searchlib.php

<?php  // $Id$

require_once($CFG->libdir.'/lexer.php');

// Constants for the various types of tokens

define("TOKEN_USER","0");
define("TOKEN_META","1");
define("TOKEN_EXACT","2");
define("TOKEN_NEGATE","3");
define("TOKEN_STRING","4");
define("TOKEN_USERID","5");
define("TOKEN_DATEFROM","6");
define("TOKEN_DATETO","7");
define("TOKEN_INSTANCE","8");

// Class to hold token/value pairs after they're parsed.

class search_token {
  var $value;
  var $type;
  function search_token($type,$value){
    $this->type = $type;
    $this->value = $this->sanitize($value);
  
  }

  // Try to clean up user input to avoid potential security issues.
  // Need to think about this some more. 

  function sanitize($userstring){
    return htmlspecialchars(addslashes($userstring));
  }
  function getValue(){  
    return $this->value;
  }
  function getType(){
    return $this->type;
  }
}



// This class does the heavy lifting of lexing the search string into tokens.
// Using a full-blown lexer is probably overkill for this application, but 
// might be useful for other tasks.

class search_lexer extends Lexer{

  function search_lexer(&$parser){

    // Call parent constructor.
    $this->Lexer($parser);

    //Set up the state machine and pattern matches for transitions.

    // Patterns to handle strings  of the form datefrom:foo

    // If we see the string datefrom: while in the base accept state, start
    // parsing a username and go to the indatefrom state.
    $this->addEntryPattern("datefrom:\S+","accept","indatefrom");

    // Snarf everything into the username until we see whitespace, then exit
    // back to the base accept state.
    $this->addExitPattern("\s","indatefrom");


    // Patterns to handle strings  of the form dateto:foo

    // If we see the string dateto: while in the base accept state, start
    // parsing a username and go to the indateto state.
    $this->addEntryPattern("dateto:\S+","accept","indateto");

    // Snarf everything into the username until we see whitespace, then exit
    // back to the base accept state.
    $this->addExitPattern("\s","indateto");


    // Patterns to handle strings  of the form instance:foo

    // If we see the string instance: while in the base accept state, start
    // parsing for instance number and go to the ininstance state.
    $this->addEntryPattern("instance:\S+","accept","ininstance");

    // Snarf everything into the username until we see whitespace, then exit
    // back to the base accept state.
    $this->addExitPattern("\s","ininstance");


    // Patterns to handle strings  of the form userid:foo

    // If we see the string userid: while in the base accept state, start
    // parsing a username and go to the inuserid state.
    $this->addEntryPattern("userid:\S+","accept","inuserid");

    // Snarf everything into the username until we see whitespace, then exit
    // back to the base accept state.
    $this->addExitPattern("\s","inuserid");


    // Patterns to handle strings  of the form user:foo

    // If we see the string user: while in the base accept state, start
    // parsing a username and go to the inusername state.
    $this->addEntryPattern("user:\S+","accept","inusername");

    // Snarf everything into the username until we see whitespace, then exit
    // back to the base accept state.
    $this->addExitPattern("\s","inusername");


    // Patterns to handle strings  of the form meta:foo
 
   // If we see the string meta: while in the base accept state, start
    // parsing a username and go to the inmeta state.
    $this->addEntryPattern("subject:\S+","accept","inmeta");

    // Snarf everything into the meta token until we see whitespace, then exit
    // back to the base accept state.
    $this->addExitPattern("\s","inmeta");

   
    // Patterns to handle required exact match strings (+foo) .

    // If we see a + sign  while in the base accept state, start
    // parsing an exact match string and enter the inrequired state
    $this->addEntryPattern("\+\S+","accept","inrequired");
    // When we see white space, exit back to accept state.
    $this->addExitPattern("\s","inrequired");

    // Handle excluded strings (-foo)

   // If we see a - sign  while in the base accept state, start
    // parsing an excluded string and enter the inexcluded state
    $this->addEntryPattern("\-\S+","accept","inexcluded");
    // When we see white space, exit back to accept state.
    $this->addExitPattern("\s","inexcluded");


    // Patterns to handle quoted strings.

    // If we see a quote  while in the base accept state, start
    // parsing a quoted string and enter the inquotedstring state.
    // Grab everything until we see the closing quote.
  
    $this->addEntryPattern("\"[^\"]+","accept","inquotedstring");

    // When we see a closing quote, reenter the base accept state.
    $this->addExitPattern("\"","inquotedstring");
 
    // Patterns to handle ordinary, nonquoted words.
  
    // When we see non-whitespace, snarf everything into the nonquoted word
    // until we see whitespace again.
    $this->addEntryPattern("\S+","accept","plainstring");

    // Once we see whitespace, reenter the base accept state.
    $this->addExitPattern("\s","plainstring");
  
  }
} 




// This class takes care of sticking the proper token type/value pairs into
// the parsed token  array.
// Most functions in this class should only be called by the lexer, the
// one exception being getParseArray() which returns the result.

class search_parser {
    var $tokens;


    // This function is called by the code that's interested in the result of the parse operation.
    function get_parsed_array(){
        return $this->tokens;
    }

    /*
     * Functions below this are part of the state machine for the parse
     * operation and should not be called directly.
     */

    // Base state. No output emitted.
    function accept() {
        return true;
    }

    // State for handling datefrom:foo constructs. Potentially emits a token.
    function indatefrom($content){
        if (strlen($content) < 10) { // State exit or missing parameter.
            return true;
        }
        // Strip off the datefrom: part and add the reminder to the parsed token array
        $param = trim(substr($content,9));
        $this->tokens[] = new search_token(TOKEN_DATEFROM,$param);
        return true;
    }

    // State for handling dateto:foo constructs. Potentially emits a token.
    function indateto($content){
        if (strlen($content) < 8) { // State exit or missing parameter.
            return true;
        }
        // Strip off the dateto: part and add the reminder to the parsed token array
        $param = trim(substr($content,7));
        $this->tokens[] = new search_token(TOKEN_DATETO,$param);
        return true;
    }

    // State for handling instance:foo constructs. Potentially emits a token.
    function ininstance($content){
        if (strlen($content) < 10) { // State exit or missing parameter.
            return true;
        }
        // Strip off the instance: part and add the reminder to the parsed token array
        $param = trim(substr($content,9));
        $this->tokens[] = new search_token(TOKEN_INSTANCE,$param);
        return true;
    }


    // State for handling userid:foo constructs. Potentially emits a token.
    function inuserid($content){
        if (strlen($content) < 8) { // State exit or missing parameter.
            return true;
        }
        // Strip off the userid: part and add the reminder to the parsed token array
        $param = trim(substr($content,7));
        $this->tokens[] = new search_token(TOKEN_USERID,$param);
        return true;
    }


    // State for handling user:foo constructs. Potentially emits a token.
    function inusername($content){
        if (strlen($content) < 6) { // State exit or missing parameter.
            return true;
        }
        // Strip off the user: part and add the reminder to the parsed token array
        $param = trim(substr($content,5));
        $this->tokens[] = new search_token(TOKEN_USER,$param);
        return true;
    }


    // State for handling meta:foo constructs. Potentially emits a token.
    function inmeta($content){   
        if (strlen($content) < 9) { // Missing parameter.
            return true;
        }
        // Strip off the meta: part and add the reminder to the parsed token array.
        $param = trim(substr($content,8));
        $this->tokens[] = new search_token(TOKEN_META,$param);
        return true;
    }


    // State entered when we've seen a required string (+foo). Potentially
    // emits a token.
    function inrequired($content){
        if (strlen($content) < 2) { // State exit or missing parameter, don't emit.
            return true;
        }
        // Strip off the + sign and add the reminder to the parsed token array.
        $this->tokens[] = new search_token(TOKEN_EXACT,substr($content,1));
        return true;
    } 

    // State entered when we've seen an excluded string (-foo). Potentially 
    // emits a token.
    function inexcluded($content){
        if (strlen($content) < 2) { // State exit or missing parameter.
            return true;
        }
        // Strip off the -sign and add the reminder to the parsed token array.
        $this->tokens[] = new search_token(TOKEN_NEGATE,substr($content,1));
        return true;
    } 


    // State entered when we've seen a quoted string. Potentially emits a token.
    function inquotedstring($content){
        if (strlen($content) < 2) { // State exit or missing parameter.
            return true;
        }
        // Strip off the opening quote and add the reminder to the parsed token array.
        $this->tokens[] = new search_token(TOKEN_STRING,substr($content,1));
        return true;
    } 

    // State entered when we've seen an ordinary, non-quoted word. Potentially
    // emits a token.
    function plainstring($content){
        if (ctype_space($content)) { // State exit
            return true;
        }
        // Add the string to the parsed token array.
        $this->tokens[] = new search_token(TOKEN_STRING,$content);
        return true;
    } 
}

// Primitive function to generate a SQL string from a parse tree
// using TEXT indexes. If searches aren't suitable to use TEXT
// this function calls the default search_generate_SQL() one.
//      
// $parsetree should be a parse tree generated by a 
// search_lexer/search_parser combination. 
// Other fields are database table names to search.
function search_generate_text_SQL($parsetree, $datafield, $metafield, $mainidfield, $useridfield,
                             $userfirstnamefield, $userlastnamefield, $timefield, $instancefield) {
    global $CFG;

/// First of all, search for reasons to switch to standard SQL generation
/// Only mysql are supported for now
    if ($CFG->dbfamily != 'mysql') {
        return search_generate_SQL($parsetree, $datafield, $metafield, $mainidfield, $useridfield,
                                   $userfirstnamefield, $userlastnamefield, $timefield, $instancefield);
    }

/// Some languages don't have "word separators" and MySQL FULLTEXT doesn't perform well with them, so
/// switch to standard SQL search generation
    if ($CFG->dbfamily == 'mysql') {
        $nonseparatedlangs = array('ja_utf8', 'th_utf8', 'zh_cn_utf8', 'zh_tw_utf8');
        if (in_array(current_language(), $nonseparatedlangs)) {
            return search_generate_SQL($parsetree, $datafield, $metafield, $mainidfield, $useridfield,
                                       $userfirstnamefield, $userlastnamefield, $timefield, $instancefield);
        }
    }

/// Here we'll acumulate non-textual tokens
    $non_text_tokens = array();

    $ntokens = count($parsetree);
    if ($ntokens == 0) {
        return "";
    }

    $SQLString = '';
    $text_sql_string = '';

    $datasearch_clause = '';
    $metasearch_clause = '';

    foreach ($parsetree as $token) {

        $type = $token->getType();
        $value = $token->getValue();

        switch($type){
            case TOKEN_STRING:
            /// If it's a multiword token, quote it
                if (strstr($value, ' ')) {
                    $datasearch_clause .= '"' . $value . '" ';
            /// Simple word token, search for it as prefix
                } else {
                    $datasearch_clause .= '+' . $value . '* ';
                }
                break;
            case TOKEN_EXACT:
            /// token must be exactly as requested
                $datasearch_clause .= '+' . $value . ' ';
                break;
            case TOKEN_NEGATE:
            /// token must not exist as prefix
                $datasearch_clause .= '-' . $value . '* ';
                break;
            case TOKEN_META:
            /// token in metafield, search for it as prefix
                $metasearch_clause .= '+' . $value . '* ';
                break;
            case TOKEN_USER:
            case TOKEN_USERID:
            case TOKEN_INSTANCE:
            case TOKEN_DATETO:
            case TOKEN_DATEFROM:
            /// delegate to standard search
                $non_text_tokens[] = $token;
                break;
            default:
                return '';
        }
    }

/// Call to standard search for pending tokens
    if (!empty($non_text_tokens)) {
        $SQLString = search_generate_SQL($non_text_tokens, $datafield, $metafield, $mainidfield, $useridfield,
                                         $userfirstnamefield, $userlastnamefield, $timefield, $instancefield);
    }
/// Build the final SQL clause
    if (!empty($datasearch_clause)) {
    /// Must have $datafield to search within
        if (!empty($datafield)) {
            $text_sql_string .= 'MATCH (' . $datafield;
        /// And optionally $metafield
            if (!empty($metafield)) {
                $text_sql_string .= ', ' . $metafield;
            }
        /// Begin with the AGAINST clause
            $text_sql_string .= ') AGAINST (' . "'";
        /// Add the search terms
            $text_sql_string .= trim($datasearch_clause);
        /// Close AGAINST clause
            $text_sql_string .= "' IN BOOLEAN MODE)";
        }
    }
/// Now add the metasearch_clause
    if (!empty($metasearch_clause)) {
    /// Must have $metafield to search within
        if (!empty($metafield)) {
        /// AND operator if needed
            if (!empty($text_sql_string)) {
                $text_sql_string .= ' AND ';
            }
            $text_sql_string .= 'MATCH (' . $metafield;
        /// Begin with the AGAINST clause
            $text_sql_string .= ') AGAINST (' . "'";
        /// Add the search terms
            $text_sql_string .= trim($metasearch_clause);
        /// Close AGAINST clause
            $text_sql_string .= "' IN BOOLEAN MODE)";
        }
    }
/// Finally add the non-text conditions
    if (!empty($SQLString)) {
    /// AND operator if needed
        if (!empty($text_sql_string)) {
            $text_sql_string .= ' AND ';
        }
        $text_sql_string .= $SQLString;
    }

    return $text_sql_string;
}

// Primitive function to generate a SQL string from a parse tree. 
// Parameters: 
//
// $parsetree should be a parse tree generated by a 
// search_lexer/search_parser combination. 
// Other fields are database table names to search.

function search_generate_SQL($parsetree, $datafield, $metafield, $mainidfield, $useridfield,
                             $userfirstnamefield, $userlastnamefield, $timefield, $instancefield) {
    global $CFG;

    $LIKE = sql_ilike();
    $NOTLIKE = 'NOT ' . $LIKE;
    if ($CFG->dbfamily == "postgres") {
        $REGEXP = "~*";
        $NOTREGEXP = "!~*";
    } else {
        $REGEXP = "REGEXP";
        $NOTREGEXP = "NOT REGEXP";
    }

    $ntokens = count($parsetree);
    if ($ntokens == 0) {
        return "";
    }

    $SQLString = '';

    for ($i=0; $i<$ntokens; $i++){
        if ($i > 0) {// We have more than one clause, need to tack on AND
            $SQLString .= ' AND ';
        }

        $type = $parsetree[$i]->getType();
        $value = $parsetree[$i]->getValue();

    /// Under Oracle and MSSQL, transform TOKEN searches into STRING searches and trim +- chars
        if ($CFG->dbfamily == 'oracle' || $CFG->dbfamily == 'mssql') {
            $value = trim($value, '+-');
            if ($type == TOKEN_EXACT) {
                $type = TOKEN_STRING;
            }
        }

        switch($type){
            case TOKEN_STRING: 
                $SQLString .= "(($datafield $LIKE '%$value%') OR ($metafield $LIKE '%$value%') )";
                break;
            case TOKEN_EXACT: 
                $SQLString .= "(($datafield $REGEXP '[[:<:]]".$value."[[:>:]]') OR ($metafield $REGEXP '[[:<:]]".$value."[[:>:]]'))";
                break; 
            case TOKEN_META: 
                if ($metafield != '') {
                    $SQLString .= "($metafield $LIKE '%$value%')";
                }
                break;
            case TOKEN_USER: 
                $SQLString .= "(($mainidfield = $useridfield) AND (($userfirstnamefield $LIKE '%$value%') OR ($userlastnamefield $LIKE '%$value%')))";
                break; 
            case TOKEN_USERID: 
                $SQLString .= "($useridfield = $value)";
                break; 
            case TOKEN_INSTANCE: 
                $SQLString .= "($instancefield = $value)";
                break; 
            case TOKEN_DATETO: 
                $SQLString .= "($timefield <= $value)";
                break; 
            case TOKEN_DATEFROM: 
                $SQLString .= "($timefield >= $value)";
                break; 
            case TOKEN_NEGATE: 
                $SQLString .= "(NOT (($datafield  $LIKE '%$value%') OR ($metafield  $LIKE '%$value%')))";
                break; 
            default:
                return '';

        } 
    } 
    return $SQLString;
}


?>