Searching for keywords in a document is a common operation for many applications. If this is a core operation, it makes sense to use a specialized store for search, such as Elasticsearch; however MongoDB can be used efficiently until scale dictates moving to a different solution.

The basic need for keyword search is to be able to search the entire document for keywords. For example, with a document in the products collection:

{ name : "Macbook Pro late 2016 15in" ,
  manufacturer : "Apple" ,
  price: 2000 ,
  keywords : [ "Macbook Pro late 2016 15in", "2000", "Apple", "macbook", "laptop", "computer" ]
 }

We can create a multi-key index in the keywords field:

> db.products.createIndex( { keywords: 1 } )

Now we can search in the keywords field for any name, manufacturer, price fields, and also any of the custom keywords we set up. This is not an efficient or flexible approach as we need to keep keywords lists in sync, can't use stemming, and can't rank results (it's more like filtering than searching) with the only upside being implementation time.

Since version 2.4 , MongoDB has had a special text index type. This can be declared in one or multiple fields and supports stemming, tokenization, exact phrase (" "), negation (-), and weighting results.

Index declaration on three fields with custom weights:

db.products.createIndex({
     name: "text",
     manufacturer: "text",
    price: "text"
   },
  {
     weights: { name: 10,
       manufacturer: 5,
      price: 1 },
     name: "ProductIndex"
   })

In this example, name is 10 times more important that price but only two from a manufacturer.

A text index can also be declared with a wildcard, matching all fields that match the pattern:

db.collection.createIndex( { "$**": "text" } )

This can be useful when we have unstructured data and we may not know all the fields that they will come with. We can drop the index by name just like with any other index.

The greatest advantage though, other than all the features, is that all record keeping is done by the database.