Improvement of Category tree loading

[naydav]

In the scope of #121 (Fix regexp when filter by root category id) was fixed functionality of category tree loading

but we got some performance degradation. It was expected due to the loading of all children leads to using more resources

Query Category                                                   233.00ms   +45.6%     (+73ms)    DEGRADATION

created-objects
Query Category (Catalog GraphQL): 128.9% (mainline = 1057 | team = 2420) -> DEGRADATION

tideways-mysql-select-requests-count
Query Category (Catalog GraphQL): 120.8% (mainline = 77 | team = 170) -> DEGRADATION

Need to investigate and provide prototype how we can load all of the needed data in the more efficient way.

[pmclain]

Most of the additional overhead is coming from loading advanced category attribute data when it hasn't been requested

graphql-ce/app/code/Magento/CatalogGraphQl/Model/Category/Hydrator.php

Line 54 in 3694c65

$categoryData = $this->dataObjectProcessor->buildOutputDataArray($category, CategoryInterface::class);

The category resolver gets around this by only requesting the additional attributes when included in the request

graphql-ce/app/code/Magento/CatalogGraphQl/Model/Resolver/Categories.php

Lines 116 to 121 in ac93407

	if (count($requestedFields) === count($foundFields)) {
	continue;
	}
	// If not all requested fields were extracted from the collection, start more complex extraction
	$categories[$item->getId()] = $this->categoryHydrator->hydrateCategory($item);

Below is a rough attempt at using a similar method in the Category Tree Extractor:

diff --git a/app/code/Magento/CatalogGraphQl/Model/Resolver/CategoryTree.php b/app/code/Magento/CatalogGraphQl/Model/Resolver/CategoryTree.php
index 4e3a8403f31..976235c41d6 100644
--- a/app/code/Magento/CatalogGraphQl/Model/Resolver/CategoryTree.php
+++ b/app/code/Magento/CatalogGraphQl/Model/Resolver/CategoryTree.php
@@ -73,7 +73,7 @@ class CategoryTree implements ResolverInterface
         $rootCategoryId = $this->getCategoryId($args);
         $categoriesTree = $this->categoryTree->getTree($info, $rootCategoryId);
         if (!empty($categoriesTree)) {
-            $result = $this->extractDataFromCategoryTree->execute($categoriesTree);
+            $result = $this->extractDataFromCategoryTree->execute($categoriesTree, $info);
             return current($result);
         } else {
             return null;
diff --git a/app/code/Magento/CatalogGraphQl/Model/Resolver/Products/DataProvider/ExtractDataFromCategoryTree.php b/app/code/Magento/CatalogGraphQl/Model/Resolver/Products/DataProvid
er/ExtractDataFromCategoryTree.php
index ac8d5709c85..39d382e63c7 100644
--- a/app/code/Magento/CatalogGraphQl/Model/Resolver/Products/DataProvider/ExtractDataFromCategoryTree.php
+++ b/app/code/Magento/CatalogGraphQl/Model/Resolver/Products/DataProvider/ExtractDataFromCategoryTree.php
@@ -9,6 +9,8 @@ namespace Magento\CatalogGraphQl\Model\Resolver\Products\DataProvider;
 
 use Magento\CatalogGraphQl\Model\Category\Hydrator;
 use Magento\Catalog\Api\Data\CategoryInterface;
+use Magento\CatalogGraphQl\Model\AttributesJoiner;
+use Magento\Framework\GraphQl\Schema\Type\ResolveInfo;
 
 /**
  * Extract data from category tree
@@ -20,22 +22,31 @@ class ExtractDataFromCategoryTree
      */
     private $categoryHydrator;
 
+    /**
+     * @var AttributesJoiner
+     */
+    private $attributesJoiner;
+
     /**
      * @param Hydrator $categoryHydrator
+     * @param AttributesJoiner $attributesJoiner
      */
     public function __construct(
-        Hydrator $categoryHydrator
+        Hydrator $categoryHydrator,
+        AttributesJoiner $attributesJoiner
     ) {
         $this->categoryHydrator = $categoryHydrator;
+        $this->attributesJoiner = $attributesJoiner;
     }
 
     /**
      * Extract data from category tree
      *
      * @param \Iterator $iterator
+     * @param ResolveInfo
      * @return array
      */
-    public function execute(\Iterator $iterator): array
+    public function execute(\Iterator $iterator, ResolveInfo $info): array
     {
         $tree = [];
         while ($iterator->valid()) {
@@ -43,10 +54,17 @@ class ExtractDataFromCategoryTree
             $category = $iterator->current();
             $iterator->next();
             $nextCategory = $iterator->current();
-            $tree[$category->getId()] = $this->categoryHydrator->hydrateCategory($category);
+            $tree[$category->getId()] = $this->categoryHydrator->hydrateCategory($category, true);
             $tree[$category->getId()]['model'] = $category;
             if ($nextCategory && (int) $nextCategory->getLevel() !== (int) $category->getLevel()) {
-                $tree[$category->getId()]['children'] = $this->execute($iterator);
+                $tree[$category->getId()]['children'] = $this->execute($iterator, $info);
+            }
+
+            $requestedFields = $this->attributesJoiner->getQueryFields($info->fieldNodes[0]);
+            $extractedFields = array_keys($tree[$category->getId()]);
+            $foundFields = array_intersect($requestedFields, $extractedFields);
+            if (count($requestedFields) !== count($foundFields)) {
+                $tree[$category->getId()] = $this->categoryHydrator->hydrateCategory($category, false);
             }
         }

[naydav]

Hello @pmclain,
Could you create PR? I will run it on internal performance builds.
Thanks

[naydav]

Closed due to #567