[drupal-devel] CCK Node Caching Strategies

[Gerhard Killesreiter]

On Sat, 5 Mar 2005, Jonathan Chaffer wrote:

> The CCK implementation is going to require a resurrection of node
> caching,

Yay!

[...]

> I think we can do a bit better, though. By placing the node cache into
> a column in the node table, we can grab it at the same time as the node
> IDs. Then node_load() would look to see if the cache is present in the
> array presented to it, and if so use that rather than perform any
> queries at all.
>      1x query to fetch node IDs and cache for the ten nodes (10 results
> + deserialize)

So, what you will do is:
$result = db_query("SELECT ....");
while ($node = db_fetch_object($result)) {
  if($node->cache == '') {
    $node = node_load(array('nid' => $node->nid), 'no cache');
  }
  else {
    $node = unserialize($node->cache);
  }
}

?

> Note that we can use this method with the cache in the cache table as
> well, but that requires a JOIN using a CONCAT() in the ON clause, which
> seems risky, and involves direct knowledge of the cache table structure
> rather than using the cache API anyway.
>
> Any other ideas?

I've never liked the structure of the cache table much. It is a one size
fits all thing. Why not split is up?

cache_page   key: url
cache_filter key: md5 of string
cache_node   key: nid
cache_menu   key: uid
cache_misc   key: as now

This way we can keep the cache out of the node table (where it does not
really belong, IMHO).

I am pretty sure that SELECTs from smaller tables will also be faster.

The unknown number of entries of the filter cache makes me feel
uncomfortable.

Cheers,
	Gerhard