package CQL::Parser; use strict; use warnings; use CQL::Lexer; use CQL::Relation; use CQL::Token; use CQL::TermNode; use CQL::AndNode; use CQL::OrNode; use CQL::NotNode; use CQL::PrefixNode; use CQL::ProxNode; use Carp qw( croak ); our $VERSION = '0.2'; my $lexer; my $token; =head1 NAME CQL::Parser - compiles CQL strings into parse trees of Node subtypes. =head1 SYNOPSIS use CQL::Parser; my $parser = CQL::Parser->new(); my $root = $parser->parse( $cql ); =head1 DESCRIPTION CQL::Parser provides a mechanism to parse Common Query Language (CQL) statements. The best description of CQL comes from the CQL homepage at the Library of Congress L CQL is a formal language for representing queries to information retrieval systems such as web indexes, bibliographic catalogs and museum collection information. The CQL design objective is that queries be human readable and human writable, and that the language be intuitive while maintaining the expressiveness of more complex languages. A CQL statement can be as simple as a single keyword, or as complicated as a set of compoenents indicating search indexes, relations, relational modifiers, proximity clauses and boolean logic. CQL::Parser will parse CQL statements and return the root node for a tree of nodes which describes the CQL statement. This data structure can then be used by a client application to analyze the statement, and possibly turn it into a query for a local repository. Each CQL component in the tree inherits from L and can be one of the following: L, L, L, L, L, L. See the documentation for those modules for their respective APIs. Here are some examples of CQL statements: =over 4 =item * george =item * dc.creator=george =item * dc.creator="George Clinton" =item * clinton and funk =item * clinton and parliament and funk =item * (clinton or bootsy) and funk =item * dc.creator="clinton" and dc.date="1976" =back =head1 METHODS =head2 new() =cut ## for convenience the lexer is located at the package level ## just need to be sure to reinitialize it in very call to parse() sub new { my ( $class, $debug ) = @_; $CQL::DEBUG = $debug ? 1 : 0; return bless { }, ref($class) || $class; } =head2 parse() Pass in a CQL query and you'll get back the root node for the CQL parse tree. If the CQL is invalid an exception will be thrown. =cut sub parse { my ($self,$query) = @_; ## initialize lexer $lexer = CQL::Lexer->new(); debug( "about to parse query: $query" ); ## create the lexer and get the first token $lexer->tokenize( $query ); $token = $lexer->nextToken(); my $root = parseQuery( 'srw.ServerChoice', CQL::Relation->new( 'scr' ) ); if ( $token->getType() != CQL_EOF ) { croak( "junk after end ".$token-getString() ); } return $root; } sub parseQuery { my ( $qualifier, $relation ) = @_; debug( "in parseQuery() with term=" . $token->getString() ); my $term = parseTerm( $qualifier, $relation ); my $type = $token->getType(); while ( $type != CQL_EOF and $type != CQL_RPAREN ) { if ( $type == CQL_AND ) { match($token); my $term2 = parseTerm( $qualifier, $relation ); $term = CQL::AndNode->new( left=>$term, right=>$term2 ); } elsif ( $type == CQL_OR ) { match($token); my $term2 = parseTerm( $qualifier, $relation ); $term = CQL::OrNode->new( left=>$term, right=>$term2 ); } elsif ( $type == CQL_NOT ) { match($token); my $term2 = parseTerm( $qualifier, $relation ); $term = CQL::NotNode->new( left=>$term, right=>$term2 ); } elsif ( $type == CQL_PROX ) { match($token); my $proxNode = CQL::ProxNode->new( $term ); gatherProxParameters( $proxNode ); my $term2 = parseTerm( $qualifier, $relation ); $proxNode->addSecondTerm( $term2 ); $term = $proxNode; } else { croak( "expected boolean got ".$token->getString() ); } $type = $token->getType(); } debug( "no more ops" ); return( $term ); } sub parseTerm { my ( $qualifier, $relation ) = @_; debug( "in parseTerm()" ); my $word; while ( 1 ) { if ( $token->getType() == CQL_LPAREN ) { debug( "parenthesized term" ); match( CQL::Token->new('(') ); my $expr = parseQuery( $qualifier, $relation ); match( CQL::Token->new(')') ); return $expr; } elsif ( $token->getType() == CQL_GT ) { match( $token ); return parsePrefix( $qualifier, $relation ); } debug( "non-parenthesised term" ); $word = matchSymbol( "qualifier or term" ); last if ! isBaseRelation(); $qualifier = $word; debug( "creating relation with word=$word" ); $relation = CQL::Relation->new( $token->getString() ); match( $token ); while ($token->getType() == CQL_MODIFIER ) { match( $token ); my $type = $token->getType(); if ( $type != CQL_RELEVANT and $type != CQL_FUZZY and $type != CQL_STEM and $type != CQL_PHONETIC and $type != CQL_WORD ) { croak( "expected relation modifier got ".$token->getString() ); } if ( $type == CQL_WORD and $token->getString() !~ /\./ ) { croak( "unknown first-class relation modifier: ". $token->getString() ); } $relation->addModifier( $token->getString() ); match( $token ); } } debug( "qualifier=$qualifier relation=$relation term=$word" ); my $node = CQL::TermNode->new( qualifier => $qualifier, relation => $relation, term => $word ); debug( "made term node: ".$node->toCQL() ); return $node; } sub parsePrefix { my ( $qualifier, $relation ) = @_; debug( "prefix mapping" ); my $name = undef; my $identifier = matchSymbol( "prefix name" ); if ( $token->getType() == CQL_EQ ) { match( $token ); $name = $identifier; $identifier = matchSymbol( "prefix identifier" ); } my $node = parseQuery( $qualifier, $relation ); return CQL::PrefixNode->new( name => $name, identifier => $identifier, subtree => $node ); } sub gatherProxParameters { my $node = shift; for (my $i=0; $i<4; $i++ ) { if ( $token->getType() != CQL_MODIFIER ) { ## end of proximity parameters return; } match($token); if ( $token->getType() != CQL_MODIFIER ) { if ( $i==0 ) { gatherProxRelation($node); } elsif ( $i==1 ) { gatherProxDistance($node); } elsif ( $i==2 ) { gatherProxUnit($node); } elsif ( $i==3 ) { gatherProxOrdering($node); } } } } sub gatherProxRelation { my $node = shift; if ( ! isProxRelation() ) { croak( "expected proximity relation got ".$token->getString() ); } $node->addModifier( "relation", $token->getString() ); match( $token ); debug( "gatherProxRelation matched ".$token->getString() ); } sub gatherProxDistance { my $node = shift; if ( $token->getString() !~ /^\d+$/ ) { croak( "expected proximity distance got ".$token->getString() ); } $node->addModifier( "distance", $token->getString() ); match( $token ); debug( "gatherProxDistance matched ".$token->getString() ); } sub gatherProxUnit { my $node = shift; my $type = $token->getType(); if( $type != CQL_PWORD and $type != CQL_SENTENCE and $type != CQL_PARAGRAPH and $type != CQL_ELEMENT ) { croak( "expected proximity unit got ".$token->getString() ); } $node->addModifier( "unit", $token->getString() ); match( $token ); debug( "gatherProxUnit matched ".$token->getString() ); } sub gatherProxOrdering { my $node = shift; my $type = $token->getType(); if ( $type != CQL_ORDERED and $type != CQL_UNORDERED ) { croak( "expected proximity ordering got ".$token->getString() ); } $node->addModifier( "ordering", $token->getString() ); match( $token ); } sub isBaseRelation { debug( "inside base relation: checking ttype=".$token->getType()." sval=". $token->getString() ); if( $token->getType() == CQL_WORD and $token->getString() !~ /\./ ) { croak( "unknown first class relation: ".$token->getString() ); } my $type = $token->getType(); return( isProxRelation() or $type==CQL_ANY or $type==CQL_ALL or $type==CQL_EXACT or $type==CQL_SCR or $type==CQL_WORD ); } sub isProxRelation { debug( "isProxRelation: checking ttype=".$token->getType()." sval=". $token->getString() ); my $type = $token->getType(); return( $type==CQL_LT or $type==CQL_GT or $type==CQL_EQ or $type==CQL_LE or $type==CQL_GE or $type==CQL_NE ); } sub match { my $expected = shift; debug( "in match(".$expected->getString().")" ); if ( $token->getType() != $expected->getType() ) { croak( "expected ".$expected->getString() . " but got " . $token->getString() ); } $token = $lexer->nextToken(); debug( "got token type=".$token->getType()." string=".$token->getString() ); } sub matchSymbol { debug( "in match symbol" ); my $return = $token->getString(); match( $token ); return $return; } sub debug { return unless $CQL::DEBUG; print STDERR "CQL::Parser: ", shift, "\n"; } =head1 XCQL CQL has an XML representation which you can generate from a CQL parse tree. Just call the toXCQL() method on the root node you get back from a call to parse(). =head1 ERRORS AND DIAGNOSTICS As mentioned above, a CQL syntax error will result in an exception being thrown. So if you have any doubts about the CQL that you are parsing you should wrap the call to parse() in an eval block, and check $@ afterwards to make sure everything went ok. eval { my $node = $parser->parse( $cql ); }; if ( $@ ) { print "uhoh, exception $@\n"; } If you'd like to see blow by blow details while your CQL is being parsed set $CQL::DEBUG equal to 1, and you will get details on STDERR. This is useful if the parse tree is incorrect and you want to locate where things are going wrong. Hopefully this won't happen, but if it does please notify the author. =head1 TODO =over 4 =item * toSwish() convert CQL to swish queries =item * toLucene() convert CQL to lucene queries =item * toYourEngineHere() please feel free to add functionality and send in patches! =back =head1 THANKYOUS CQL::Parser is essentially a Perl port of Mike Taylor's cql-java package http://zing.z3950.org/cql/java/. Mike and IndexData were kind enough to allow the author to write this port, and to make it available under the terms of the Artistic License. Thanks Mike! The CQL::Lexer package relies heavily on Stevan Little's excellent String::Tokenizer. Thanks Stevan! CQL::Parser was developed as a component of the Ockham project, which is funded by the National Science Foundation. See http://www.ockham.org for more information about Ockham. =head1 AUTHOR =over 4 =item * Ed Summers =cut 1;