Automatically adjust feed frequencies

This commit is contained in:
Michael 2020-08-16 17:59:37 +00:00
parent a7f8604d42
commit b7e8c1595d
1 changed files with 130 additions and 10 deletions

View File

@ -300,6 +300,7 @@ class Feed
}
$items = [];
$creation_dates = [];
// Limit the number of items that are about to be fetched
$total_items = ($entries->length - 1);
@ -354,16 +355,6 @@ class Feed
$item["parent-uri"] = $item["uri"];
if (!$dryRun) {
$condition = ["`uid` = ? AND `uri` = ? AND `network` IN (?, ?)",
$importer["uid"], $item["uri"], Protocol::FEED, Protocol::DFRN];
$previous = Item::selectFirst(['id'], $condition);
if (DBA::isResult($previous)) {
Logger::info("Item with uri " . $item["uri"] . " for user " . $importer["uid"] . " already existed under id " . $previous["id"]);
continue;
}
}
$item["title"] = XML::getFirstNodeValue($xpath, 'atom:title/text()', $entry);
if (empty($item["title"])) {
@ -403,6 +394,19 @@ class Feed
$item["edited"] = $updated;
}
if (!$dryRun) {
$condition = ["`uid` = ? AND `uri` = ? AND `network` IN (?, ?)",
$importer["uid"], $item["uri"], Protocol::FEED, Protocol::DFRN];
$previous = Item::selectFirst(['id', 'created'], $condition);
if (DBA::isResult($previous)) {
// Use the creation date when the post had been stored. It can happen this date changes in the feed.
$creation_dates[] = $previous['created'];
Logger::info("Item with uri " . $item["uri"] . " for user " . $importer["uid"] . " already existed under id " . $previous["id"]);
continue;
}
$creation_dates[] = DateTimeFormat::utc($item['created']);
}
$creator = XML::getFirstNodeValue($xpath, 'author/text()', $entry);
if (empty($creator)) {
@ -598,9 +602,125 @@ class Feed
}
}
if (!$dryRun) {
self::adjustPollFrequency($contact, $creation_dates);
}
return ["header" => $author, "items" => $items];
}
/**
* Automatically adjust the poll frequency according to the post frequency
*
* @param array $contact
* @param array $creation_dates
* @return void
*/
private static function adjustPollFrequency(array $contact, array $creation_dates)
{
if (($contact['priority'] > 3) || ($contact['network'] != Protocol::FEED)) {
Logger::info('Contact is no feed or has a low priority, skip.', ['id' => $contact['id'], 'uid' => $contact['uid'], 'url' => $contact['url'], 'priority' => $contact['priority'], 'network' => $contact['network']]);
return;
}
if (!empty($creation_dates)) {
// Count the post frequency and the earliest and latest post date
$frequency = [];
$oldest = time();
$newest = 0;
$oldest_date = $newest_date = '';
foreach ($creation_dates as $date) {
$timestamp = strtotime($date);
$day = intdiv($timestamp, 86400);
$hour = $timestamp % 86400;
// Only have a look at values from the last seven days
if (((time() / 86400) - $day) < 7) {
if (empty($frequency[$day])) {
$frequency[$day] = ['count' => 1, 'low' => $hour, 'high' => $hour];
} else {
++$frequency[$day]['count'];
if ($frequency[$day]['low'] > $hour) {
$frequency[$day]['low'] = $hour;
}
if ($frequency[$day]['high'] < $hour) {
$frequency[$day]['high'] = $hour;
}
}
}
if ($oldest > $day) {
$oldest = $day;
$oldest_date = $date;
}
if ($newest < $day) {
$newest = $day;
$newest_date = $date;
}
}
if (($newest == $oldest) && count($creation_dates) > 1) {
Logger::info('Feed has no different creation dates, quitting', ['date' => $newest_date, 'id' => $contact['id'], 'uid' => $contact['uid'], 'url' => $contact['url']]);
return;
}
if (((time() / 86400) - $newest) > 30) {
Logger::info('Feed had not posted for a month, switching to daily polling', ['newest' => $newest_date, 'id' => $contact['id'], 'uid' => $contact['uid'], 'url' => $contact['url']]);
$priority = 3; // Poll once a day
}
if (count($creation_dates) == 1) {
Logger::info('Feed had posted a single time, switching to daily polling', ['newest' => $newest_date, 'id' => $contact['id'], 'uid' => $contact['uid'], 'url' => $contact['url']]);
$priority = 3; // Poll once a day
}
if (empty($priority) && (($newest - $oldest) > count($creation_dates))) {
Logger::info('Less than a post per day, switching to daily polling', ['posts' => count($creation_dates), 'oldest' => $oldest_date, 'newest' => $newest_date, 'id' => $contact['id'], 'uid' => $contact['uid'], 'url' => $contact['url']]);
$priority = 3; // Poll once a day
}
if (empty($priority)) {
// Calculate the highest "posts per day" value
$max = 1;
foreach ($frequency as $entry) {
if (($entry['count'] == 1) || ($entry['high'] == $entry['low'])) {
continue;
}
// We take the earliest and latest post day and interpolate the number of post per day
// that would had been created with this post frequency
// Assume at least four hours between oldest and newest post per day - should be okay for news outlets
$duration = max($entry['high'] - $entry['low'], 14400);
$ppd = (86400 / $duration) * $entry['count'];
if ($ppd > $max) {
$max = $ppd;
}
}
if ($max >= 24) {
$priority = 0; // Poll with the minimum poll intervall
} elseif ($max >= 12) {
$priority = 1; // Poll hourly
} elseif ($max > 1) {
$priority = 2; // Poll twice a day
} else {
/// @todo In the future we could calculate the days between the posts to set even lower priorities
$priority = 3; // Poll once a day
}
Logger::info('Calculated priority by the posts per day', ['priority' => $priority, 'max' => round($max, 2), 'id' => $contact['id'], 'uid' => $contact['uid'], 'url' => $contact['url']]);
}
} else {
Logger::info('No posts, switching to daily polling', ['id' => $contact['id'], 'uid' => $contact['uid'], 'url' => $contact['url']]);
$priority = 3; // Poll once a day
}
if ($contact['priority'] != $priority) {
Logger::info('Adjusting priority', ['old' => $contact['priority'], 'new' => $priority, 'id' => $contact['id'], 'uid' => $contact['uid'], 'url' => $contact['url']]);
}
}
/**
* Convert a tag array to a tag string
*