Memory problem for large taxonomies in taxonomy_save()

There are many possible tokens that could be using the term's parent, so we can't just skip that chunk if [catpath] is used. This will be solved with #810294: Add an 'URL alias update queue' for entities associated with an updated entity, so I'm tempted to mark it as a duplicate.

subscribe

Perhaps it is just my personal use of taxonomies but huge amount of terms is generally linked with free tagging. An free tagging vocabularies don't have any hierarchy...

In this idea, a solution could by to generate children url aliases only when associated vocabulary is not defined as "free tagging" kind. This can be done by modifying the code like this :

          // For all children generate new alias (important if [catpath] used)
          $vocabulary = taxonomy_get_vocabulary($term->vid);
          if ($vocabulary->tags!=1) {
             foreach (taxonomy_get_tree($term->vid, $term->tid) as $subcategory) {
               $count = _taxonomy_pathauto_alias($subcategory, $op);
             }
          }

I agree with Yoran. There needs to be some way of handling large free tag vocabularies. His solution works for me.

Free tagged terms can still be organized into parents and children in the term listing screen right?

Necroing this sucker like a madman (as I recently worked on a site with 350k+ terms in a vocab).

Yes, even if you don't use a hierarchy, there's technically no way to forcefully disable hierarchical terms in core. Vocabularies have a hierarchy property on them, but it's purely reflective of whether or not any term in the vocabulary has one ($vocabulary->hierarchy = 1) or more ($vocabulary->hierarchy = 2) parents. It gets updated when the term form is submitted and a term's parents have changed or when a term is deleted.

I implemented a variety of workarounds on that site (including disabling changing parents on terms and deleting terms since they were remotely managed). I also had to workaround this use of taxonomy_get_tree() in Pathauto by unregistering the hook and then selectively invoking it for unmanaged vocabularies:

/**
 * Implements hook_module_implements_alter().
 */
function example_module_implements_alter(&$implementations, $hook) {
  // If hook_taxonomy_term_update() is being invoked, look to see if the
  // Pathauto implementation is due to be called. If so, unset it, because it
  // calls taxonomy_get_tree(), which will result in memory errors as the
  // large taxonomy vocabulary (100k+ terms) is populated. This module's
  // own implementation of the hook will call the Pathauto function in a
  // more performant way for terms in this vocabulary.
  if ($hook == 'taxonomy_term_update' && isset($implementations['pathauto'])) {
    unset($implementations['pathauto']);
  }
}

/**
 * Implements hook_taxonomy_term_update().
 */
function example_taxonomy_term_update($term) {
  // Check for the Pathauto module, as this hook solely exists to provide a
  // more performant alternative for the large vocabulary.
  if (module_exists('pathauto')) {
    // If the term is in the large vocabulary...
    if ($term->vocabulary_machine_name == 'large_vocabulary_name') {
      // To avoid the call to taxonomy_get_tree(), we disable updating the
      // aliases of child terms. Sites using term aliases that include parent
      // term tokens that may be updated during imports will need to implement
      // an alternate update strategy.
      pathauto_taxonomy_term_update_alias($term, 'update', array('alias children' => FALSE));
    }
    else {
      pathauto_taxonomy_term_update_alias($term, 'update', array('alias children' => TRUE));
    }
  }
}

Leaving it here in case it helps someone else and turning my attentions toward the referenced issue. (I may also look at updating child terms without the use of taxonomy_get_tree().)

#2114323: Field based aliases incorrectly generated for taxonomy term children when updating top level term has been committed which switches to use taxonomy_get_children(). The rest will be handled by #810294: Add an 'URL alias update queue' for entities associated with an updated entity.