/**
* Gets the next doc from the iterator
* @param bool $no_process do not do any processing on page data
* @return array associative array for doc or string if no_process true
*/
function nextPage($no_process = false)
{
if (!$this->checkFileHandle()) {
return NULL;
}
$indexable_records = array('response', 'resource');
do {
$this->getRecordStart();
$page_info = $this->getWarcHeaders();
if ($page_info == NULL || !isset($page_info[self::SIZE])) {
return NULL;
}
$length = intval($page_info[self::SIZE]);
$page_info[self::SIZE] = $length;
$header_and_page = ltrim($this->fileRead($length + 2));
$this->fileGets();
$this->fileGets();
if (!$header_and_page) {
return NULL;
}
} while (!in_array($page_info['warc-type'], $indexable_records) || substr($page_info[self::URL], 0, 4) == 'dns:');
//ignore warcinfo, request, metadata, revisit, etc. records
if ($no_process) {
return $header_and_page;
}
unset($page_info['line']);
unset($page_info['warc-type']);
$site = $page_info;
$site_contents = FetchUrl::parseHeaderPage($header_and_page);
$site = array_merge($site, $site_contents);
$site[self::HASH] = FetchUrl::computePageHash($site[self::PAGE]);
$site[self::WEIGHT] = 1;
if (!isset($site[self::TYPE])) {
$site[self::TYPE] = "text/plain";
}
return $site;
}
/**
* Gets the next doc from the iterator
* @param bool $no_process do not do any processing on page data
* @return array associative array for doc or string if no_process true
*/
function nextPage($no_process = false)
{
if (!$this->checkFileHandle()) {
return NULL;
}
do {
$page_info = $this->fileGets();
if (trim($page_info) == "") {
return NULL;
}
$info_parts = explode(" ", $page_info);
$num_parts = count($info_parts);
$length = intval($info_parts[$num_parts - 1]);
$header_and_page = $this->fileRead($length + 1);
if (!$header_and_page) {
return NULL;
}
} while (substr($page_info, 0, 3) == 'dns' || substr($page_info, 0, 8) == 'filedesc');
//ignore dns entries in arc and ignore first record
if ($no_process) {
return $header_and_page;
}
$site = array();
$site[self::URL] = $info_parts[0];
$site[self::IP_ADDRESSES] = array($info_parts[1]);
$site[self::TIMESTAMP] = date("U", strtotime($info_parts[2]));
$site[self::TYPE] = $info_parts[3];
$site_contents = FetchUrl::parseHeaderPage($header_and_page);
$site = array_merge($site, $site_contents);
$site[self::HASH] = FetchUrl::computePageHash($site[self::PAGE]);
$site[self::WEIGHT] = 1;
return $site;
}
