A new requirement came into existence to search for products that have a specific attribute value set.

As in, the user should be able to search for a particular product based on any of its attributes. However, these attributes are not fixed and might change on a daily basis. Hence, it's not possible to keep this as a key-value pair. More keys mean more fields and each field costs Lucene a reverse index under the hood. So, a better approach would be to model the data as follows:

/products/product/LCD_TV

{
  "name" : "LCD TV",
  "tags" : [
    {
      "tagName" : "company" ,
      "value" : "Sony"
    },
    {
      "tagName" : "competitor" ,
      "value" : "Toshiba"
    }
  ]
}

With this approach, everything looks good, but then, the following query works true for the preceding document:

tags.tagName:company AND tags.value:Toshiba

This is not what we expected. The product actually belongs to the company Sony, but then, it matched against the company Toshiba...