// $Id: HqlParser.java 8475 2005-10-27 11:43:19Z oneovthafew $
package org.hibernate.hql.ast;
import java.io.PrintStream;
import java.io.PrintWriter;
import java.io.StringReader;
import antlr.ASTPair;
import antlr.MismatchedTokenException;
import antlr.RecognitionException;
import antlr.Token;
import antlr.TokenStream;
import antlr.TokenStreamException;
import antlr.collections.AST;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.hibernate.hql.antlr.HqlBaseParser;
import org.hibernate.hql.antlr.HqlTokenTypes;
import org.hibernate.hql.ast.util.ASTPrinter;
import org.hibernate.hql.ast.util.ASTUtil;
import org.hibernate.QueryException;
/**
* Implements the semantic action methods defined in the HQL base parser to keep the grammar
* source file a little cleaner. Extends the parser class generated by ANTLR.
*
* @author Joshua Davis (pgmjsd@sourceforge.net)
*/
public final class HqlParser extends HqlBaseParser {
/**
* A logger for this class.
*/
private static final Logger log = LoggerFactory.getLogger( HqlParser.class );
private ParseErrorHandler parseErrorHandler;
private ASTPrinter printer = getASTPrinter();
private static ASTPrinter getASTPrinter() {
return new ASTPrinter( org.hibernate.hql.antlr.HqlTokenTypes.class );
}
public static HqlParser getInstance(String hql) {
// [jsd] The fix for HHH-558...
HqlLexer lexer = new HqlLexer( new StringReader( hql ) );
return new HqlParser( lexer );
}
private HqlParser(TokenStream lexer) {
super( lexer );
initialize();
}
public void reportError(RecognitionException e) {
parseErrorHandler.reportError( e ); // Use the delegate.
}
public void reportError(String s) {
parseErrorHandler.reportError( s ); // Use the delegate.
}
public void reportWarning(String s) {
parseErrorHandler.reportWarning( s );
}
public ParseErrorHandler getParseErrorHandler() {
return parseErrorHandler;
}
/**
* Overrides the base behavior to retry keywords as identifiers.
*
* @param token The token.
* @param ex The recognition exception.
* @return AST - The new AST.
* @throws antlr.RecognitionException if the substitution was not possible.
* @throws antlr.TokenStreamException if the substitution was not possible.
*/
public AST handleIdentifierError(Token token, RecognitionException ex) throws RecognitionException, TokenStreamException {
// If the token can tell us if it could be an identifier...
if ( token instanceof HqlToken ) {
HqlToken hqlToken = ( HqlToken ) token;
// ... and the token could be an identifer and the error is
// a mismatched token error ...
if ( hqlToken.isPossibleID() && ( ex instanceof MismatchedTokenException ) ) {
MismatchedTokenException mte = ( MismatchedTokenException ) ex;
// ... and the expected token type was an identifier, then:
if ( mte.expecting == HqlTokenTypes.IDENT ) {
// Use the token as an identifier.
reportWarning( "Keyword '"
+ token.getText()
+ "' is being interpreted as an identifier due to: " + mte.getMessage() );
// Add the token to the AST.
ASTPair currentAST = new ASTPair();
token.setType( HqlTokenTypes.WEIRD_IDENT );
astFactory.addASTChild( currentAST, astFactory.create( token ) );
consume();
AST identifierAST = currentAST.root;
return identifierAST;
}
} // if
} // if
// Otherwise, handle the error normally.
return super.handleIdentifierError( token, ex );
}
/**
* Returns an equivalent tree for (NOT (a relop b) ), for example:<pre>
* (NOT (GT a b) ) => (LE a b)
* </pre>
*
* @param x The sub tree to transform, the parent is assumed to be NOT.
* @return AST - The equivalent sub-tree.
*/
public AST negateNode(AST x) {
//TODO: switch statements are always evil! We already had bugs because
// of forgotten token types. Use polymorphism for this!
switch ( x.getType() ) {
case OR:
x.setType(AND);
x.setText("{and}");
negateNode( x.getFirstChild() );
negateNode( x.getFirstChild().getNextSibling() );
return x;
case AND:
x.setType(OR);
x.setText("{or}");
negateNode( x.getFirstChild() );
negateNode( x.getFirstChild().getNextSibling() );
return x;
case EQ:
x.setType( NE );
x.setText( "{not}" + x.getText() );
return x; // (NOT (EQ a b) ) => (NE a b)
case NE:
x.setType( EQ );
x.setText( "{not}" + x.getText() );
return x; // (NOT (NE a b) ) => (EQ a b)
case GT:
x.setType( LE );
x.setText( "{not}" + x.getText() );
return x; // (NOT (GT a b) ) => (LE a b)
case LT:
x.setType( GE );
x.setText( "{not}" + x.getText() );
return x; // (NOT (LT a b) ) => (GE a b)
case GE:
x.setType( LT );
x.setText( "{not}" + x.getText() );
return x; // (NOT (GE a b) ) => (LT a b)
case LE:
x.setType( GT );
x.setText( "{not}" + x.getText() );
return x; // (NOT (LE a b) ) => (GT a b)
case LIKE:
x.setType( NOT_LIKE );
x.setText( "{not}" + x.getText() );
return x; // (NOT (LIKE a b) ) => (NOT_LIKE a b)
case NOT_LIKE:
x.setType( LIKE );
x.setText( "{not}" + x.getText() );
return x; // (NOT (NOT_LIKE a b) ) => (LIKE a b)
case IN:
x.setType( NOT_IN );
x.setText( "{not}" + x.getText() );
return x;
case NOT_IN:
x.setType( IN );
x.setText( "{not}" + x.getText() );
return x;
case IS_NULL:
x.setType( IS_NOT_NULL );
x.setText( "{not}" + x.getText() );
return x; // (NOT (IS_NULL a b) ) => (IS_NOT_NULL a b)
case IS_NOT_NULL:
x.setType( IS_NULL );
x.setText( "{not}" + x.getText() );
return x; // (NOT (IS_NOT_NULL a b) ) => (IS_NULL a b)
case BETWEEN:
x.setType( NOT_BETWEEN );
x.setText( "{not}" + x.getText() );
return x; // (NOT (BETWEEN a b) ) => (NOT_BETWEEN a b)
case NOT_BETWEEN:
x.setType( BETWEEN );
x.setText( "{not}" + x.getText() );
return x; // (NOT (NOT_BETWEEN a b) ) => (BETWEEN a b)
/* This can never happen because this rule will always eliminate the child NOT.
case NOT:
return x.getFirstChild(); // (NOT (NOT x) ) => (x)
*/
default:
return super.negateNode( x ); // Just add a 'not' parent.
}
}
/**
* Post process equality expressions, clean up the subtree.
*
* @param x The equality expression.
* @return AST - The clean sub-tree.
*/
public AST processEqualityExpression(AST x) {
if ( x == null ) {
log.warn( "processEqualityExpression() : No expression to process!" );
return null;
}
int type = x.getType();
if ( type == EQ || type == NE ) {
boolean negated = type == NE;
if ( x.getNumberOfChildren() == 2 ) {
AST a = x.getFirstChild();
AST b = a.getNextSibling();
// (EQ NULL b) => (IS_NULL b)
if ( a.getType() == NULL && b.getType() != NULL ) {
return createIsNullParent( b, negated );
}
// (EQ a NULL) => (IS_NULL a)
else if ( b.getType() == NULL && a.getType() != NULL ) {
return createIsNullParent( a, negated );
}
else if ( b.getType() == EMPTY ) {
return processIsEmpty( a, negated );
}
else {
return x;
}
}
else {
return x;
}
}
else {
return x;
}
}
private AST createIsNullParent(AST node, boolean negated) {
node.setNextSibling( null );
int type = negated ? IS_NOT_NULL : IS_NULL;
String text = negated ? "is not null" : "is null";
return ASTUtil.createParent( astFactory, type, text, node );
}
private AST processIsEmpty(AST node, boolean negated) {
node.setNextSibling( null );
// NOTE: Because we're using ASTUtil.createParent(), the tree must be created from the bottom up.
// IS EMPTY x => (EXISTS (QUERY (SELECT_FROM (FROM x) ) ) )
AST ast = createSubquery( node );
ast = ASTUtil.createParent( astFactory, EXISTS, "exists", ast );
// Add NOT if it's negated.
if ( !negated ) {
ast = ASTUtil.createParent( astFactory, NOT, "not", ast );
}
return ast;
}
private AST createSubquery(AST node) {
AST ast = ASTUtil.createParent( astFactory, RANGE, "RANGE", node );
ast = ASTUtil.createParent( astFactory, FROM, "from", ast );
ast = ASTUtil.createParent( astFactory, SELECT_FROM, "SELECT_FROM", ast );
ast = ASTUtil.createParent( astFactory, QUERY, "QUERY", ast );
return ast;
}
public void showAst(AST ast, PrintStream out) {
showAst( ast, new PrintWriter( out ) );
}
private void showAst(AST ast, PrintWriter pw) {
printer.showAst( ast, pw );
}
private void initialize() {
// Initialize the error handling delegate.
parseErrorHandler = new ErrorCounter();
setASTFactory(new HqlASTFactory()); // Create nodes that track line and column number.
}
public void weakKeywords() throws TokenStreamException {
int t = LA( 1 );
switch ( t ) {
case ORDER:
case GROUP:
// Case 1: Multi token keywords GROUP BY and ORDER BY
// The next token ( LT(2) ) should be 'by'... otherwise, this is just an ident.
if ( LA( 2 ) != LITERAL_by ) {
LT( 1 ).setType( IDENT );
if ( log.isDebugEnabled() ) {
log.debug( "weakKeywords() : new LT(1) token - " + LT( 1 ) );
}
}
break;
default:
// Case 2: The current token is after FROM and before '.'.
if (LA(0) == FROM && t != IDENT && LA(2) == DOT) {
HqlToken hqlToken = (HqlToken)LT(1);
if (hqlToken.isPossibleID()) {
hqlToken.setType(IDENT);
if ( log.isDebugEnabled() ) {
log.debug( "weakKeywords() : new LT(1) token - " + LT( 1 ) );
}
}
}
break;
}
}
public void handleDotIdent() throws TokenStreamException {
// This handles HHH-354, where there is a strange property name in a where clause.
// If the lookahead contains a DOT then something that isn't an IDENT...
if (LA(1) == DOT && LA(2) != IDENT) {
// See if the second lookahed token can be an identifier.
HqlToken t = (HqlToken)LT(2);
if (t.isPossibleID())
{
// Set it!
LT( 2 ).setType( IDENT );
if ( log.isDebugEnabled() ) {
log.debug( "handleDotIdent() : new LT(2) token - " + LT( 1 ) );
}
}
}
}
public void processMemberOf(Token n, AST p, ASTPair currentAST) {
AST inAst = n == null ? astFactory.create( IN, "in" ) : astFactory.create( NOT_IN, "not in" );
astFactory.makeASTRoot( currentAST, inAst );
AST ast = createSubquery( p );
ast = ASTUtil.createParent( astFactory, IN_LIST, "inList", ast );
inAst.addChild( ast );
}
static public void panic() {
//overriden to avoid System.exit
throw new QueryException("Parser: panic");
}
}