<?php
/**
 * A text search query, converting queries to PCRE and SQL matchers.
 *
 * This represents an enhanced "Google-like" text search query:
 * <dl>
 * <dt> default: case-insensitive glob-style search with special operators OR AND NOT -
 * <dt> wiki -test
 *   <dd> Match strings containing the substring 'wiki', and NOT containing the
 *        substring 'test'.
 * <dt> wiki word or page
 *   <dd> Match strings containing the substring 'wiki' AND either the substring
 *        'word' OR the substring 'page'.
 * <dt> auto-detect regex hints, glob-style or regex-style, and converts them
 *      to PCRE and SQL matchers:
 *   <dd> "^word$" => EXACT(phrase)
 *   <dd> "^word"  => STARTS_WITH(phrase)
 *   <dd> "word$"  => ENDS_WITH(phrase)
 *   <dd> "^word" ... => STARTS_WITH(word)
 *   <dd> "word$" ... => ENDS_WITH(word)
 *   <dd> "word*"  => STARTS_WITH(word)
 *   <dd> "*word"  => ENDS_WITH(word)
 *   <dd> "/^word.* /" => REGEX(^word.*)
 *   <dd> "word*word" => REGEX(word.*word)
 * </dl>
 *
 * The full query syntax, in order of precedence, is roughly:
 *
 * The unary 'NOT' or '-' operator (they are equivalent) negates the
 * following search clause.
 *
 * Search clauses may be joined with the (left-associative) binary operators
 * 'AND' and 'OR'. (case-insensitive)
 *
 * Two adjoining search clauses are joined with an implicit 'AND'.  This has
 * lower precedence than either an explicit 'AND' or 'OR', so "a b OR c"
 * parses as "a AND ( b OR c )", while "a AND b OR c" parses as
 * "( a AND b ) OR c" (due to the left-associativity of 'AND' and 'OR'.)
 *
 * Search clauses can be grouped with parentheses.
 *
 * Phrases (or other things which don't look like words) can be forced to
 * be interpreted as words by quoting them, either with single (') or double (")
 * quotes.  If you wan't to include the quote character within a quoted string,
 * double-up on the quote character: 'I''m hungry' is equivalent to
 * "I'm hungry".
 *
 * Force regex on "re:word" => posix-style, "/word/" => pcre-style
 * or use regex='glob' to use file wildcard-like matching. (not yet)
 *
 * The parsed tree is then converted to the needed PCRE (highlight,
 * simple backends) or SQL functions.
 *
 * @author: Jeff Dairiki
 * @author: Reini Urban (case and regex detection, enhanced sql callbacks)
 */

// regex-style: 'auto', 'none', 'glob', 'posix', 'pcre', 'sql'
define ('TSQ_REGEX_NONE', 0);
define ('TSQ_REGEX_AUTO', 1);
define ('TSQ_REGEX_POSIX', 2);
define ('TSQ_REGEX_GLOB', 4);
define ('TSQ_REGEX_PCRE', 8);
define ('TSQ_REGEX_SQL', 16);

define ('TSQ_TOK_VOID', 0);
define ('TSQ_TOK_BINOP', 1);
define ('TSQ_TOK_NOT', 2);
define ('TSQ_TOK_LPAREN', 4);
define ('TSQ_TOK_RPAREN', 8);
define ('TSQ_TOK_WORD', 16);
define ('TSQ_TOK_STARTS_WITH', 32);
define ('TSQ_TOK_ENDS_WITH', 64);
define ('TSQ_TOK_EXACT', 128);
define ('TSQ_TOK_REGEX', 256);
define ('TSQ_TOK_REGEX_GLOB', 512);
define ('TSQ_TOK_REGEX_PCRE', 1024);
define ('TSQ_TOK_REGEX_SQL', 2048);
define ('TSQ_TOK_ALL', 4096);
// all bits from word to the last.
define ('TSQ_ALLWORDS', (4096 * 2) - 1 - (16 - 1));

class TextSearchQuery
{
    private $sql_clause_cb;
    private $word;

    /**
     * Create a new query.
     *
     * @param $search_query string The query.  Syntax is as described above.
     * Note that an empty $search_query will match anything.
     * @param $case_exact boolean
     * @param $regex string one of 'auto', 'none', 'glob', 'posix', 'pcre', 'sql'
     * @see TextSearchQuery
     */
    function TextSearchQuery($search_query, $case_exact = false, $regex = 'auto')
    {
        if ($regex == 'none' or !$regex)
            $this->_regex = 0;
        elseif (defined("TSQ_REGEX_" . strtoupper($regex)))
            $this->_regex = constant("TSQ_REGEX_" . strtoupper($regex)); else {
            trigger_error(fmt("Unsupported argument: %s=%s", 'regex', $regex));
            $this->_regex = 0;
        }
        $this->_regex_modifier = ($case_exact ? '' : 'i') . 'sS';
        $this->_case_exact = $case_exact;
        if ($regex != 'pcre') {
            $parser = new TextSearchQuery_Parser;
            $this->_tree = $parser->parse($search_query, $case_exact, $this->_regex);
            $this->optimize(); // broken under certain circumstances: "word -word -word"
            if (defined("FULLTEXTSEARCH_STOPLIST"))
                $this->_stoplist = FULLTEXTSEARCH_STOPLIST;
            else // default stoplist, localizable.
                $this->_stoplist = _("(A|An|And|But|By|For|From|In|Is|It|Of|On|Or|The|To|With)");
        } else {
            $this->_tree = new TextSearchQuery_node_regex_pcre($search_query);
            if (preg_match("/^\/(.*)\/(\w*)$/", $search_query, $m)) {
                $this->_tree->word = $m[1];
                $this->_regex_modifier = $m[2]; // overrides case_exact
            }
        }
    }

    function getType()
    {
        return 'text';
    }

    private function optimize()
    {
        $this->_tree = $this->_tree->optimize();
    }

    /**
     * Get a PCRE regexp which matches the query.
     */
    function asRegexp()
    {
        if (!isset($this->_regexp)) {
            if (!isset($this->_regex_modifier))
                $this->_regex_modifier = ($this->_case_exact ? '' : 'i') . 'sS';
            if ($this->_regex)
                $this->_regexp = '/' . $this->_tree->regexp() . '/' . $this->_regex_modifier;
            else
                $this->_regexp = '/^' . $this->_tree->regexp() . '/' . $this->_regex_modifier;
        }
        return $this->_regexp;
    }

    /**
     * Match query against string.
     * EXACT ("Term") ignores the case_exact setting.
     *
     * @param $string string The string to match.
     * @return boolean True if the string matches the query.
     */
    function match($string)
    {
        if ($this->_tree->_op == TSQ_TOK_ALL) return true;
        if ($this->_tree->_op == TSQ_TOK_EXACT) return $this->_tree->word == $string;
        return preg_match($this->asRegexp(), $string);
    }

    /* How good does it match? Returns a number */
    function score($string)
    {
        $score = 0.0;
        $i = 10;
        foreach (array_unique($this->_tree->highlight_words()) as $word) {
            if ($nummatch = preg_match_all("/" . preg_quote($word, '/') . "/" .
                    $this->_regex_modifier,
                $string, $out)
            )
                $score += ($i-- * $nummatch);
        }
        return min(1.0, $score / 10.0);
    }

    /**
     * Get a regular expression suitable for highlighting matched words.
     *
     * This returns a PCRE regular expression which matches any non-negated
     * word in the query.
     *
     * @return string The PCRE regexp.
     */
    function getHighlightRegexp()
    {
        if (!isset($this->_hilight_regexp)) {
            $words = array_unique($this->_tree->highlight_words());
            if (!$words) {
                $this->_hilight_regexp = false;
            } else {
                foreach ($words as $key => $word)
                    $words[$key] = preg_quote($word, '/');
                $this->_hilight_regexp = '(?' . ($this->_case_exact ? '' : 'i') . ':'
                    . join('|', $words) . ')';
            }
        }
        return $this->_hilight_regexp;
    }

    /** Get away with the callback and use a db-specific search class instead.
     * @see WikiDB_backend_PearDB_search
     */
    function makeSqlClauseObj(&$sql_search_cb)
    {
        $this->sql_clause_cb = $sql_search_cb;
        return $this->sql_clause_obj($this->_tree);
    }

    private function sql_clause_obj($node)
    {
        switch ($node->_op) {
            case TSQ_TOK_NOT:
                return "NOT (" . $this->sql_clause_cb->call($node->leaves[0]) . ")";
            case TSQ_TOK_BINOP:
                $subclauses = array();
                foreach ($node->leaves as $leaf)
                    $subclauses[] = "(" . $this->sql_clause_obj($leaf) . ")";
                return join(" $node->op ", $subclauses);
            case TSQ_TOK_VOID:
                return '0=1';
            case TSQ_TOK_ALL:
                return '1=1';
            default:
                return $this->sql_clause_cb->call($node);
        }
    }

    /*
     postgresql tsearch2 uses no WHERE operators, just & | and ! in the searchstring
     */
    function makeTsearch2SqlClauseObj(&$sql_search_cb)
    {
        $this->sql_clause_cb = $sql_search_cb;
        return $this->Tsearch2Sql_clause_obj($this->_tree);
    }

    private function Tsearch2Sql_clause_obj($node)
    {
        // TODO: "such a phrase"
        switch ($node->_op) {
            case TSQ_TOK_NOT:
                return "!" . $node->leaves[0];
            case TSQ_TOK_BINOP:
                $subclauses = array();
                foreach ($node->leaves as $leaf)
                    $subclauses[] = $this->Tsearch2Sql_clause_obj($leaf);
                return join($node->_op == 'OR' ? "|" : "&", $subclauses);
            case TSQ_TOK_VOID:
                return '';
            case TSQ_TOK_ALL:
                return '1';
            default:
                return $this->sql_clause_cb->call($node);
        }
    }

    function sql()
    {
        return '%' . $this->_sql_quote($this->word) . '%';
    }

    /**
     * Get printable representation of the parse tree.
     *
     * This is for debugging only.
     * @return string Printable parse tree.
     */
    function asString()
    {
        return $this->as_string($this->_tree);
    }

    private function as_string($node, $indent = '')
    {
        switch ($node->_op) {
            case TSQ_TOK_WORD:
                return $indent . "WORD: $node->word";
            case TSQ_TOK_VOID:
                return $indent . "VOID";
            case TSQ_TOK_ALL:
                return $indent . "ALL";
            default:
                $lines = array($indent . $node->op . ":");
                $indent .= "  ";
                foreach ($node->leaves as $leaf)
                    $lines[] = $this->as_string($leaf, $indent);
                return join("\n", $lines);
        }
    }
}

/**
 * This is a TextSearchQuery which matches nothing.
 */
class NullTextSearchQuery extends TextSearchQuery
{
    /**
     * Create a new query.
     *
     * @see TextSearchQuery
     */
    function NullTextSearchQuery()
    {
    }

    function asRegexp()
    {
        return '/^(?!a)a/x';
    }

    function match($string)
    {
        return false;
    }

    function getHighlightRegexp()
    {
        return "";
    }

    function asString()
    {
        return "NullTextSearchQuery";
    }
}

/**
 * A simple algebraic matcher for numeric attributes.
 *  NumericSearchQuery can do ("population < 20000 and area > 1000000", array("population", "area"))
 *  ->match(array('population' => 100000, 'area' => 10000000))
 *
 * Supports all mathematical PHP comparison operators, plus ':=' for equality.
 *   "(x < 2000000 and x >= 10000) or (x >= 100 and x < 2000)"
 *   "x := 100000" is the same as "x == 100000"
 *
 * Since this is basic numerics only, we simply try to get away with
 * replacing the variable values at the right positions and do an eval then.
 *
 * @package NumericSearchQuery
 * @author Reini Urban
 * @see SemanticAttributeSearchQuery
 */
class NumericSearchQuery
{
    protected $bound;
    protected $workquery;

    /**
     * Create a new query.
     *   NumericSearchQuery("population > 20000 or population < 200", "population")
     *   NumericSearchQuery("population < 20000 and area > 1000000", array("population", "area"))
     *
     * With a single variable it is easy: The valid name must be matched elsewhere, just
     * replace the given number in match in the query.
     *   ->match(2000)
     *
     * With matching a struct we need strict names, no * as name is allowed.
     * So always when the placeholder is an array, the names of the target struct must match
     * and all vars be defined. Use the method can_match($struct) therefore.
     *
     * @access public
     * @param $search_query string   A numerical query with placeholders as variable.
     * @param $placeholders array or string  All placeholders in the query must be defined
     *     here, and will be replaced by the matcher.
     */
    function NumericSearchQuery($search_query, $placeholders)
    {
        // added some basic security checks against user input
        $this->_query = $search_query;
        $this->_placeholders = $placeholders;

        // we should also allow the M_ constants
        $this->_allowed_functions = explode(':', 'abs:acos:acosh:asin:asinh:atan2:atan:atanh:base_convert:bindec:ceil:cos:cosh:decbin:dechex:decoct:deg2rad:exp:expm1:floor:fmod:getrandmax:hexdec:hypot:is_finite:is_infinite:is_nan:lcg_value:log10:log1p:log:max:min:mt_getrandmax:mt_rand:mt_srand:octdec:pi:pow:rad2deg:rand:round:sin:sinh:sqrt:srand:tan:tanh');
        $this->_allowed_operators = explode(',', '-,<,<=,>,>=,==,!=,*,+,/,(,),%,and,or,xor,<<,>>,===,!==,&,^,|,&&,||');
        $this->_parser_check = array();
        // check should be fast, so make a hash
        foreach ($this->_allowed_functions as $f)
            $this->_parser_check[$f] = 1;
        foreach ($this->_allowed_operators as $f)
            $this->_parser_check[$f] = 1;
        if (is_array($placeholders))
            foreach ($placeholders as $f)
                $this->_parser_check[$f] = 1;
        else $this->_parser_check[$placeholders] = 1;

        // This is a speciality: := looks like the attribute definition and is
        // therefore a dummy check for this definition.
        // php-4.2.2 has a problem with /\b:=\b/ matching "population := 1223400"
        $this->_query = preg_replace("/:=/", "==", $this->_query);
        $this->_query = $this->check_query($this->_query);
    }

    function getType()
    {
        return 'numeric';
    }

    /**
     * Check the symbolic definition query against unwanted functions and characters.
     * "population < 20000 and area > 1000000" vs
     *   "area > 1000000 and mail($me,file("/etc/passwd"),...)"
     * http://localhost/wikicvs/SemanticSearch?attribute=*&attr_op=<0 and find(1)>&s=-0.01&start_debug=1
     */
    function check_query($query)
    {
        $tmp = $query; // check for all function calls, in case the tokenizer is not available.
        while (preg_match("/([a-z][a-z0-9]+)\s*\((.*)$/i", $tmp, $m)) {
            if (!in_array($m[1], $this->_allowed_functions)
                and !in_array($m[1], $this->_allowed_operators)
            ) {
                trigger_error("Illegal function in query: " . $m[1], E_USER_WARNING);
                return '';
            }
            $tmp = $m[2];
        }

        // Strictly check for illegal functions and operators, which are no placeholders.
        if (function_exists('token_get_all')) {
            $parsed = token_get_all("<?$query?>");
            foreach ($parsed as $x) { // flat, non-recursive array
                if (is_string($x) and !isset($this->_parser_check[$x])) {
                    // single char op or name
                    trigger_error("Illegal string or operator in query: \"$x\"", E_USER_WARNING);
                    $query = '';
                } elseif (is_array($x)) {
                    $n = token_name($x[0]);
                    if ($n == 'T_OPEN_TAG' or $n == 'T_WHITESPACE'
                        or $n == 'T_CLOSE_TAG' or $n == 'T_LNUMBER'
                        or $n == 'T_CONST' or $n == 'T_DNUMBER'
                    ) continue;
                    if ($n == 'T_VARIABLE') { // but we do allow consts
                        trigger_error("Illegal variable in query: \"$x[1]\"", E_USER_WARNING);
                        $query = '';
                    }
                    if (is_string($x[1]) and !isset($this->_parser_check[$x[1]])) {
                        // multi-char char op or name
                        trigger_error("Illegal $n in query: \"$x[1]\"", E_USER_WARNING);
                        $query = '';
                    }
                }
            }
            //echo "$query <br>";
            //$this->_parse_token($parsed);
            //echo "<br>\n";
            //var_dump($parsed);
            /*
    "_x > 0" =>
    { T_OPEN_TAG "<?"} { T_STRING "_x"} { T_WHITESPACE " "} ">" { T_WHITESPACE " "} { T_LNUMBER "0"} { T_CLOSE_TAG "?>"}
        Interesting: on-char ops, as ">" are not tokenized.
    "_x <= 0"
    { T_OPEN_TAG "< ?" } { T_STRING "_x" } { T_WHITESPACE " " } { T_IS_SMALLER_OR_EQUAL "<=" } { T_WHITESPACE " " } { T_LNUMBER "0" } { T_CLOSE_TAG "?>" }
             */
        } else {
            // Detect illegal characters besides nums, words and ops.
            // So attribute names can not be utf-8
            $c = "/([^\d\w.,\s" . preg_quote(join("", $this->_allowed_operators), "/") . "])/";
            if (preg_match($c, $query, $m)) {
                trigger_error("Illegal character in query: " . $m[1], E_USER_WARNING);
                return '';
            }
        }
        return $query;
    }

    /**
     * Check the bound, numeric-only query against unwanted functions and sideeffects.
     * "4560000 < 20000 and 1456022 > 1000000"
     */
    private function live_check()
    {
        // TODO: check $this->_workquery again?
        return !empty($this->workquery);
    }

    /**
     * A numeric query can only operate with predefined variables. "x < 0 and y < 1"
     *
     * @return array The names as array of strings. => ('x', 'y') the placeholders.
     */
    function getVars()
    {
        if (is_array($this->_placeholders)) return $this->_placeholders;
        else return array($this->_placeholders);
    }

    /**
     * Strip non-numeric chars from the variable (as the groupseperator) and replace
     * it in the symbolic query for evaluation.
     *
     * @access private
     * @param $value number   A numerical value: integer, float or string.
     * @param $x string       The variable name to be replaced in the query.
     * @return string
     */
    private function bind($value, $x)
    {
        // TODO: check is_number, is_float, is_integer and do casting
        $this->bound[] = array('linkname' => $x,
            'linkvalue' => $value);
        $value = preg_replace("/[^-+0123456789.,]/", "", $value);
        //$c = "/\b".preg_quote($x,"/")."\b/";
        $this->workquery = preg_replace("/\b" . preg_quote($x, "/") . "\b/", $value, $this->workquery);
        // FIXME: do again a final check. now only numbers and some operators are allowed.
        return $this->workquery;
    }

    /* array of successfully bound vars, and in case of success, the resulting vars
     */
    private function bound()
    {
        return $this->bound;
    }

    /**
     * With an array of placeholders we need a hash to check against, if all required names are given.
     * Purpose: Be silent about missing vars, just return false.
    `*
     * @access public
     * @param string $variables string or hash of name => value  The keys must satisfy all placeholders in the definition.
     * We want the full hash and not just the keys because a hash check is faster than the array of keys check.
     * @return boolean
     */
    function can_match(&$variables)
    {
        if (empty($this->_query))
            return false;
        $p =& $this->_placeholders;
        if (!is_array($variables) and !is_array($p))
            return $variables == $p; // This was easy.
        // Check if all placeholders have definitions. can be overdefined but not underdefined.
        if (!is_array($p)) {
            if (!isset($variables[$p])) return false;
        } else {
            foreach ($p as $x) {
                if (!isset($variables[$x])) return false;
            }
        }
        return true;
    }

    /**
     * We can match against a single variable or against a hash of variables.
     * With one placeholder we need just a number.
     * With an array of placeholders we need a hash.
     *
     * @access public
     * @param $variable number or array of name => value  The keys must satisfy all placeholders in the definition.
     * @return boolean
     */
    function match(&$variable)
    {
        $p =& $this->_placeholders;
        $this->workquery = $this->_query;
        if (!is_array($p)) {
            if (is_array($variable)) { // which var to match? we cannot decide this here
                if (!isset($variable[$p]))
                    trigger_error("Required NumericSearchQuery->match variable $p not defined.", E_USER_ERROR);
                $this->bind($variable[$p], $p);
            } else {
                $this->bind($variable, $p);
            }
        } else {
            foreach ($p as $x) {
                if (!isset($variable[$x]))
                    trigger_error("Required NumericSearchQuery->match variable $x not defined.", E_USER_ERROR);
                $this->bind($variable[$x], $x);
            }
        }
        if (!$this->live_check()) // check returned an error
            return false;
        $search = $this->workquery;
        $result = false;
        //if (DEBUG & _DEBUG_VERBOSE)
        //    trigger_error("\$result = (boolean)($search);", E_USER_NOTICE);
        // We might have a numerical problem:
        // php-4.2.2 eval'ed as module: "9.636e+08 > 1000" false;
        // php-5.1.2 cgi true, 4.2.2 cgi true
        eval("\$result = (boolean)($search);");
        if ($result and is_array($p)) {
            return $this->bound();
        }
        return $result;
    }
}

////////////////////////////////////////////////////////////////
//
// Remaining classes are private.
//
////////////////////////////////////////////////////////////////
/**
 * Virtual base class for nodes in a TextSearchQuery parse tree.
 *
 * Also serves as a 'VOID' (contentless) node.
 */
class TextSearchQuery_node
{
    public $op = 'VOID';
    public $_op = 0;
    public $word;

    /**
     * Optimize this node.
     * @return object Optimized node.
     */
    function optimize()
    {
        return $this;
    }

    /**
     * @return string regexp matching this node.
     */
    function regexp()
    {
        return '';
    }

    /**
     * @param bool $negated True if this node has been negated (higher in the parse tree.)
     * @return array A list of all non-negated words contained by this node.
     */
    function highlight_words($negated = false)
    {
        return array();
    }

    function sql()
    {
        return $this->word;
    }

    protected function sql_quote()
    {
        global $request;
        $word = preg_replace('/(?=[%_\\\\])/', "\\", $this->word);
        return $request->_dbi->_backend->qstr($word);
    }
}

/**
 * A word. Exact or substring?
 */
class TextSearchQuery_node_word
    extends TextSearchQuery_node
{
    public $op = "WORD";
    public $_op = TSQ_TOK_WORD;

    function TextSearchQuery_node_word($word)
    {
        $this->word = $word;
    }

    function regexp()
    {
        return '(?=.*\b' . preg_quote($this->word, '/') . '\b)';
    }

    function highlight_words($negated = false)
    {
        return $negated ? array() : array($this->word);
    }

    function sql()
    {
        return '%' . $this->sql_quote($this->word) . '%';
    }
}

class TextSearchQuery_node_all
    extends TextSearchQuery_node
{
    public $op = "ALL";
    public $_op = TSQ_TOK_ALL;

    function regexp()
    {
        return '(?=.*)';
    }

    function sql()
    {
        return '%';
    }
}

class TextSearchQuery_node_starts_with
    extends TextSearchQuery_node_word
{
    public $op = "STARTS_WITH";
    public $_op = TSQ_TOK_STARTS_WITH;

    function regexp()
    {
        return '(?=.*\b' . preg_quote($this->word, '/') . ')';
    }

    function sql()
    {
        return $this->sql_quote($this->word) . '%';
    }
}

// ^word: full phrase starts with
class TextSearchQuery_phrase_starts_with
    extends TextSearchQuery_node_starts_with
{
    function regexp()
    {
        return '(?=^' . preg_quote($this->word, '/') . ')';
    }
}

class TextSearchQuery_node_ends_with
    extends TextSearchQuery_node_word
{
    public $op = "ENDS_WITH";
    public $_op = TSQ_TOK_ENDS_WITH;

    function regexp()
    {
        return '(?=.*' . preg_quote($this->word, '/') . '\b)';
    }

    function sql()
    {
        return '%' . $this->sql_quote($this->word);
    }
}

// word$: full phrase ends with
class TextSearchQuery_phrase_ends_with
    extends TextSearchQuery_node_ends_with
{
    function regexp()
    {
        return '(?=' . preg_quote($this->word, '/') . '$)';
    }
}

class TextSearchQuery_node_exact
    extends TextSearchQuery_node_word
{
    public $op = "EXACT";
    public $_op = TSQ_TOK_EXACT;

    function regexp()
    {
        return '(?=\b' . preg_quote($this->word, '/') . '\b)';
    }

    function sql()
    {
        return $this->_sql_squote($this->word);
    }
}

class TextSearchQuery_node_regex // posix regex. FIXME!
    extends TextSearchQuery_node_word
{
    public $op = "REGEX"; // using REGEXP or ~ extension
    public $_op = TSQ_TOK_REGEX;

    function regexp()
    {
        return '(?=.*\b' . $this->word . '\b)';
    }

    function sql()
    {
        return $this->sql_quote($this->word);
    }
}

class TextSearchQuery_node_regex_glob
    extends TextSearchQuery_node_regex
{
    public $op = "REGEX_GLOB";
    public $_op = TSQ_TOK_REGEX_GLOB;

    function regexp()
    {
        return '(?=.*\b' . glob_to_pcre($this->word) . '\b)';
    }
}

class TextSearchQuery_node_regex_pcre // how to handle pcre modifiers? /i
    extends TextSearchQuery_node_regex
{
    public $op = "REGEX_PCRE";
    public $_op = TSQ_TOK_REGEX_PCRE;

    function regexp()
    {
        return $this->word;
    }
}

class TextSearchQuery_node_regex_sql
    extends TextSearchQuery_node_regex
{
    public $op = "REGEX_SQL"; // using LIKE
    public $_op = TSQ_TOK_REGEX_SQL;

    function regexp()
    {
        return str_replace(array("/%/", "/_/"), array(".*", "."), $this->word);
    }

    function sql()
    {
        return $this->word;
    }
}

/**
 * A negated clause.
 */
class TextSearchQuery_node_not
    extends TextSearchQuery_node
{
    public $op = "NOT";
    public $_op = TSQ_TOK_NOT;

    function TextSearchQuery_node_not($leaf)
    {
        $this->leaves = array($leaf);
    }

    function optimize()
    {
        $leaf = &$this->leaves[0];
        $leaf = $leaf->optimize();
        if ($leaf->_op == TSQ_TOK_NOT)
            return $leaf->leaves[0]; // ( NOT ( NOT x ) ) -> x
        return $this;
    }

    function regexp()
    {
        $leaf = &$this->leaves[0];
        return '(?!' . $leaf->regexp() . ')';
    }

    function highlight_words($negated = false)
    {
        return $this->leaves[0]->highlight_words(!$negated);
    }
}

/**
 * Virtual base class for 'AND' and 'OR conjoins.
 */
class TextSearchQuery_node_binop
    extends TextSearchQuery_node
{
    public $_op = TSQ_TOK_BINOP;

    function TextSearchQuery_node_binop($leaves)
    {
        $this->leaves = $leaves;
    }

    protected function flatten()
    {
        // This flattens e.g. (AND (AND a b) (OR c d) e)
        //        to (AND a b e (OR c d))
        $flat = array();
        foreach ($this->leaves as $leaf) {
            $leaf = $leaf->optimize();
            if ($this->op == $leaf->op)
                $flat = array_merge($flat, $leaf->leaves);
            else
                $flat[] = $leaf;
        }
        $this->leaves = $flat;
    }

    function optimize()
    {
        $this->flatten();
        assert(!empty($this->leaves));
        if (count($this->leaves) == 1)
            return $this->leaves[0]; // (AND x) -> x
        return $this;
    }

    function highlight_words($negated = false)
    {
        $words = array();
        foreach ($this->leaves as $leaf)
            array_splice($words, 0, 0,
                $leaf->highlight_words($negated));
        return $words;
    }
}

/**
 * A (possibly multi-argument) 'AND' conjoin.
 */
class TextSearchQuery_node_and
    extends TextSearchQuery_node_binop
{
    public $op = "AND";

    function optimize()
    {
        $this->flatten();

        // Convert (AND (NOT a) (NOT b) c d) into (AND (NOT (OR a b)) c d).
        // Since OR's are more efficient for regexp matching:
        //   (?!.*a)(?!.*b)  vs   (?!.*(?:a|b))

        // Suck out the negated leaves.
        $nots = array();
        foreach ($this->leaves as $key => $leaf) {
            if ($leaf->_op == TSQ_TOK_NOT) {
                $nots[] = $leaf->leaves[0];
                unset($this->leaves[$key]);
            }
        }

        // Combine the negated leaves into a single negated or.
        if ($nots) {
            $node = (new TextSearchQuery_node_not
            (new TextSearchQuery_node_or($nots)));
            array_unshift($this->leaves, $node->optimize());
        }

        assert(!empty($this->leaves));
        if (count($this->leaves) == 1)
            return $this->leaves[0]; // (AND x) -> x
        return $this;
    }

    /* FIXME!
     * Either we need all combinations of all words to be position independent,
     * or we have to use multiple match calls for each AND
     * (AND x y) => /(?(:x)(:y))|(?(:y)(:x))/
     */
    function regexp()
    {
        $regexp = '';
        foreach ($this->leaves as $leaf)
            $regexp .= $leaf->regexp();
        return $regexp;
    }
}

/**
 * A (possibly multi-argument) 'OR' conjoin.
 */
class TextSearchQuery_node_or
    extends TextSearchQuery_node_binop
{
    public $op = "OR";

    function regexp()
    {
        // We will combine any of our direct descendents which are WORDs
        // into a single (?=.*(?:word1|word2|...)) regexp.

        $regexps = array();
        $words = array();

        foreach ($this->leaves as $leaf) {
            if ($leaf->op == TSQ_TOK_WORD)
                $words[] = preg_quote($leaf->word, '/');
            else
                $regexps[] = $leaf->regexp();
        }

        if ($words)
            array_unshift($regexps,
                '(?=.*' . $this->join($words) . ')');

        return $this->join($regexps);
    }

    private function join($regexps)
    {
        assert(count($regexps) > 0);

        if (count($regexps) > 1)
            return '(?:' . join('|', $regexps) . ')';
        else
            return $regexps[0];
    }
}

////////////////////////////////////////////////////////////////
//
// Parser:
//   op's (and, or, not) are forced to lowercase in the tokenizer.
//
////////////////////////////////////////////////////////////////
class TextSearchQuery_Parser
{
    /*
     * This is a simple recursive descent parser, based on the following grammar:
     *
     * toplist    :
     *        | toplist expr
     *        ;
     *
     *
     * list    : expr
     *        | list expr
     *        ;
     *
     * expr    : atom
     *        | expr BINOP atom
     *        ;
     *
     * atom    : '(' list ')'
     *        | NOT atom
     *        | WORD
     *        ;
     *
     * The terminal tokens are:
     *
     *
     * and|or          BINOP
     * -|not          NOT
     * (          LPAREN
     * )          RPAREN
     * /[^-()\s][^()\s]*  WORD
     * /"[^"]*"/      WORD
     * /'[^']*'/      WORD
     *
     * ^WORD              TextSearchQuery_phrase_starts_with
     * WORD*              STARTS_WITH
     * *WORD              ENDS_WITH
     * ^WORD$             EXACT
     * *                  ALL
     */

    public $lexer;
    private $regex;

    function parse($search_expr, $case_exact = false, $regex = TSQ_REGEX_AUTO)
    {
        $this->lexer = new TextSearchQuery_Lexer($search_expr, $case_exact, $regex);
        $this->regex = $regex;
        $tree = $this->get_list('toplevel');
        // Assert failure when using the following URL in debug mode.
        // /TitleSearch?action=FullTextSearch&s=WFXSSProbe'")/>&case_exact=1&regex=sql
        //        assert($this->lexer->eof());
        unset($this->lexer);
        return $tree;
    }

    function get_list($is_toplevel = false)
    {
        $list = array();

        // token types we'll accept as words (and thus expr's) for the
        // purpose of error recovery:
        $accept_as_words = TSQ_TOK_NOT | TSQ_TOK_BINOP;
        if ($is_toplevel)
            $accept_as_words |= TSQ_TOK_LPAREN | TSQ_TOK_RPAREN;

        while (($expr = $this->get_expr())
            || ($expr = $this->get_word($accept_as_words))) {
            $list[] = $expr;
        }

        if (!$list) {
            if ($is_toplevel)
                return new TextSearchQuery_node;
            else
                return false;
        }
        if ($is_toplevel and count($list) == 1) {
            if ($this->lexer->query_str[0] == '^')
                return new TextSearchQuery_phrase_starts_with($list[0]->word);
            else
                return $list[0];
        }
        return new TextSearchQuery_node_and($list);
    }

    function get_expr()
    {
        if (($expr = $this->get_atom()) === false) // protect against '0'
            return false;

        $savedpos = $this->lexer->tell();
        // Bug#1791564: allow string '0'
        while (($op = $this->lexer->get(TSQ_TOK_BINOP)) !== false) {
            if (!($right = $this->get_atom())) {
                break;
            }

            if ($op == 'and')
                $expr = new TextSearchQuery_node_and(array($expr, $right));
            else {
                assert($op == 'or');
                $expr = new TextSearchQuery_node_or(array($expr, $right));
            }

            $savedpos = $this->lexer->tell();
        }
        $this->lexer->seek($savedpos);

        return $expr;
    }

    function get_atom()
    {
        if ($atom = $this->get_word(TSQ_ALLWORDS)) // Bug#1791564 not involved: '*'
            return $atom;

        $savedpos = $this->lexer->tell();
        if ($this->lexer->get(TSQ_TOK_LPAREN)) {
            if (($list = $this->get_list()) && $this->lexer->get(TSQ_TOK_RPAREN)) {
                return $list;
            } else {
                // Fix Bug#1792170
                // Handle " ( " or "(test" without closing ")" as plain word
                $this->lexer->seek($savedpos);
                return new TextSearchQuery_node_word($this->lexer->get(-1));
            }
        } elseif ($this->lexer->get(TSQ_TOK_NOT)) {
            if (($atom = $this->get_atom()))
                return new TextSearchQuery_node_not($atom);
        }
        $this->lexer->seek($savedpos);
        return false;
    }

    function get_word($accept = TSQ_ALLWORDS)
    {
        // Performance shortcut for ( and ). This is always false
        if (!empty($this->lexer->tokens[$this->lexer->pos])) {
            list ($type, $val) = $this->lexer->tokens[$this->lexer->pos];
            if ($type == TSQ_TOK_LPAREN or $type == TSQ_TOK_RPAREN)
                return false;
        }
        foreach (array("WORD", "STARTS_WITH", "ENDS_WITH", "EXACT",
                     "REGEX", "REGEX_GLOB", "REGEX_PCRE", "ALL") as $tok) {
            $const = constant("TSQ_TOK_" . $tok);
            // Bug#1791564: allow word '0'
            if ($accept & $const and
                (($word = $this->lexer->get($const)) !== false)
            ) {
                // phrase or word level?
                if ($tok == 'STARTS_WITH' and $this->lexer->query_str[0] == '^')
                    $classname = "TextSearchQuery_phrase_" . strtolower($tok);
                elseif ($tok == 'ENDS_WITH' and
                    string_ends_with($this->lexer->query_str, '$')
                )
                    $classname = "TextSearchQuery_phrase_" . strtolower($tok); else
                    $classname = "TextSearchQuery_node_" . strtolower($tok);
                return new $classname($word);
            }
        }
        return false;
    }
}

class TextSearchQuery_Lexer
{
    function TextSearchQuery_Lexer($query_str, $case_exact = false,
                                   $regex = TSQ_REGEX_AUTO)
    {
        $this->tokens = $this->tokenize($query_str, $case_exact, $regex);
        $this->query_str = $query_str;
        $this->pos = 0;
    }

    function tell()
    {
        return $this->pos;
    }

    function seek($pos)
    {
        $this->pos = $pos;
    }

    function eof()
    {
        return $this->pos == count($this->tokens);
    }

    /**
     * TODO: support more regex styles, esp. prefer the forced ones over auto
     * re: and // stuff
     */
    function tokenize($string, $case_exact = false, $regex = TSQ_REGEX_AUTO)
    {
        $tokens = array();
        $buf = $case_exact ? ltrim($string) : strtolower(ltrim($string));
        while (!empty($buf)) {
            if (preg_match('/^([()])\s*/', $buf, $m)) {
                $val = $m[1];
                $type = $m[1] == '(' ? TSQ_TOK_LPAREN : TSQ_TOK_RPAREN;
            } // * => ALL
            elseif ($regex & (TSQ_REGEX_AUTO | TSQ_REGEX_POSIX | TSQ_REGEX_GLOB)
                and preg_match('/^\*\s*/', $buf, $m)
            ) {
                $val = "*";
                $type = TSQ_TOK_ALL;
            } // .* => ALL
            elseif ($regex & (TSQ_REGEX_PCRE)
                and preg_match('/^\.\*\s*/', $buf, $m)
            ) {
                $val = ".*";
                $type = TSQ_TOK_ALL;
            } // % => ALL
            elseif ($regex & (TSQ_REGEX_SQL)
                and preg_match('/^%\s*/', $buf, $m)
            ) {
                $val = "%";
                $type = TSQ_TOK_ALL;
            } // ^word
            elseif ($regex & (TSQ_REGEX_AUTO | TSQ_REGEX_POSIX | TSQ_REGEX_PCRE)
                and preg_match('/^\^([^-()][^()\s]*)\s*/', $buf, $m)
            ) {
                $val = $m[1];
                $type = TSQ_TOK_STARTS_WITH;
            } // word*
            elseif ($regex & (TSQ_REGEX_AUTO | TSQ_REGEX_POSIX | TSQ_REGEX_GLOB)
                and preg_match('/^([^-()][^()\s]*)\*\s*/', $buf, $m)
            ) {
                $val = $m[1];
                $type = TSQ_TOK_STARTS_WITH;
            } // *word
            elseif ($regex & (TSQ_REGEX_AUTO | TSQ_REGEX_POSIX | TSQ_REGEX_GLOB)
                and preg_match('/^\*([^-()][^()\s]*)\s*/', $buf, $m)
            ) {
                $val = $m[1];
                $type = TSQ_TOK_ENDS_WITH;
            } // word$
            elseif ($regex & (TSQ_REGEX_AUTO | TSQ_REGEX_POSIX | TSQ_REGEX_PCRE)
                and preg_match('/^([^-()][^()\s]*)\$\s*/', $buf, $m)
            ) {
                $val = $m[1];
                $type = TSQ_TOK_ENDS_WITH;
            } // ^word$
            elseif ($regex & (TSQ_REGEX_AUTO | TSQ_REGEX_POSIX | TSQ_REGEX_PCRE)
                and preg_match('/^\^([^-()][^()\s]*)\$\s*/', $buf, $m)
            ) {
                $val = $m[1];
                $type = TSQ_TOK_EXACT;
            } elseif (preg_match('/^(and|or)\b\s*/i', $buf, $m)) {
                $val = strtolower($m[1]);
                $type = TSQ_TOK_BINOP;
            } elseif (preg_match('/^(-|not\b)\s*/i', $buf, $m)) {
                $val = strtolower($m[1]);
                $type = TSQ_TOK_NOT;
            } // "words "
            elseif (preg_match('/^ " ( (?: [^"]+ | "" )* ) " \s*/x', $buf, $m)) {
                $val = str_replace('""', '"', $m[1]);
                $type = TSQ_TOK_WORD;
            } // 'words '
            elseif (preg_match("/^ ' ( (?:[^']+|'')* ) ' \s*/x", $buf, $m)) {
                $val = str_replace("''", "'", $m[1]);
                $type = TSQ_TOK_WORD;
            } // word
            elseif (preg_match('/^([^-()][^()\s]*)\s*/', $buf, $m)) {
                $val = $m[1];
                $type = TSQ_TOK_WORD;
            } else {
                assert(empty($buf));
                break;
            }
            $buf = substr($buf, strlen($m[0]));

            /* refine the simple parsing from above: bla*bla, bla?bla, ...
            if ($regex and $type == TSQ_TOK_WORD) {
                if (substr($val,0,1) == "^")
                    $type = TSQ_TOK_STARTS_WITH;
                elseif (substr($val,0,1) == "*")
                    $type = TSQ_TOK_ENDS_WITH;
                elseif (substr($val,-1,1) == "*")
                    $type = TSQ_TOK_STARTS_WITH;
            }
            */
            $tokens[] = array($type, $val);
        }
        return $tokens;
    }

    function get($accept)
    {
        if ($this->pos >= count($this->tokens))
            return false;

        list ($type, $val) = $this->tokens[$this->pos];
        if (($type & $accept) == 0)
            return false;

        $this->pos++;
        return $val;
    }
}

// Local Variables:
// mode: php
// tab-width: 8
// c-basic-offset: 4
// c-hanging-comment-ender-p: nil
// indent-tabs-mode: nil
// End: