '<=' ;
EQ : '=' ;

LPAREN : '(' ;
RPAREN : ')' ;

DECIMAL : '-'?[0-9]+('.'[0-9]+)? ;

IDENTIFIER : [a-zA-Z_][a-zA-Z_0-9]* ;

WS : [ \r\t\u000C\n]+ -> skip;

nestedCondition : LPAREN condition+ RPAREN (binary nestedCondition)*;
condition: predicate (binary predicate)*
            | predicate (binary component)*;
component: predicate | multiAttrComp;
multiAttrComp : LPAREN predicate (and predicate)+ RPAREN;
predicate : IDENTIFIER comparator IDENTIFIER;
comparator : GT | GE | LT | LE | EQ ;
binary: AND | OR ;
unary: NOT;
and: AND;

And here's the Java Code that I'm using to parse it:

ANTLRInputStream inputStr = new ANTLRInputStream(input);
SimpleBooleanLexer lexer = new SimpleBooleanLexer(inputStr);
TokenStream tokens = new CommonTokenStream(lexer);
SimpleBooleanParser parser = new SimpleBooleanParser(tokens);
parser.getBuildParseTree();
ParseTree tree = parser.rule_set();
System.out.println(tree.toStringTree(parser));

The output is:

(rule_set (nestedCondition ( (condition (predicate A (comparator =) a) (binary OR) (component (predicate B (comparator =) b)) (binary OR) (component (predicate C (comparator =) c)) (binary AND) (component (multiAttrComp ( (predicate ( D (comparator =) d) (and AND) (predicate E (comparator =) e) ))) (binary OR) (component (multiAttrComp ( (predicate F (comparator =) f) (and AND) (predicate G (comparator =) g) )))) ) )) <EOF>)

I'm looking for help on how to parse this tree to get the conditions in the correct order? In ANTLR 3, we could specify ^ and ! to decide how the tree is built(refer this thread), but I learnt that this is not supported in ANTLR 4.

Can someone suggest how I can parse the String in the correct order in Java using the ParseTree created by ANTLR?

Archdiocese

@BartKiers Do you have any idea how to solve this problem? – Jotter

@BartKiers You're right. GT | GE | LT | LE | EQ all have the same precedence and they should be evaluated before AND | OR. The parsing should be based on the brackets ( ). What I'm looking for is help on how to parse the String in Java using the ParseTree shown in the code above. – Archdiocese

In our case, whenever there is AND between two components, it would always be inside brackets ( ). – Archdiocese

In your example, (A = a OR B = b OR C = c AND ((D = d AND E = e) OR (F = f AND G = g))), the first AND is not inside parens. Will A = a OR B = b OR C = c be evaluated first, or will C = c AND ((D = d AND E = e) OR (F = f AND G = g)) first be evaluated? – Wolford

Answer 1

I'd just wrap all the expressions into a single expression rule. Be sure to define the comparator expressions alternative before your binary expression alternative to make sure comparator operators bind more tightly than OR and AND:

grammar SimpleBoolean;

parse
 : expression EOF
 ;

expression
 : LPAREN expression RPAREN                       #parenExpression
 | NOT expression                                 #notExpression
 | left=expression op=comparator right=expression #comparatorExpression
 | left=expression op=binary right=expression     #binaryExpression
 | bool                                           #boolExpression
 | IDENTIFIER                                     #identifierExpression
 | DECIMAL                                        #decimalExpression
 ;

comparator
 : GT | GE | LT | LE | EQ
 ;

binary
 : AND | OR
 ;

bool
 : TRUE | FALSE
 ;

AND        : 'AND' ;
OR         : 'OR' ;
NOT        : 'NOT';
TRUE       : 'TRUE' ;
FALSE      : 'FALSE' ;
GT         : '>' ;
GE         : '>=' ;
LT         : '<' ;
LE         : '<=' ;
EQ         : '=' ;
LPAREN     : '(' ;
RPAREN     : ')' ;
DECIMAL    : '-'? [0-9]+ ( '.' [0-9]+ )? ;
IDENTIFIER : [a-zA-Z_] [a-zA-Z_0-9]* ;
WS         : [ \r\t\u000C\n]+ -> skip;

To test the grammar above, use the following quick-and-dirty test classes:

public class Main {

  public static void main(String[] args) throws Exception {

    Map<String, Object> variables = new HashMap<String, Object>() {{
      put("A", true);
      put("a", true);
      put("B", false);
      put("b", false);
      put("C", 42.0);
      put("c", 42.0);
      put("D", -999.0);
      put("d", -1999.0);
      put("E", 42.001);
      put("e", 142.001);
      put("F", 42.001);
      put("f", 42.001);
      put("G", -1.0);
      put("g", -1.0);
    }};

    String[] expressions = {
        "1 > 2",
        "1 >= 1.0",
        "TRUE = FALSE",
        "FALSE = FALSE",
        "A OR B",
        "B",
        "A = B",
        "c = C",
        "E > D",
        "B OR (c = B OR (A = A AND c = C AND E > D))",
        "(A = a OR B = b OR C = c AND ((D = d AND E = e) OR (F = f AND G = g)))"
    };

    for (String expression : expressions) {
      SimpleBooleanLexer lexer = new SimpleBooleanLexer(new ANTLRInputStream(expression));
      SimpleBooleanParser parser = new SimpleBooleanParser(new CommonTokenStream(lexer));
      Object result = new EvalVisitor(variables).visit(parser.parse());
      System.out.printf("%-70s -> %s\n", expression, result);
    }
  }
}

class EvalVisitor extends SimpleBooleanBaseVisitor<Object> {

  private final Map<String, Object> variables;

  public EvalVisitor(Map<String, Object> variables) {
    this.variables = variables;
  }

  @Override
  public Object visitParse(SimpleBooleanParser.ParseContext ctx) {
    return super.visit(ctx.expression());
  }

  @Override
  public Object visitDecimalExpression(SimpleBooleanParser.DecimalExpressionContext ctx) {
    return Double.valueOf(ctx.DECIMAL().getText());
  }

  @Override
  public Object visitIdentifierExpression(SimpleBooleanParser.IdentifierExpressionContext ctx) {
    return variables.get(ctx.IDENTIFIER().getText());
  }

  @Override
  public Object visitNotExpression(SimpleBooleanParser.NotExpressionContext ctx) {
    return !((Boolean)this.visit(ctx.expression()));
  }

  @Override
  public Object visitParenExpression(SimpleBooleanParser.ParenExpressionContext ctx) {
    return super.visit(ctx.expression());
  }

  @Override
  public Object visitComparatorExpression(SimpleBooleanParser.ComparatorExpressionContext ctx) {
    if (ctx.op.EQ() != null) {
      return this.visit(ctx.left).equals(this.visit(ctx.right));
    }
    else if (ctx.op.LE() != null) {
      return asDouble(ctx.left) <= asDouble(ctx.right);
    }
    else if (ctx.op.GE() != null) {
      return asDouble(ctx.left) >= asDouble(ctx.right);
    }
    else if (ctx.op.LT() != null) {
      return asDouble(ctx.left) < asDouble(ctx.right);
    }
    else if (ctx.op.GT() != null) {
      return asDouble(ctx.left) > asDouble(ctx.right);
    }
    throw new RuntimeException("not implemented: comparator operator " + ctx.op.getText());
  }

  @Override
  public Object visitBinaryExpression(SimpleBooleanParser.BinaryExpressionContext ctx) {
    if (ctx.op.AND() != null) {
      return asBoolean(ctx.left) && asBoolean(ctx.right);
    }
    else if (ctx.op.OR() != null) {
      return asBoolean(ctx.left) || asBoolean(ctx.right);
    }
    throw new RuntimeException("not implemented: binary operator " + ctx.op.getText());
  }

  @Override
  public Object visitBoolExpression(SimpleBooleanParser.BoolExpressionContext ctx) {
    return Boolean.valueOf(ctx.getText());
  }

  private boolean asBoolean(SimpleBooleanParser.ExpressionContext ctx) {
    return (boolean)visit(ctx);
  }

  private double asDouble(SimpleBooleanParser.ExpressionContext ctx) {
    return (double)visit(ctx);
  }
}

Running the Main class will result in the following output:

1 > 2                                                                  -> false
1 >= 1.0                                                               -> true
TRUE = FALSE                                                           -> false
FALSE = FALSE                                                          -> true
A OR B                                                                 -> true
B                                                                      -> false
A = B                                                                  -> false
c = C                                                                  -> true
E > D                                                                  -> true
B OR (c = B OR (A = A AND c = C AND E > D))                            -> true
(A = a OR B = b OR C = c AND ((D = d AND E = e) OR (F = f AND G = g))) -> true

Wolford

Thanks @BartKiers. I'll try this and get back to you shortly. – Archdiocese

I'm getting errors like incompatible types: double cannot be converted to Object – Flavouring

@johnktejik a little late (I just saw your comment), but I just tested it with ANTLR 4.7.1 and it works like a charm. I get the exact output as in my answer of 3 years ago. Perhaps you're compiling with an old(er) Java version? I recommend using Java 8. – Wolford