spider/lib/Doc.php

<?php
namespace Spider\Lib;
use Dever;

class Doc
{
	private $url;
	private $host;
	private $path;
	private $rule;
	public $type = 'dom';
	private $log = false;
	private $cur = '';
	static protected $instance;

	static public function getInstance($url, $rule = '')
	{
		$key = $url . md5($rule);
		if (empty(self::$instance[$key])) {
			# 此处开协程：Dever::coroutine(self::$instance[$key]);
			self::$instance[$key] = new self($url, $rule);
		}

		return self::$instance[$key];
	}

	public function __construct($url, $rule = '')
	{
		$this->url($url);
		$this->rule = $rule;
		if (strpos($this->rule, '$json') !== false) {
			$this->type = 'json';
		} else {
			$this->type = 'dom';
		}
	}

	private function url($url = false)
	{
		if (!$url) {
			return;
		}
		$this->url = $url;
		$value = parse_url($this->url);
		$this->path = $this->host = $value['scheme'] . '://' . $value['host'];
		if (isset($value['path']) && $value['path']) {
			$temp = explode('/', $value['path']);
			unset($temp[count($temp)-1]);
			$this->path .= implode('/', $temp);
		}
		$this->path .= '/';
	}

	public function get($param = array())
	{
		$doc = $this->doc(false, $param);
		if (!$this->cur) {
			$this->cur = $doc;
		}
		if ($this->rule) {
			$doc = $this->find($doc, $this->rule);
		}
		return $doc;
	}

	public function doc($url = false, $param = array())
	{
		$this->url($url);
		$html = $this->download($this->url, $param);
		return ($this->getClass())::init($html);
	}

	public function getCur()
	{
		return $this->cur;
	}

	private function download($url, $header = '', $param = '')
	{
		$this->addLog($url . '下载中...');
		$download = new Download($url, $header, $param);
		$this->addLog($url . '下载完成');
		return $download->get($this->type);
	}

	private function collect($data, $include, $exclude, $filter)
	{
		if ($include) {
			$include = explode("\n", str_replace("\r", '', $include));
			foreach ($include as $k => $v) {
				$state = preg_match('/' . $v . '/i', $data);
				if ($state) {
					break;
				}
			}

			if (!$state) {
				return 'error';
			}
		}
		if ($exclude) {
			$exclude = explode("\n", str_replace("\r", '', $exclude));
			foreach ($exclude as $k => $v) {
				$state = preg_match('/' . $v . '/i', $data);
				if (!$state) {
					return 'error';
				}
			}
		}
		if ($filter) {
			$filter = explode("\n", str_replace("\r", '', $filter));
			foreach ($filter as $k => $v) {
				$s = '';
				if (strstr($v, '=>')) {
					$temp = explode('=>', $v);
					$v = $temp[0];
					$s = $temp[1];
				}
				$data = preg_replace('/' . $v . '/i', $s, $data);
			}
		}
		
		return $data;
	}

	private function getClass()
	{
		return 'Spider\\Lib\\Doc\\' . ucfirst($this->type);
	}

	public function find($doc, $rule)
	{
		return ($this->getClass())::find($doc, $rule);
	}

	public function init($data)
	{
		if (is_string($data) && filter_var($data, FILTER_VALIDATE_URL) !== false) {
			$data = $this->doc($data);
		} else {
			$data = ($this->getClass())::init($data);
		}
		return $data;
	}

	public function rule($data, $col, $config)
	{
		$name = '字段[' . $config['name'] . '('.$config['key'].')]' . '"';
		$this->addLog($name . '正在按照规则['.$config['collect_rule'].']进行解析');

		$method = 'rule_' . $this->type;

		$result = $this->getRule($data, $col, $config['collect_rule'], $config['key']);

		if (isset($config['collect_url']) && $config['collect_url']) {
			$collect_url = explode("\n", str_replace("\r", '', $config['collect_url']));
			if (!isset($collect_url[1])) {
				$collect_url[1] = '';
			}
			$temp = array();
			$temp[] = $result;
			$this->getNext($temp, $data, $col, $collect_url[0], $collect_url[1], $config['collect_rule'], $config['key']);
			$result = implode(',', $temp);
		}
		
		$this->addLog($name . '解析完成');

		return $this->collect($result, $config['collect_include'], $config['collect_exclude'], $config['collect_filter']);
	}

	public function getNext(&$result, $data, $col, $collect_url, $collect_include, $collect_rule, $key)
	{
		$url = $this->getUrl($data, $col, $collect_url, $key);
		if ($url) {
			if ($collect_include && !strstr($url, $collect_include)) {
				return;
			}
			$data = $this->init($url);
			if ($data) {
				$temp = $this->getRule($data, $col, $collect_rule, $key);
				if ($temp) {
					$result[] = $temp;
					$this->getNext($result, $data, $col, $collect_url, $collect_include, $collect_rule, $key);
				}
			}
		}
	}

	public function getUrl($data, $col, $collect_rule, $key)
	{
		$url = $this->getRule($data, $col, $collect_rule, $key);
		if (!$url) {
			return '';
		}
		if (strpos($url, 'http') === false) {
			if ($url[0] == '/') {
				$url = $this->host . $url;
			} elseif (strstr($url, '.')) {
				$url = $this->path . $url;
			} else {
				$url = $this->url . $url;
			}
		}
		return $url;
	}

	public function getRule($data, $col, $collect_rule, $key)
	{
		return ($this->getClass())::rule($this, $data, $col, $collect_rule, $key);
	}

	public function addLog($string)
	{
		if ($this->log) {
			$this->log->add($string);
		}
	}

	public function saveLog()
	{
		if ($this->log) {
			$this->log->save();
		}
	}

	public function outLog()
	{
		if ($this->log) {
			$this->log->out();
		}
	}

	public function log(Log $log)
	{
		$this->log = $log;
	}
}