/**
* Handles admin request related to controlling file options to be used
* in a crawl
*
* This activity allows a user to specify the page range size to be
* be used during a crawl as well as which file types can be downloaded
*/
function pageOptions()
{
global $INDEXED_FILE_TYPES;
/* get processors for different file types (populating
$INDEXED_FILE_TYPES) */
foreach (glob(BASE_DIR . "/lib/processors/*_processor.php") as $filename) {
require_once $filename;
}
$parent = $this->parent;
$crawl_model = $parent->model("crawl");
$profile_model = $parent->model("profile");
$data["ELEMENT"] = "pageoptions";
$data['SCRIPT'] = "";
$machine_urls = $parent->model("machine")->getQueueServerUrls();
$num_machines = count($machine_urls);
if ($num_machines < 1 || $num_machines == 1 && UrlParser::isLocalhostUrl($machine_urls[0])) {
$machine_urls = NULL;
}
$data['available_options'] = array(tl('crawl_component_use_below'), tl('crawl_component_use_defaults'));
$crawls = $crawl_model->getCrawlList(false, true, $machine_urls);
$data['options_default'] = tl('crawl_component_use_below');
foreach ($crawls as $crawl) {
if (strlen($crawl['DESCRIPTION']) > 0) {
$data['available_options'][$crawl['CRAWL_TIME']] = $crawl['DESCRIPTION'];
}
}
$seed_info = $crawl_model->getSeedInfo();
$data['RECRAWL_FREQS'] = array(-1 => tl('crawl_component_recrawl_never'), 1 => tl('crawl_component_recrawl_1day'), 2 => tl('crawl_component_recrawl_2day'), 3 => tl('crawl_component_recrawl_3day'), 7 => tl('crawl_component_recrawl_7day'), 14 => tl('crawl_component_recrawl_14day'));
$data['SIZE_VALUES'] = array(10000 => 10000, 50000 => 50000, 100000 => 100000, 500000 => 500000, 1000000 => 1000000, 5000000 => 5000000, 10000000 => 10000000);
$data['LEN_VALUES'] = array(2000 => 2000, 10000 => 10000, 50000 => 50000, 100000 => 100000, 500000 => 500000, 1000000 => 1000000, 5000000 => 5000000, 10000000 => 10000000);
$data['available_summarizers'] = array(self::BASIC_SUMMARIZER => tl('crawl_component_basic'), self::CENTROID_SUMMARIZER => tl('crawl_component_centroid'));
if (!isset($seed_info["indexed_file_types"]["extensions"])) {
$seed_info["indexed_file_types"]["extensions"] = $INDEXED_FILE_TYPES;
}
$loaded = false;
if (isset($_REQUEST['load_option']) && $_REQUEST['load_option'] > 0) {
if ($_REQUEST['load_option'] == 1) {
$seed_loaded = $crawl_model->getSeedInfo(true);
} else {
$timestamp = substr($parent->clean($_REQUEST['load_option'], "int"), 0, TIMESTAMP_LEN);
$seed_loaded = $crawl_model->getCrawlSeedInfo($timestamp, $machine_urls);
}
$copy_options = array("general" => array("page_recrawl_frequency", "page_range_request", "max_description_len", "cache_pages", 'summarizer_option'), "indexed_file_types" => array("extensions"), "indexing_plugins" => array("plugins", "plugins_data"));
foreach ($copy_options as $main_option => $sub_options) {
foreach ($sub_options as $sub_option) {
if (isset($seed_loaded[$main_option][$sub_option])) {
$seed_info[$main_option][$sub_option] = $seed_loaded[$main_option][$sub_option];
}
}
}
if (isset($seed_loaded['page_rules'])) {
$seed_info['page_rules'] = $seed_loaded['page_rules'];
}
if (isset($seed_loaded['active_classifiers'])) {
$seed_info['active_classifiers'] = $seed_loaded['active_classifiers'];
} else {
$seed_info['active_classifiers'] = array();
$seed_info['active_classifiers']['label'] = array();
}
$loaded = true;
} else {
$seed_info = $crawl_model->getSeedInfo();
if (isset($_REQUEST["page_recrawl_frequency"]) && in_array($_REQUEST["page_recrawl_frequency"], array_keys($data['RECRAWL_FREQS']))) {
$seed_info["general"]["page_recrawl_frequency"] = $_REQUEST["page_recrawl_frequency"];
}
if (isset($_REQUEST["page_range_request"]) && in_array($_REQUEST["page_range_request"], $data['SIZE_VALUES'])) {
$seed_info["general"]["page_range_request"] = $_REQUEST["page_range_request"];
}
if (isset($_REQUEST['summarizer_option']) && in_array($_REQUEST['summarizer_option'], array_keys($data['available_summarizers']))) {
$seed_info['general']['summarizer_option'] = $_REQUEST['summarizer_option'];
}
if (isset($_REQUEST["max_description_len"]) && in_array($_REQUEST["max_description_len"], $data['LEN_VALUES'])) {
$seed_info["general"]["max_description_len"] = $_REQUEST["max_description_len"];
}
if (isset($_REQUEST["cache_pages"])) {
$seed_info["general"]["cache_pages"] = true;
} else {
if (isset($_REQUEST['posted'])) {
//form sent but check box unchecked
$seed_info["general"]["cache_pages"] = false;
}
}
if (isset($_REQUEST['page_rules'])) {
$seed_info['page_rules']['rule'] = $parent->convertStringCleanArray($_REQUEST['page_rules'], 'rule');
}
}
if (!isset($seed_info["general"]["page_recrawl_frequency"])) {
$seed_info["general"]["page_recrawl_frequency"] = PAGE_RECRAWL_FREQUENCY;
}
$data['summarizer_option'] = $seed_info['general']['summarizer_option'];
$data['PAGE_RECRAWL_FREQUENCY'] = $seed_info["general"]["page_recrawl_frequency"];
if (!isset($seed_info["general"]["cache_pages"])) {
$seed_info["general"]["cache_pages"] = false;
}
$data["CACHE_PAGES"] = $seed_info["general"]["cache_pages"];
if (!isset($seed_info["general"]["page_range_request"])) {
$seed_info["general"]["page_range_request"] = PAGE_RANGE_REQUEST;
}
$data['PAGE_SIZE'] = $seed_info["general"]["page_range_request"];
if (!isset($seed_info["general"]["max_description_len"])) {
$seed_info["general"]["max_description_len"] = MAX_DESCRIPTION_LEN;
}
$data['MAX_LEN'] = $seed_info["general"]["max_description_len"];
$data['INDEXING_PLUGINS'] = array();
$included_plugins = array();
if (isset($_REQUEST["posted"]) && !$loaded) {
$seed_info['indexing_plugins']['plugins'] = isset($_REQUEST["INDEXING_PLUGINS"]) ? $_REQUEST["INDEXING_PLUGINS"] : array();
}
$included_plugins = isset($seed_info['indexing_plugins']['plugins']) ? $seed_info['indexing_plugins']['plugins'] : array();
foreach ($parent->indexing_plugins as $plugin) {
$plugin_name = ucfirst($plugin);
$data['INDEXING_PLUGINS'][$plugin_name]['checked'] = in_array($plugin_name, $included_plugins) ? "checked='checked'" : "";
/* to use method_exists we want that the require_once for the plugin
class has occurred so we instantiate the object via the plugin
method call which will also do the require if needed.
*/
$plugin_object = $parent->plugin(lcfirst($plugin_name));
$class_name = $plugin_name . "Plugin";
if ($loaded && method_exists($class_name, 'setConfiguration') && method_exists($class_name, 'loadDefaultConfiguration')) {
if (isset($seed_info['indexing_plugins']['plugins_data'][$plugin_name])) {
$plugin_object->setConfiguration($seed_info['indexing_plugins']['plugins_data'][$plugin_name]);
} else {
$plugin_object->loadDefaultConfiguration();
}
$plugin_object->saveConfiguration();
}
if (method_exists($class_name, 'configureHandler') && method_exists($class_name, 'configureView')) {
$data['INDEXING_PLUGINS'][$plugin_name]['configure'] = true;
$plugin_object->configureHandler($data);
} else {
$data['INDEXING_PLUGINS'][$plugin_name]['configure'] = false;
}
}
$profile = $profile_model->getProfile(WORK_DIRECTORY);
if (!isset($_REQUEST['load_option'])) {
$data = array_merge($data, $profile);
} else {
$parent->updateProfileFields($data, $profile, array('IP_LINK', 'CACHE_LINK', 'SIMILAR_LINK', 'IN_LINK', 'SIGNIN_LINK', 'SUBSEARCH_LINK', 'WORD_SUGGEST'));
}
$weights = array('TITLE_WEIGHT' => 4, 'DESCRIPTION_WEIGHT' => 1, 'LINK_WEIGHT' => 2, 'MIN_RESULTS_TO_GROUP' => 200, 'SERVER_ALPHA' => 1.6);
$change = false;
foreach ($weights as $weight => $value) {
if (isset($_REQUEST[$weight])) {
$data[$weight] = $parent->clean($_REQUEST[$weight], 'float', 1);
$profile[$weight] = $data[$weight];
$change = true;
} else {
if (isset($profile[$weight]) && $profile[$weight] != "") {
$data[$weight] = $profile[$weight];
} else {
$data[$weight] = $value;
$profile[$weight] = $data[$weight];
$change = true;
}
}
}
if ($change == true) {
$profile_model->updateProfile(WORK_DIRECTORY, array(), $profile);
}
$data['INDEXED_FILE_TYPES'] = array();
$filetypes = array();
foreach ($INDEXED_FILE_TYPES as $filetype) {
$ison = false;
if (isset($_REQUEST["filetype"]) && !$loaded) {
if (isset($_REQUEST["filetype"][$filetype])) {
$filetypes[] = $filetype;
$ison = true;
$change = true;
}
} else {
if (in_array($filetype, $seed_info["indexed_file_types"]["extensions"])) {
$filetypes[] = $filetype;
$ison = true;
}
}
$data['INDEXED_FILE_TYPES'][$filetype] = $ison ? "checked='checked'" : '';
}
$seed_info["indexed_file_types"]["extensions"] = $filetypes;
$data['CLASSIFIERS'] = array();
$data['RANKERS'] = array();
$active_classifiers = array();
$active_rankers = array();
foreach (Classifier::getClassifierList() as $classifier) {
$label = $classifier->class_label;
$ison = false;
if (isset($_REQUEST['classifier']) && !$loaded) {
if (isset($_REQUEST['classifier'][$label])) {
$ison = true;
}
} else {
if ($loaded || !isset($_REQUEST['posted']) && isset($seed_info['active_classifiers']['label'])) {
if (in_array($label, $seed_info['active_classifiers']['label'])) {
$ison = true;
}
}
}
if ($ison) {
$data['CLASSIFIERS'][$label] = 'checked="checked"';
$active_classifiers[] = $label;
} else {
$data['CLASSIFIERS'][$label] = '';
}
$ison = false;
if (isset($_REQUEST['ranker']) && !$loaded) {
if (isset($_REQUEST['ranker'][$label])) {
$ison = true;
}
} else {
if ($loaded || !isset($_REQUEST['posted']) && isset($seed_info['active_rankers']['label'])) {
if (isset($seed_info['active_rankers']['label']) && in_array($label, $seed_info['active_rankers']['label'])) {
$ison = true;
}
}
}
if ($ison) {
$data['RANKERS'][$label] = 'checked="checked"';
$active_rankers[] = $label;
} else {
$data['RANKERS'][$label] = '';
}
}
$parent->pagingLogic($data, 'CLASSIFIERS', 'CLASSIFIERS', DEFAULT_ADMIN_PAGING_NUM / 5, array(), "", array('name' => 'class_label'));
$seed_info['active_classifiers']['label'] = $active_classifiers;
$seed_info['active_rankers']['label'] = $active_rankers;
if (isset($seed_info['page_rules']['rule'])) {
if (isset($seed_info['page_rules']['rule']['rule'])) {
$data['page_rules'] = $parent->convertArrayLines($seed_info['page_rules']['rule']['rule']);
} else {
$data['page_rules'] = $parent->convertArrayLines($seed_info['page_rules']['rule']);
}
} else {
$data['page_rules'] = "";
}
$allowed_options = array('crawl_time', 'search_time', 'test_options');
if (isset($_REQUEST['option_type']) && in_array($_REQUEST['option_type'], $allowed_options)) {
$data['option_type'] = $_REQUEST['option_type'];
} else {
$data['option_type'] = 'crawl_time';
}
if ($data['option_type'] == 'crawl_time') {
$data['crawl_time_active'] = "active";
$data['search_time_active'] = "";
$data['test_options_active'] = "";
$data['SCRIPT'] .= "\nswitchTab('crawltimetab'," . "'searchtimetab', 'testoptionstab')\n";
} else {
if ($data['option_type'] == 'search_time') {
$data['search_time_active'] = "active";
$data['crawl_time_active'] = "";
$data['test_options_active'] = "";
$data['SCRIPT'] .= "\nswitchTab('searchtimetab'," . "'crawltimetab', 'testoptionstab')\n";
} else {
$data['search_time_active'] = "";
$data['crawl_time_active'] = "";
$data['test_options_active'] = "active";
$data['SCRIPT'] .= "\nswitchTab('testoptionstab'," . "'crawltimetab', 'searchtimetab');\n";
}
}
$crawl_model->setSeedInfo($seed_info);
if ($change == true && $data['option_type'] != 'test_options') {
$data['SCRIPT'] .= "doMessage('<h1 class=\"red\" >" . tl('crawl_component_page_options_updated') . "</h1>')";
}
$test_processors = array("text/html" => "html", "text/asp" => "html", "text/xml" => "xml", "text/robot" => "robot", "application/xml" => "xml", "application/xhtml+xml" => "html", "application/rss+xml" => "rss", "application/atom+xml" => "rss", "text/csv" => "text", "text/gopher" => "gopher", "text/plain" => "text", "text/rtf" => "rtf", "text/tab-separated-values" => "text");
$data['MIME_TYPES'] = array_keys($test_processors);
$data['page_type'] = "text/html";
if (isset($_REQUEST['page_type']) && in_array($_REQUEST['page_type'], $data['MIME_TYPES'])) {
$data['page_type'] = $_REQUEST['page_type'];
}
$data['TESTPAGE'] = isset($_REQUEST['TESTPAGE']) ? $parent->clean($_REQUEST['TESTPAGE'], 'string') : "";
if ($data['option_type'] == 'test_options' && $data['TESTPAGE'] != "") {
$data['SCRIPT'] .= "doMessage('<h1 class=\"red\" >" . tl('crawl_component_page_options_running_tests') . "</h1>')";
$site = array();
$site[self::ENCODING] = "UTF-8";
$site[self::URL] = "http://test-site.yioop.com/";
$site[self::IP_ADDRESSES] = array("1.1.1.1");
$site[self::HTTP_CODE] = 200;
$site[self::MODIFIED] = date("U", time());
$site[self::TIMESTAMP] = time();
$site[self::TYPE] = "text/html";
$site[self::HEADER] = "page options test extractor";
$site[self::SERVER] = "unknown";
$site[self::SERVER_VERSION] = "unknown";
$site[self::OPERATING_SYSTEM] = "unknown";
$site[self::LANG] = 'en';
$site[self::JUST_METAS] = false;
if (isset($_REQUEST['page_type']) && in_array($_REQUEST['page_type'], $data['MIME_TYPES'])) {
$site[self::TYPE] = $_REQUEST['page_type'];
}
if ($site[self::TYPE] == 'text/html') {
$site[self::ENCODING] = guessEncodingHtml($_REQUEST['TESTPAGE']);
}
$prefix_name = $test_processors[$site[self::TYPE]];
$processor_name = ucfirst($prefix_name) . "Processor";
$plugin_processors = array();
if (isset($seed_info['indexing_plugins']['plugins'])) {
foreach ($seed_info['indexing_plugins']['plugins'] as $plugin) {
$plugin_name = $plugin . "Plugin";
$supported_processors = $plugin_name::getProcessors();
foreach ($supported_processors as $supported_processor) {
$parent_processor = $processor_name;
do {
if ($supported_processor == $parent_processor) {
$plugin_object = $parent->plugin(lcfirst($plugin));
if (method_exists($plugin_name, "loadConfiguration")) {
$plugin_object->loadConfiguration();
}
$plugin_processors[] = $plugin_object;
break;
}
} while (($parent_processor = get_parent_class($parent_processor)) && $parent_processor != "PageProcessor");
}
}
}
$page_processor = new $processor_name($plugin_processors, $seed_info["general"]["max_description_len"], $seed_info["general"]["summarizer_option"]);
restore_error_handler();
$data["PAGE_RANGE_REQUEST"] = $seed_info["general"]["page_range_request"];
$doc_info = $page_processor->handle(substr($_REQUEST['TESTPAGE'], 0, $data["PAGE_RANGE_REQUEST"]), $site[self::URL]);
set_error_handler("yioop_error_handler");
if (!$doc_info) {
$data["AFTER_PAGE_PROCESS"] = "";
$data["AFTER_RULE_PROCESS"] = "";
$data["EXTRACTED_WORDS"] = "";
$data["EXTRACTED_META_WORDS"] = "";
return $data;
}
if ($processor_name != "RobotProcessor" && !isset($doc_info[self::JUST_METAS])) {
$doc_info[self::LINKS] = UrlParser::pruneLinks($doc_info[self::LINKS]);
}
foreach ($doc_info as $key => $value) {
$site[$key] = $value;
}
if (isset($site[self::PAGE])) {
unset($site[self::PAGE]);
}
if (isset($site[self::ROBOT_PATHS])) {
$site[self::JUST_METAS] = true;
}
$reflect = new ReflectionClass("CrawlConstants");
$crawl_constants = $reflect->getConstants();
$crawl_keys = array_keys($crawl_constants);
$crawl_values = array_values($crawl_constants);
$inverse_constants = array_combine($crawl_values, $crawl_keys);
$after_process = array();
foreach ($site as $key => $value) {
$out_key = isset($inverse_constants[$key]) ? $inverse_constants[$key] : $key;
$after_process[$out_key] = $value;
}
$data["AFTER_PAGE_PROCESS"] = wordwrap($parent->clean(print_r($after_process, true), "string"), 75, "\n", true);
$rule_string = implode("\n", $seed_info['page_rules']['rule']);
$rule_string = html_entity_decode($rule_string, ENT_QUOTES);
$page_rule_parser = new PageRuleParser($rule_string);
$page_rule_parser->executeRuleTrees($site);
$after_process = array();
foreach ($site as $key => $value) {
$out_key = isset($inverse_constants[$key]) ? $inverse_constants[$key] : $key;
$after_process[$out_key] = $value;
}
$data["AFTER_RULE_PROCESS"] = wordwrap($parent->clean(print_r($after_process, true), "string"), 75, "\n", true);
$lang = NULL;
if (isset($site[self::LANG])) {
$lang = $site[self::LANG];
}
$meta_ids = PhraseParser::calculateMetas($site);
if (!$site[self::JUST_METAS]) {
$host_words = UrlParser::getWordsIfHostUrl($site[self::URL]);
$path_words = UrlParser::getWordsLastPathPartUrl($site[self::URL]);
$phrase_string = $host_words . " " . $site[self::TITLE] . " " . $path_words . " " . $site[self::DESCRIPTION];
if ($site[self::TITLE] != "") {
$lang = guessLocaleFromString($site[self::TITLE], $lang);
} else {
$lang = guessLocaleFromString(substr($site[self::DESCRIPTION], 0, AD_HOC_TITLE_LENGTH), $lang);
}
$word_lists = PhraseParser::extractPhrasesInLists($phrase_string, $lang);
$len = strlen($phrase_string);
if (PhraseParser::computeSafeSearchScore($word_lists, $len) < 0.012) {
$meta_ids[] = "safe:true";
$safe = true;
} else {
$meta_ids[] = "safe:false";
$safe = false;
}
}
if (!isset($word_lists)) {
$word_lists = array();
}
$data["EXTRACTED_WORDS"] = wordwrap($parent->clean(print_r($word_lists, true), "string"), 75, "\n", true);
$data["EXTRACTED_META_WORDS"] = wordwrap($parent->clean(print_r($meta_ids, true), "string"), 75, "\n", true);
}
return $data;
}
/**
* Splits an http response document into the http headers sent
* and the web page returned. Parses out useful information from
* the header and return an array of these two parts and the useful info.
*
* @param string& $header_and_page reference to string of downloaded data
* @param string $value field to store the page portion of page
* @return array info array consisting of a header, page for an http
* response, as well as parsed from the header the server, server
* version, operating system, encoding, and date information.
*/
static function parseHeaderPage(&$header_and_page, $value = CrawlConstants::PAGE)
{
$cache_page_validators = array();
$cache_page_validators['etag'] = -1;
$cache_page_validators['expires'] = -1;
$new_offset = 0;
// header will include all redirect headers
$site = array();
$site[CrawlConstants::LOCATION] = array();
do {
$continue = false;
$CRLFCRLF = strpos($header_and_page, "\r\n\r\n", $new_offset);
$LFLF = strpos($header_and_page, "\n\n", $new_offset);
//either two CRLF (what spec says) or two LF's to be safe
$old_offset = $new_offset;
$header_offset = $CRLFCRLF > 0 ? $CRLFCRLF : $LFLF;
$header_offset = $header_offset ? $header_offset : 0;
$new_offset = $CRLFCRLF > 0 ? $header_offset + 4 : $header_offset + 2;
$redirect_pos = stripos($header_and_page, 'Location:', $old_offset);
$redirect_str = "Location:";
if ($redirect_pos === false) {
$redirect_pos = stripos($header_and_page, 'Refresh:', $old_offset);
$redirect_str = "Refresh:";
}
if (isset($header_and_page[$redirect_pos - 1]) && ord($header_and_page[$redirect_pos - 1]) > 32) {
$redirect_pos = $new_offset;
//ignore X-XRDS-Location header
} else {
if ($redirect_pos !== false && $redirect_pos < $new_offset) {
$redirect_pos += strlen($redirect_str);
$pre_line = substr($header_and_page, $redirect_pos, strpos($header_and_page, "\n", $redirect_pos) - $redirect_pos);
$loc = @trim($pre_line);
if (strlen($loc) > 0) {
$site[CrawlConstants::LOCATION][] = @$loc;
}
$continue = true;
}
}
} while ($continue);
if ($header_offset > 0) {
$site[CrawlConstants::HEADER] = substr($header_and_page, 0, $header_offset);
$site[$value] = ltrim(substr($header_and_page, $header_offset));
} else {
//header message no body; maybe 301?
$site[CrawlConstants::HEADER] = $header_and_page;
$site[$value] = " ";
}
$lines = explode("\n", $site[CrawlConstants::HEADER]);
$first_line = array_shift($lines);
$response = preg_split("/(\\s+)/", $first_line);
$site[CrawlConstants::HTTP_CODE] = @trim($response[1]);
$site[CrawlConstants::ROBOT_METAS] = array();
foreach ($lines as $line) {
$line = trim($line);
if (stristr($line, 'Server:')) {
$server_parts = preg_split("/Server\\:/i", $line);
$server_name_parts = @explode("/", $server_parts[1]);
$site[CrawlConstants::SERVER] = @trim($server_name_parts[0]);
if (isset($server_name_parts[1])) {
$version_parts = explode("(", $server_name_parts[1]);
$site[CrawlConstants::SERVER_VERSION] = @trim($version_parts[0]);
if (isset($version_parts[1])) {
$os_parts = explode(")", $version_parts[1]);
$site[CrawlConstants::OPERATING_SYSTEM] = @trim($os_parts[0]);
}
}
}
if (stristr($line, 'Content-type:')) {
list(, $mimetype, ) = preg_split("/:|;/i", $line);
$site[CrawlConstants::TYPE] = trim($mimetype);
}
if (stristr($line, 'charset=')) {
$line_parts = preg_split("/charset\\=/i", $line);
$site[CrawlConstants::ENCODING] = strtoupper(@trim($line_parts[1]));
}
if (stristr($line, 'Last-Modified:')) {
$line_parts = preg_split("/Last\\-Modified\\:/i", $line);
$site[CrawlConstants::MODIFIED] = strtotime(@trim($line_parts[1]));
}
if (stristr($line, 'X-Robots-Tag:')) {
// robot directives pdfs etc
$line_parts = preg_split("/X\\-Robots\\-Tag\\:/i", $line);
$robot_metas = explode(",", $line_parts[1]);
foreach ($robot_metas as $robot_meta) {
$site[CrawlConstants::ROBOT_METAS][] = strtoupper(trim($robot_meta));
}
}
$canonical_regex = "/Link\\:\\s*\\<\\s*(http.*)\\s*\\>\\s*\\;\\s*" . "rel\\s*\\=\\s*(\"|')?canonical(\"|')?/";
// levenshtein gives notices on strings longer than 255
if (preg_match($canonical_regex, $line, $matches) && isset($site[CrawlConstants::URL]) && strlen($matches[1]) < 252 && (strlen($site[CrawlConstants::URL]) >= 255 || levenshtein($matches[1], $site[CrawlConstants::URL]) > 3)) {
// for rel canonical headers
$site[CrawlConstants::LOCATION][] = $matches[1];
$site[CrawlConstants::ROBOT_METAS][] = 'NOFOLLOW';
}
if (USE_ETAG_EXPIRES && stristr($line, 'ETag:')) {
$line_parts = preg_split("/ETag\\:/i", $line);
if (isset($line_parts[1])) {
$etag_data = explode(" ", $line_parts[1]);
if (isset($etag_data[1])) {
$etag = $etag_data[1];
$cache_page_validators['etag'] = $etag;
}
}
}
if (USE_ETAG_EXPIRES && stristr($line, 'Expires:')) {
$line_parts = preg_split("/Expires\\:/i", $line);
$all_dates = $line_parts[1];
$date_parts = explode(",", $all_dates);
if (count($date_parts) == 2) {
$cache_page_validators['expires'] = strtotime($date_parts[1]);
} else {
if (count($date_parts) > 2) {
/*Encountered some pages with more than one Expires date
:O */
$timestamps = array();
for ($i = 1; $i < count($date_parts); $i += 2) {
$ds = strtotime($date_parts[$i]);
$timestamps[] = $ds;
}
$lowest = min($timestamps);
$cache_page_validators['expires'] = $lowest;
}
}
}
if (USE_ETAG_EXPIRES && !($cache_page_validators['etag'] == -1 && $cache_page_validators['expires'] == -1)) {
$site[CrawlConstants::CACHE_PAGE_VALIDATORS] = $cache_page_validators;
}
}
/*
If the doc is HTML and it uses a http-equiv to set the encoding
then we override what the server says (if anything). As we
are going to convert to UTF-8 we remove the charset info
from the meta tag so cached pages will display correctly and
redirects without char encoding won't be given a different hash.
*/
$encoding_info = guessEncodingHtml($site[$value], true);
if (is_array($encoding_info)) {
list($site[CrawlConstants::ENCODING], $start_charset, $len_c) = $encoding_info;
$site[$value] = substr_replace($site[$value], "", $start_charset, $len_c);
} else {
$site[CrawlConstants::ENCODING] = $encoding_info;
}
if (!isset($site[CrawlConstants::SERVER])) {
$site[CrawlConstants::SERVER] = "unknown";
}
return $site;
}
