Filtering product collection by exact SKU returns wrong results when SKU contains special characters

[diwipl]

\Smile\ElasticsuiteCatalog\Model\ResourceModel\Product\Fulltext\Collection is not filtering collection properly when SKU condition is used and passed SKUs contain special characters

This might be related to changes done in SKU searching described in wiki:
https://github.com/Smile-SA/elasticsuite/wiki/SearchingBySkuBasics
https://github.com/Smile-SA/elasticsuite/wiki/SearchingBySkuDetailled

Nevertheless filtering itself should not be affected and should return correct results when exact SKUs are provided.

Preconditions

• Magento 2.3.1 installed with ElasticSuite 2.7.6 and Magento sample data.

Magento Version : 2.3.1

ElasticSuite Version : 2.7.6

Environment : Developer

Steps to reproduce

ObjectManager is used directly only to simplify reproduction.

<?php
        $collection = \Magento\Framework\App\ObjectManager::getInstance()
            ->create( \Smile\ElasticsuiteCatalog\Model\ResourceModel\Product\Fulltext\Collection::class);

        $skus = ['24-MB01', '24-MB04'];
        
        $collection->addFieldToFilter('sku', ['in' => $skus]);
        $products = $collection->getItems();

Expected result

1. Two entities should be returned by getItems method

Actual result

1. Empty array is returned

[romainruaud]

Yes it's actually due to the fact that using in will result in a terms query : https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-term-query.html

Please pay attention to the "Why doesn’t the term query match my document?" part especially.

Since the sku field is analyzed during indexing, it's impossible to match it with a term query if it contains dashes.

From what I can see, there is an easy workaround : just set the sku field to isFilterable via the elasticsuite_indices.xml file, like this :

<indices xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
         xsi:noNamespaceSchemaLocation="urn:magento:module:Smile_ElasticsuiteCore:etc/elasticsuite_indices.xsd">

    <index identifier="catalog_product" defaultSearchType="product">
        <type name="product" idFieldName="entity_id">
            <mapping>
                <field name="sku" type="text">
                    <isFilterable>1</isFilterable>
                </field>
...

This will allow you to match with the in condition you described. (I tested and it's working, as it does generate a term query on the sku.untouched field, allowing exact matching).

Actually I'm not sure if we should add this behavior to the Core of ElasticSuite or if it's up to people to decide if they want to use this.

@rbayet any opinion ? I do not fully measure if indexing the untouched version of SKU would collisionate with your recent work on SKU search.

Regards

[rbayet]

Hi @romainruaud,

No basically, for correct exact value filtering, the sku should indeed be declared as filterable so its mapping inherits the untouched subfield.
I have no issue with that, the only impact would be the index size.

It will not affect what I described in https://github.com/Smile-SA/elasticsuite/wiki/SearchingBySkuBasics and https://github.com/Smile-SA/elasticsuite/wiki/SearchingBySkuDetailled, the untouched subfield is never a target in the context of a fulltext search for any field whatsoever atm, so it will not impact SKU search.

Regards,

[romainruaud]

Sounds good to me, so PR #1396 will fix this issue.

Regards