Article page load spending 1.4 seconds doing PHP things.

[Alfa1]

When looking at my VW articles in debug mode, I find that there is quite a difference between articles in loading time. Its not the queries, as they are mostly quick. One of my most viewed articles (50.000 views) gives this result:

Page generated in 1.48419094086 seconds with 45 queries, spending 0.0618517398834 doing MySQL queries and 1.42233920097 doing PHP things.

Do you know what causes this?

[pegasus]

Did you find the issue? All of the pages on your site are taking less than a second for PHP, and most articles are taking less than .2 seconds.

[Alfa1]

Not yet. Its not server strain.
Last night my server load was going trough the roof(54!) and my site became unreachable. There where only 6.500 users, while 10k users is normally doing OK.
Today server load is normal again and php page load is still 1.4 seconds

[pegasus]

This slowdown occurs because the longer an article is and the number of BB-Codes used make auto-link discovery more difficult. We intend to add an option to disable auto-links on a per-article basis, but not until VaultWiki 4.

I have devised this, which will improve auto-link speed by about 50%, but I cannot speak to the accuracy of the discovery mechanism after this change. Only time and testing will tell. I don't think it's possible to improve the speed much more without serious repercussions to accuracy (i.e. creating links where links don't belong).

In vault/class/bbcode/autolink.php, replace the complete

function process

(Everything until the next function definition) with:

	function process($text, $parent, $use_html = false)
	{
		global $vbulletin, $vault;

		$pregfind = array();
		$pregreplace = array();
		$wiki_links = array();

		$force_return = false;

		if (
			$parent == CATEGORY_SPACE OR
			$parent == IMAGE_SPACE OR
			$parent == TEMPLATE_SPACE
		)
		{
			$force_return = true;
		}

		if (!$force_return)
		{
			$vault->fetch_instance('article');

			if (!$vault->article->fetch('autolinks', $parent, VAULT_LANGUAGEID, 'count'))
			{
				$force_return = true;
			}
		}

		if (!$force_return)
		{
			$this->init_tags();
			$preg_no_auto = array();

			if (!empty($this->skip_tags))
			{
				$preg_no_auto[] = '#\[(' . implode('|', $this->skip_tags) . ')\].*\[/\1\]#siUe';
			}

			$preg_no_auto[] = '#\[[^\[]*\]#sUe';

			if ($use_html)
			{
				$preg_no_auto[] = '#<(a|script|code).*</\1>#siUe';
				$preg_no_auto[] = '#<.*>#sUe';
			}

			if ($vbulletin->options['vault_legacy_enabled'] == 1)
			{
				$preg_no_auto[] = '#\{.*(\||\})#sUe';
			}
		}

		($hook = vBulletinHook::fetch_hook('vault_autolinks_start')) ? eval($hook) : false;

		if ($force_return)
		{
			return $text;
		}

		global $post;
		static $redirects_to_skip;

		$articles_to_skip = $vault->parser['autolink_articles_to_skip'];
		$matching_articles = array();
		$text = htmlspecialchars_uni($text);

		$vault->fetch_instance('string');
		$lowertext = $vault->string->tolower($text);

		// empty this in case we somehow parse multiple articles on the same page
		$vault->parser['autolink_articles_to_skip'] = array();
		$linkids = $vault->article->fetch('autolinks', $parent, VAULT_LANGUAGEID, 'dump');
		$entities = get_html_translation_table(HTML_ENTITIES);

		if ($vbulletin->options['vault_autolink_stop'])
		{
			$ignore_list = preg_split('#[\r\n]+#s', $vault->string->tolower($vbulletin->options['vault_autolink_stop']), -1, PREG_SPLIT_NO_EMPTY);
		}

		if (empty($ignore_list))
		{
			$ignore_list = false;
		}

		// turn on UTF-8 mode?
		$vault->fetch_instance('language');
		$langid = $vault->language->get(0);
		$modifiers = 'iU';

		if ($vault->language->charset["$langid"] == 'U')
		{
			$modifiers .= 'u';
		}

		foreach ($linkids AS $linkid => $linktitle)
		{
			$linktitle = $linktitle[VAULT_LANGUAGEID];

			if (!$linktitle)
			{
				continue;
			}

			$linktitle = $vault->string->tolower($linktitle);

			if (
				strlen($linktitle) < $vbulletin->options['vault_autolink_min'] OR
				$linktitle == 'ces' OR
				$linktitle == 'dol' OR
				$linktitle == 'xst' OR
				$linktitle == 'l' OR
				$linktitle == 'autolink' OR
				intval($linktitle) > 0 OR
				(
					$ignore_list AND
					in_array($linktitle, $ignore_list)
				) OR
				strpos($lowertext, $linktitle) === false OR
				(
					in_array($linktitle, $entities) OR
					(
						$number = str_replace(array('#', 'x', 'u', ';'), '', $number) AND
						intval($number) > 0
					)
				)
			)
			{
				continue;
			}

			$matching_articles[strlen($linktitle)]["$linkid"] = preg_quote($linktitle, '#');
		}

		krsort($matching_articles);
		$text = str_replace('$', '#!#CES-XST()DOL#L', $text);
		$text = preg_replace($preg_no_auto, '\$this->save("\0")', $text);

		if (
			!empty($linkids) AND
			!empty($articles_to_skip) AND
			VAULT == 'true' AND
			empty($redirects_to_skip[$post['threadid']])
		)
		{
			$redirects_to_skip[$post['threadid']] = array();

			$redirects = $vbulletin->db->query_read_slave("
				SELECT thread.threadid
				FROM " . TABLE_PREFIX . "thread AS thread
				LEFT JOIN " . TABLE_PREFIX . "vault_link AS link ON (thread.firstpostid = link.postid)
				WHERE thread.threadid IN (" . implode(',', array_keys($linkids)) . ")
					AND link.redirect = 1
					AND link.targetid IN (" . implode(',', array_keys($articles_to_skip)) . ")
				GROUP BY thread.threadid
			");

			if ($vbulletin->db->num_rows($redirects))
			{
				while ($redirect = $vbulletin->db->fetch_array($redirects))
				{
					$redirects_to_skip[$post['threadid']][$redirect['threadid']] = $redirect['threadid'];
				}
				$vbulletin->db->free_result($redirects);
				unset($redirect);
			}
			else
			{
				$redirects_to_skip[$post['threadid']] = true;
			}
		}

		if (is_array($redirects_to_skip[$post['threadid']]))
		{
			// numerical keys aren't preserved with array_merge
			foreach ($redirects_to_skip[$post['threadid']] AS $redkey => $redval)
			{
				$articles_to_skip["$redkey"] = $redval;
			}
		}

		$i = 0;
		$totallen = 0;
		$search = array();

		foreach ($matching_articles AS $article_group)
		{
			foreach ($articles_to_skip AS $linkid => $ditto)
			{
				unset($article_group["$linkid"]);
			}

			if (empty($article_group))
			{
				continue;
			}

			$search = array_merge($search, $article_group);
		}

		if (!empty($search))
		{
			$searchstring = '#((?:^|[^\&\w]){1})(' . implode('|', $search) . ')((?:$|[^\w]){1})#' . $modifiers;
			$text = preg_replace($searchstring, "\\1[AUTOLINK]\\2[/AUTOLINK]\\3", $text);
		}

		$text = $this->save('', $text, 1);
		$text = str_replace('#!#CES-XST()DOL#L', '$', $text);
		$text = unhtmlspecialchars($text);

		return $text;
	}

[Alfa1]

Is there a possibility to set a maximum of 1 instance per autolink per page?
If not, then I really think this should be added.

[pegasus]

The maximum you speak of is automatically applied, but not at the discovery level. Believe it or not, filtering during discovery takes longer than just discovering all locations and then ignoring duplicates.

In my previous post, I was talking about the accuracy of auto-links being created where they shouldn't, like inside code blocks, noparse sections, or links that were explicitly made, or auto-link matches not being discovered where they should.

[Alfa1]

I understand. NuWiki had that annoying issue, my members frequently complained about. i.e. if you have an article for 'bear' and someone writes 'bearing', then the first 4 characters of the word 'bearing' would become an autolink to the article. This becomes more annoying when you have similar articles.

[InformationNirvana]

another important rule of wikis is ... keep the articles short !

[pegasus]

Yes, I have a test article that is over 500,000 characters long and it takes over 10 seconds to load. Not much that can be done on the coding side, the article is just too long.

[Alfa1]

But the whole point of it is to accumulate comprehensive articles. Hahaha.

[InformationNirvana]

But the whole point of it is to accumulate comprehensive articles. Hahaha.

Break up comprehensive articles into smaller articles.