CatalogSearch - exclude child product names from fulltext

Question

I've been playing with the catalogsearch for a while, and I'm almost done, but there is one odd thing happening to me. The only "usable in search" attributes are name, child_skus and synonyme. In child_skus, I'm storing the skus of the associated products of my grouped products, as I'm working with grouped products only on the frontend. synonyme contains a list of coma-separated words which related to the said product.

What I'm using for the fulltext search is the following code :

$query = Mage::getModel('catalogsearch/query')->setQueryText($searchText)->prepare();
$fulltextResource = Mage::getResourceModel('catalogsearch/fulltext')->prepareResult(
            Mage::getModel('catalogsearch/fulltext'), 
            $searchText, 
            $query
        );
$products->getSelect()->joinInner(
            array('search_result' => $products->getTable('catalogsearch/result')),
            $products->getConnection()->quoteInto(
                'search_result.product_id=e.entity_id AND search_result.query_id=?',
                $query->getId()
            ),
            array('relevance' => 'relevance')
        );
$products->getSelect()->order('relevance DESC');  

All works fine for now, except that somehow, it also indexes the associated product's name in the fulltext table, which ends up jamming my search because it yields biased results.

Does anyone have any idea on how to change that behaviour, or control it through admin?

Answer 1

Take a look at the method Mage_CatalogSearch_Model_Resource_Fulltext::_rebuildStoreIndex. That's the one that actually rebuilds the fultext search index.
Somewhere inside it you will find this:

$productChildren = $this->_getProductChildIds($productData['entity_id'], $productData['type_id']);
$productRelations[$productData['entity_id']] = $productChildren;

and later

if ($productChildren = $productRelations[$productData['entity_id']]) {
   foreach ($productChildren as $productChildId) {
       if (isset($productAttributes[$productChildId])) {
           $productIndex[$productChildId] = $productAttributes[$productChildId];
       }
   }
}

This means that for configurable, grouped and bundle products the child product attributes are indexed also.
You can rewrite the model and remove this parts, or leave it like it is and simply rewrite the method _getProductChildIds in the same class and make it return null so it will be like there are no child products for any other product.
this way the child products will not be indexed in the parent product data.

Answer 2

You can already set the synonyms in the back-office with "search term" in the catalog menu. To do that the search term as to have been already used in front-office and then you can redirect to the product, category or cms page you like. For products, you can set visibility accordingly to what you want to display in search results so you don't really need to bother about the child_skus.

I would suggest you to install elasticsearch or solr with an appropriate module to really improve search results. Moreover, you can really manage better with fuzzy queries, wild cards and so on.