Translate this monsterous query w/HABTM related data into a CakePHP find - mysql

Question

It's late and I have written this monstrosity of a query to get related products based on a product I have already found.
I need to fetch the products in the same category (HABTM), the parent product, products with the same parent (siblings/neighbours), and products that are direct children of the current product (there is only one level of nesting). I have the product ID and its parent_id of the current product. If it could be possible to put conditions on the product as for Product.published = 1 that would be great, but if it's going to make the query so big I can always check that after. Additionally, I need to exclude the current product.
SELECT `products`.*
FROM `products`, `categories_products`
WHERE
(
(
`categories_products`.`product_id` = `products`.`id`
AND `categories_products`.`category_id` IN (
SELECT `category_id`
FROM `categories_products`
WHERE `categories_products`.`product_id` = '$product_id'
)
)
OR `products`.`parent_id` = '$parent_id'
OR `products`.`parent_id` = '$product_id'
OR `products`.`id` = '$parent_id'
)
AND `product`.`id` <> '$product_id'
GROUP BY `products`.`id`
It might even be possible to optimize it a bit more, so far I have:
public function related($productData, $limit = 4) {
$conditions = array(
'OR' => array(array('Product.parent_id' => $productData['Product']['id'])), // Children of product),
'Product.id <>' => $productData['Product']['id']
);
if(!empty($product['parent_id'])) {
$conditions['OR'][] = array('Product.parent_id' => $productData['Product']['parent_id']); // Siblings
$conditions['OR'][] = array('Product.id' => $productData['Product']['parent_id']); // Parent of product
}
return $this->find('all', array(
'conditions' => $conditions,
'contain' => array('Category'),
'group' => 'Product.id',
'limit' => $limit
));
}

Answer 1

You will need to use cake's Complex Find Conditions syntax (scroll down to Sub-queries secrion).

Answer 2

To be honest I didn't think its the best approach to get "related" products from mysql, thats what search engines are for. especially your similar category approach is would doom to fail when dealing with big data.
After saying this much this is a a re-write of your current sql which would hopefully have you gain some performance.
SELECT * FROM products p
WHERE
p.published = 1 AND
p.id != $product_id AND
(
p.id IN
(
SELECT DISTINCT(cp2.product_id) FROM categories_product cp1
LEFT JOIN categories_product cp2 ON cp1.category_id = cp2.category_id
WHERE cp1.product_id = $product_id
UNION SELECT $parent_id
)
OR p.parent_id IN($parent_id, $product_id)
)
;
I tried to get rid of unnecessary group by statement. Hope this helps.
P.S : There can be syntax errors since I wrote this in text editor.

Answer 3

I would prevent the need of the recursive attempt to SELECT IN on the category. Pre build that based on the one product in question and get all its distinct categories. From that, get distinct products that match the category. Now, you have a prequery of "CommonByCategory" that will ALREADY be a single instance of IDs.
Next, do a hard join to products again "OriginalProduct" based on the SPECIFIC ID you are trying to qualify against. Since it will always exist and never change, we can use this as the pointer for the siblings to compare against, and also for a parent ID match (in case not null -- via the IFNULL() tests applied
Since each product will only be scanned ONCE and not return multiple entries due to the multiple category possibilities, no "GROUP BY" is required.
SELECT STRAIGHT_JOIN
p.*
from
products p
left join
( SELECT DISTINCT
cp2.product_id
from
( SELECT cp.Category_ID
from categories_products cp
where cp.product_id = '$product_id' ) JustCats
join categories_products cp2
ON JustCats.Category_ID = cp2.Category_ID ) as CommonByCategory
ON p.ID = CommonByCategory.product_ID
join products OriginalProduct
ON OriginalProduct.ID = '$product_id'
where
p.id <> '$product_id'
and ( IFNULL( CommonByCategory.Product_ID, -1) > 0
OR p.id = IFNULL( OriginalProduct.Parent_ID, -1 )
OR p.parent_id = OriginalProduct.id

Answer 4

This issue came round again, and I figured it out 100%! Here was my finished code:
// Model
public function related($product, $limit = 9) {
// Children of product
$conditions = array(
'OR' => array(array('Product.parent_id' => $product['Product']['id'])), // Children of product),
'Product.id <>' => $product['Product']['id'],
'Product.published' => 1
);
// Siblings and parent of product if applicable
if (!empty($product['Product']['parent_id'])) {
$conditions['OR'][] = array('Product.parent_id' => $product['Product']['parent_id']);
$conditions['OR'][] = array('Product.id' => $product['Product']['parent_id']);
}
// Products in the same categories
// Get category IDs in an array
$categoryIds = Set::extract($product['Category'], '{n}.id');
$conditionsSubQuery['category_id IN(?)'] = implode(',', $categoryIds);
$db = $this->getDataSource();
$subQuery = $db->buildStatement(
array(
'fields' => array('product_id'),
'table' => 'categories_products',
'joins' => array(),
'alias' => 'c_p',
'conditions' => $conditionsSubQuery,
'order' => null,
'group' => null,
'limit' => null
), $this->CategoryProduct
);
$subQuery = 'Product.id IN (' . $subQuery . ') ';
$subQueryExpression = $db->expression($subQuery);
$conditions['OR'][] = $subQueryExpression;
return $this->find('all', array(
'conditions' => $conditions,
'contain' => array('Category'),
'group' => 'Product.id',
'limit' => $limit
));