/**
* 采集
* @param int $id collect_task_id
* @url string 需要读取的文件地址
* @task array 采集任务记录数组
*/
public function ContentReg($id, $url, $task)
{
$re = CollectTask::read_html($url);
if ($re['success']) {
$content = $re['content'];
} else {
echo '<script>alert("' . $re['error'] . '");window.close();</script>';
exit;
}
$url_arr = parse_url($url);
if ($url_arr['path']) {
$a = array_filter(explode("/", $url_arr['path']));
array_pop($a);
$pa = implode("/", $a) . "/";
} else {
$pa = "";
}
$content = CollectTask::re2ab($content, $url_arr['scheme'] . "://" . $url_arr['host'] . "/" . $pa);
$template = CollectTemplate::get_template_by_id($task['collect_template_id'], false);
if ($task['collect_task_totalpagereg']) {
$totalpagereg = $task['collect_task_totalpagereg'];
} elseif ($template['collect_template_totalpagereg']) {
$totalpagereg = $template['collect_template_totalpagereg'];
} else {
$totalpagereg = "";
}
$pagerule = $task['collect_task_pagerule'] ? $task['collect_task_pagerule'] : $template['collect_template_pagerule'];
$totalpage = 0;
if ($totalpagereg) {
$totalpagereg = preg_replace("/\\[.*\\]/U", "(\\d)", $totalpagereg);
preg_match("/" . $totalpagereg . "/", $content, $p);
if (intval($p[1])) {
$totalpage = intval($p[1]);
}
}
$pageurl_arr = array();
if ($totalpage) {
for ($i = 2; $i <= $totalpage; $i++) {
$pageurl_arr[] = preg_replace("/{$pagerule}\\d/", $pagerule . $i, $url);
}
} else {
//不是js分页时，则查找分页地址
$pagestart = $task['collect_task_pagestart'] ? $task['collect_task_pagestart'] : $template['collect_template_pagestart'];
$pageend = $task['collect_task_pageend'] ? $task['collect_task_pageend'] : $template['collect_template_pageend'];
if ($pagestart && $pageend) {
$page_area_reg = str_replace("/", "\\/", $pagestart) . "([\\s\\S]*)" . str_replace("/", "\\/", $pageend);
preg_match("/{$page_area_reg}/Ui", $content, $mm);
if ($mm[1]) {
preg_match_all('/\\<a.*href=["|\'](.*)?["|\'].*\\>(.*)\\<\\/a\\>/iU', $content, $arr);
if (count($arr[1])) {
$links = array_unique($arr[1]);
sort($links);
$source_url = preg_replace("/{$pagerule}\\d/", "", $url);
if ($pagerule) {
foreach ($links as $k => $v) {
if ($v != $url) {
$baseurl = preg_replace("/{$pagerule}\\d/", "", $v);
if ($baseurl == $source_url && $v != $url) {
$pageurl_arr[] = $v;
}
}
}
}
}
}
}
}
if (is_array($pageurl_arr) && count($pageurl_arr)) {
//取得所有分页内容
foreach ($pageurl_arr as $v) {
$c = CollectTask::read_html($v);
if ($c['content']) {
$c['content'] = CollectTask::re2ab($c['content'], $url_arr['scheme'] . "://" . $url_arr['host'] . "/");
$content .= $c['content'];
}
}
}
$content = str_replace("none", "block", $content);
//有些分页是默认不显示内容的
$filter = array();
if ($task['collect_task_filter']) {
$filter = array_merge($filter, json_decode($task['collect_task_filter']));
}
if ($template['collect_template_filter']) {
$filter = array_merge($filter, json_decode($template['collect_template_filter']));
}
$_charset = CollectTask::get_charset();
$content = mb_convert_encoding($content, "UTF-8", "gb2312,gbk,utf-8");
if (count($filter)) {
foreach ($filter as $f) {
$f = preg_replace("/\\[.*\\]/U", "([\\s\\S]*)", $f);
$f = str_replace("/", "\\/", $f);
$content = preg_replace("/" . $f . "/Ui", " ", $content);
}
}
if ($task['collect_task_rulearr']) {
eval('$task_rule_arr = ' . $task['collect_task_rulearr'] . ';');
$task_rule_arr = array_filter($task_rule_arr);
} else {
$task_rule_arr = array();
}
if ($template['collect_template_fieldsreg']) {
eval('$template_rule_arr = ' . $template['collect_template_fieldsreg'] . ';');
}
if (count($task_rule_arr)) {
foreach ($task_rule_arr as $k => $v) {
$template_rule_arr[$k] = $v;
}
}
$template_rule_arr = array_filter($template_rule_arr);
$fields = array();
if (count($template_rule_arr) && is_array($template_rule_arr)) {
foreach ($template_rule_arr as $k => $v) {
$reg = preg_replace("/\\[.*\\]/U", "([\\s\\S]*)", $v);
$reg = str_replace("/", "\\/", $reg);
preg_match_all("/" . $reg . "/U", $content, $ma);
if (count($ma[1])) {
foreach ($ma[1] as $k2 => $v2) {
if (CollectTask::is_image($v2)) {
if ($task['collect_task_saveimg']) {
$ma[1][$k2] = CollectTask::save_image($v2);
} else {
$ma[1][$k2] = $v2;
}
} else {
if ($task['collect_task_saveimg']) {
preg_match_all('/<img.*src=["|\']+(.*)?["|\']+>/Ui', $v2, $arr);
//正则出所有图片保存替换
if (is_array($arr[1]) && count($arr[1])) {
foreach ($arr[1] as $img) {
$ma[1][$k2] = str_replace($img, CollectTask::save_image($img), $v2);
}
}
}
}
}
preg_match("/\\[(.*)\\]/U", $v, $match);
$fields[] = array('identify' => $k, 'name' => $match[1], 'value' => $ma[1][0]);
}
}
}
return $fields;
}
if (is_array($f['value'])) {
foreach ($f['value'] as $v) {
if (CollectTask::is_image($v)) {
//echo CollectTask::signName($v).".".CollectTask::getFileType($v);
?>
<img src='<?php
echo $v;
?>
' /><br />
<?php
} else {
echo $v . '<br />';
}
}
} else {
if (CollectTask::is_image($f['value'])) {
?>
<img src='<?php
echo $v;
?>
' /><br />
<?php
} else {
echo $v . '<br />';
}
}
?>
</td>
</tr>
<?php
}
