. */ namespace Doctrine\Common; /** * Simple generic lexical scanner * * @license http://www.opensource.org/licenses/lgpl-license.php LGPL * @link www.doctrine-project.org * @since 2.0 * @version $Revision: 3938 $ * @author Guilherme Blanco * @author Jonathan Wage * @author Roman Borschel */ abstract class Lexer { /** * Array of scanned tokens. * * @var array */ private $_tokens = array(); /** * @todo Doc */ private $_position = 0; /** * @todo Doc */ private $_peek = 0; /** * @var array The next token in the query string. */ public $lookahead; /** * @var array The last matched/seen token. */ public $token; /** * Inputs data to be tokenized * * @param string $input input to be tokenized */ public function setInput($input) { $this->_tokens = array(); $this->reset(); $this->_scan($input); } /** * Resets the scanner * */ public function reset() { $this->lookahead = null; $this->token = null; $this->_peek = 0; $this->_position = 0; } /** * Resets the peek pointer to 0 * */ public function resetPeek() { $this->_peek = 0; } /** * Resets the lexer position on the input to the given position * * @param integer $position Position to place the lexical scanner */ public function resetPosition($position = 0) { $this->_position = $position; } /** * Checks whether a given token matches the current lookahead. * * @param integer|string $token * @return boolean */ public function isNextToken($token) { $la = $this->lookahead; return ($la['type'] === $token || $la['value'] === $token); } /** * Moves to the next token in the input string. * * A token is an associative array containing three items: * - 'value' : the string value of the token in the input string * - 'type' : the type of the token (identifier, numeric, string, input * parameter, none) * - 'position' : the position of the token in the input string * * @return array|null the next token; null if there is no more tokens left */ public function moveNext() { $this->token = $this->lookahead; $this->_peek = 0; if (isset($this->_tokens[$this->_position])) { $this->lookahead = $this->_tokens[$this->_position++]; return true; } else { $this->lookahead = null; return false; } } /** * Tells the lexer to skip input tokens until it sees a token with the given value. * * @param $value The value to skip until. */ public function skipUntil($value) { while ($this->lookahead !== null && $this->lookahead['value'] !== $value) { $this->moveNext(); } } /** * @todo Doc */ public function isA($value, $token) { $type = $this->_getType($value); return $type === $token; } /** * Moves the lookahead token forward. * * @return array | null The next token or NULL if there are no more tokens ahead. */ public function peek() { if (isset($this->_tokens[$this->_position + $this->_peek])) { return $this->_tokens[$this->_position + $this->_peek++]; } else { return null; } } /** * Peeks at the next token, returns it and immediately resets the peek. * * @return array|null The next token or NULL if there are no more tokens ahead. */ public function glimpse() { $peek = $this->peek(); $this->_peek = 0; return $peek; } /** * Scans the input string for tokens. * * @param string $input a query string */ protected function _scan($input) { static $regex; if ( ! isset($regex)) { $regex = '/(' . implode(')|(', $this->getCatchablePatterns()) . ')|' . implode('|', $this->getNonCatchablePatterns()) . '/i'; } $flags = PREG_SPLIT_NO_EMPTY | PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_OFFSET_CAPTURE; $matches = preg_split($regex, $input, -1, $flags); foreach ($matches as $match) { $value = $match[0]; $type = $this->_getType($value); $this->_tokens[] = array( 'value' => $value, 'type' => $type, 'position' => $match[1] ); } } /** * Lexical catchable patterns * * @return array */ abstract protected function getCatchablePatterns(); /** * Lexical non-catchable patterns * * @return array */ abstract protected function getNonCatchablePatterns(); /** * Retrieve token type. Also processes the token value if necessary. * * @param string $value * @return integer */ abstract protected function _getType(&$value); }