Use hook_cron instead of hook_exit

Strange that you can't reproduce the problem. I had the same problem with a module written by me. I've implemented hook_init() and the message appeared as well. I thought only using aggressive caching can cause problems. So right now I don't know what to do. Do you think it's a big problem using hook_cron()? I'll try it today. Thanks.

Edit: I think I know why my module was throwing errors. Some guys discuss it here: http://drupal.org/node/140573. If bootstrap is set in the system table it should work. I've implemented hook_init() after enabling the module. So bootstrap is still not set. I've checked my blog_reactions installation but there bootstrap is set... So I'll try moving my the functions form hook_exit() to hook_cron().

Okay I've checked my bootstrap settings. As I said bootstrap is set to 1 for blog_reactions. I still get the error message when I refresh the browser. Very strange. Do you know another setting which should cause the problem using normal caching? Right now I think using the code to hook_cron is a lot of work isn't it? Thanks, Bjoern

We have some of the things right? But it sounds like you don't want to do it because the module is working for you and there's no point for changing it?

Okay I'll see what I can do.

Edit: This is working for me. What do you think (I don't want to provide a patch because I've changed some more lines).

Removed hook_exit() and implemented hook_cron().

/**
 * Implementation of hook_cron().
 */
function blog_reactions_cron() {
  $limit = variable_get('blog_reactions_cron_quantity', 20);
  $where = "'".implode("', '", variable_get('blog_reactions_node_types', array('blog')))."'";

  $result = db_query_range("SELECT nid FROM {node} WHERE type IN(".$where.") ORDER BY changed ASC", 0, $limit);
  while ($node = db_fetch_object($result)) {
    $node = node_load(array('nid' => $node->nid));
    if (($node->blog_reactions['last_update']+(60*60)) <= time()) {
      $uri = url('node/'. $node->nid, NULL, NULL, TRUE);
      $reactions = blog_reactions_fetch($node->nid, $uri);
      blog_reactions_set_items($reactions);
      watchdog('blog_reactions', t('Node @title (@type) was updated', array('@title' => $node->title, '@type' => $node->type)), WATCHDOG_NOTICE);
    }
  }
}

Furthermore I've added a new field to blog_reactions_admin_settings().

  $form['blog_reactions_cron_quantity'] = array(
    '#type' => 'select',
    '#title' => t('Number of nodes per cron run'),
    '#options' => drupal_map_assoc(array(10,20,30,40,50,75,100)),
    '#default_value' => variable_get('blog_reactions_cron_quantity', 20),
    '#description' => t('Choose the number of nodes wich will be updated per cron run. Default is 20.'),
  );

I've thought about the cron and your comment

* Check which nodes haven't been updated the longest
* Track last nid via variable_set() and respect chosen node types

I think my code is not fulfilling all your requirements. I'm just checking for last node.changed. But I've seen that updating blog reactions has no impact on node.changed. Now I understand why we need an changed date. Can't we use the field blog_reactions.updated? If it doesn't exist or if the date is too long ago we have to run the blog_reactions_fetch() and blog_reactions_set_items() functions.

Your code is just great. Thanks a lot! But one problem is left. The query doesn't select new nodes, i.e. if there's no nid/timestamp in the table blog_reactions the cron won't check for blog reactions for this node. Is this a problem of the inner join? So what can we do to check also new nodes which haven't been checked yet?

Edit: I've changed "INNER JOIN" to "LEFT JOIN" and now four new nodes were checked and added to the blog_reactions table.

Hi there,

I've tested your new code. Thanks a lot! I've set blog_reactions_cron_quantity to 5 nodes per cron and it worked very well. I'll let it run on my system and I'll inform you how it's working.

Awesome!