public static function crawl($_pid = NULL, $_rid = NULL, $_url = NULL, $_title = NULL)
{
ini_get('safe_mode') or set_time_limit(0);
$sid = spider::$sid;
if ($sid) {
$sRs = iDB::row("SELECT * FROM `#iCMS@__spider_url` WHERE `id`='{$sid}' LIMIT 1;");
$title = $sRs->title;
$cid = $sRs->cid;
$pid = $sRs->pid;
$url = $sRs->url;
$rid = $sRs->rid;
} else {
$rid = spider::$rid;
$pid = spider::$pid;
$title = spider::$title;
$url = spider::$url;
$_rid === NULL or $rid = $_rid;
$_pid === NULL or $pid = $_pid;
$_title === NULL or $title = $_title;
$_url === NULL or $url = $_url;
}
if ($pid) {
$project = spider::project($pid);
$prule_list_url = $project['list_url'];
}
$ruleA = spider::rule($rid);
$rule = $ruleA['rule'];
$dataArray = $rule['data'];
if ($prule_list_url) {
$rule['list_url'] = $prule_list_url;
}
if (spider::$dataTest) {
echo "<b>抓取规则信息</b><pre>";
print_r(iS::escapeStr($ruleA));
print_r(iS::escapeStr($project));
echo "</pre><hr />";
}
spider::$curl_proxy = $rule['proxy'];
$responses = array();
$html = spiderTools::remote($url);
if (empty($html)) {
$msg = '错误:001..采集 ' . $url . '文件内容为空!请检查采集规则';
if (spider::$work == 'shell') {
echo "{$msg}\n";
return false;
} else {
iPHP::alert($msg);
}
}
// $http = spider::check_content_code($html);
//
// if($http['match']==false){
// return false;
// }
// $content = $http['content'];
spider::$allHtml = "";
$rule['__url__'] = spider::$url;
$responses['reurl'] = spider::$url;
$responses['__title__'] = $title;
foreach ((array) $dataArray as $key => $data) {
$content_html = $html;
$dname = $data['name'];
/**
* [UNSET:name]
* 注销[name]
* @var string
*/
if (strpos($dname, 'UNSET:') !== false) {
$_dname = str_replace('UNSET:', '', $dname);
unset($responses[$_dname]);
continue;
}
/**
* [DATA:name]
* 把之前[name]处理完的数据当作原始数据
* 如果之前有数据会叠加
* 用于数据多次处理
* @var string
*/
if (strpos($dname, 'DATA:') !== false) {
$_dname = str_replace('DATA:', '', $dname);
$content_html = $responses[$_dname];
unset($responses[$dname]);
}
/**
* [PRE:name]
* 把PRE:name采集到的数据 当做原始数据
* 一般用于下载内容
* @var string
*/
$pre_dname = 'PRE:' . $dname;
if (isset($responses[$pre_dname])) {
$content_html = $responses[$pre_dname];
unset($responses[$pre_dname]);
}
/**
* [EMPTY:name]
* 如果[name]之前抓取结果数据为空使用这个数据项替换
* @var string
*/
if (strpos($dname, 'EMPTY:') !== false) {
$_dname = str_replace('EMPTY:', '', $dname);
if (empty($responses[$_dname])) {
$dname = $_dname;
} else {
//有值不执行抓取
continue;
}
}
$content = spiderContent::crawl($content_html, $data, $rule, $responses);
unset($content_html);
if (strpos($dname, 'ARRAY:') !== false) {
// if(strpos($data['rule'], 'RULE@')!==false){
$dname = str_replace('ARRAY:', '', $dname);
// $contentArray = $responses[$dname];
// // $contentArray = $responses[$dname];
$cArray = array();
foreach ((array) $content as $k => $value) {
foreach ((array) $value as $key => $val) {
$cArray[$key][$k] = $val;
}
}
if ($cArray) {
$content = $cArray;
unset($cArray);
}
}
/**
* [name.xxx]
* 采集内容做为数组
*/
if (strpos($dname, '.') !== false) {
$f_key = substr($dname, 0, stripos($dname, "."));
$s_key = substr(strrchr($dname, "."), 1);
if (isset($responses[$f_key][$s_key])) {
if (is_array($responses[$f_key][$s_key])) {
$responses[$f_key][$s_key] = array_merge($responses[$f_key][$s_key], $content);
} else {
$responses[$f_key][$s_key] .= $content;
}
} else {
$responses[$f_key][$s_key] = $content;
}
} else {
/**
* 多个name 内容合并
*/
if (isset($responses[$dname])) {
if (is_array($responses[$dname])) {
$responses[$dname] = array_merge($responses[$dname], $content);
} else {
$responses[$dname] .= $content;
}
} else {
$responses[$dname] = $content;
}
}
/**
* 对匹配多条的数据去重过滤
*/
if (!is_array($responses[$dname]) && $data['multi']) {
if (strpos($responses[$dname], ',') !== false) {
$_dnameArray = explode(',', $responses[$dname]);
$dnameArray = array();
foreach ((array) $_dnameArray as $key => $value) {
$value = trim($value);
$value && ($dnameArray[] = $value);
}
$dnameArray = array_filter($dnameArray);
$dnameArray = array_unique($dnameArray);
$responses[$dname] = implode(',', $dnameArray);
unset($dnameArray, $_dnameArray);
}
}
gc_collect_cycles();
}
if (isset($responses['title']) && empty($responses['title'])) {
$responses['title'] = $responses['__title__'];
}
spider::$allHtml = null;
unset($html);
gc_collect_cycles();
if (spider::$dataTest) {
echo "<pre style='width:99%;word-wrap: break-word;'>";
print_r(iS::escapeStr($responses));
echo '<hr />';
echo '使用内存:' . iFS::sizeUnit(memory_get_usage()) . ' 执行时间:' . iPHP::timer_stop() . 's';
echo "</pre>";
}
iFS::$CURLOPT_ENCODING = '';
iFS::$CURLOPT_REFERER = '';
iFS::$watermark_config['pos'] = iCMS::$config['watermark']['pos'];
iFS::$watermark_config['x'] = iCMS::$config['watermark']['x'];
iFS::$watermark_config['y'] = iCMS::$config['watermark']['y'];
iFS::$watermark_config['img'] = iCMS::$config['watermark']['img'];
$rule['fs']['encoding'] && (iFS::$CURLOPT_ENCODING = $rule['fs']['encoding']);
$rule['fs']['referer'] && (iFS::$CURLOPT_REFERER = $rule['fs']['referer']);
if ($rule['watermark_mode']) {
iFS::$watermark_config['pos'] = $rule['watermark']['pos'];
iFS::$watermark_config['x'] = $rule['watermark']['x'];
iFS::$watermark_config['y'] = $rule['watermark']['y'];
$rule['watermark']['img'] && (iFS::$watermark_config['img'] = $rule['watermark']['img']);
}
if (spider::$callback['data'] && is_callable(spider::$callback['data'])) {
$responses = call_user_func_array(spider::$callback['data'], array($responses));
}
return $responses;
}
