Load all nodes of comments at once

Hm...

- This conflicts with #1778178: Convert comments to the new Entity Field API, you're not supposed to add arbitrary stuff to $comments anmore
- Note that it only helps when actually calling comment_view_multiple(), which views is currently not.
- I thin kthe comment render controller would be a better place for this.

That said, we don't actually *have* to attach it. It's enough if we would just call node_load_multiple() to prepare the entity cache, then it will be loaded from the cache anyway. That could even happen within the view class and without having to use comment_view_multiple()...

+++ b/core/modules/comment/lib/Drupal/comment/CommentRenderController.phpundefined
@@ -29,8 +29,15 @@ public function buildContent(array $entities = array(), $view_mode = 'full', $la
+    $nodes =node_load_multiple($nids);

Missing space :)

#7: drupal-1827582-7.patch queued for re-testing.

+++ b/core/modules/comment/lib/Drupal/comment/CommentRenderController.phpundefined
@@ -32,8 +32,15 @@ public function buildContent(array $entities = array(), $view_mode = 'full', $la
+    // Collect all nodes and load them at once.
+    $nids = array();
     foreach ($entities as $entity) {
-      $node = node_load($entity->nid);
+      $nids[$entity->nid] = $entity->nid;
+    }
+    $nodes = node_load_multiple(array_unique($nids));

Sorry, another one.

The array_unique() is not necessary, you already use the nid as key.

Not sure if we can do better with the comment here, but I don't know how, so I guess it's fine.

+++ b/core/modules/comment/lib/Drupal/comment/CommentRenderController.phpundefined
@@ -32,8 +32,15 @@ public function buildContent(array $entities = array(), $view_mode = 'full', $la
+    foreach ($entities as $entity) {
+      $node = $nodes[$entity->nid];
       if (!$node) {
         throw new \InvalidArgumentException(t('Invalid node for comment.'));

Hm. If $node doesn't exist, then this results in a php notice. We should move the if (!$node) up and do a if (!isset($nodes[$entity->nid]) instead.

Looks good now :)

I would also argue that this is a (performance-improving) task, not a feature, it shouldn't be blocked because we're currently over thresholds.

Agreed.

Committed and pushed to 8.x. Thanks!