JIYIK CN >

Today, we will discuss fuzzy search and how to do it with MongoDB.

We will start by using $regexthe operator and $textthe query. In addition, we will learn to Fuse.jsperform fuzzy search on documents using a JavaScript library called .

What is fuzzy search

Using fuzzy search, we can search for text that does not match exactly but closely. This is useful to find relevant results even if the search term is misspelled.

For example, Google will show us various web pages related to our search term even if we type it wrong. Using regular expressions (also known as regex) is also a very beneficial and time-saving way to implement fuzzy search.

Create a sample collection in MongoDB

We will learn about fuzzy search from basic to advanced level. To practice it, let's create a collection_onesample collection called which has one field for each document, ie name.

_idis created automatically; we don't have to create it. You can use the following query to perform the same operation.

Sample code:

> db.createCollection('collection_one')
> db.collection_one.insertMany([
    { name : 'Mehvish Ashiq'},
    { name : 'Jennifer Johnson'},
    { name : 'Natalie Robinson'},
    { name : 'John Ferguson'},
    { name : 'Samuel Patterson'},
    { name : 'Salvatore Callahan'},
    { name : 'Mikaela Christensen'}
])
> db.collection_one.find()

Output:

{ "_id" : ObjectId("62939a37b3a0d806d251ddae"), "name" : "Mehvish Ashiq" }
{ "_id" : ObjectId("62939a37b3a0d806d251ddaf"), "name" : "Jennifer Johnson" }
{ "_id" : ObjectId("62939a37b3a0d806d251ddb0"), "name" : "Natalie Robinson" }
{ "_id" : ObjectId("62939a37b3a0d806d251ddb1"), "name" : "John Ferguson" }
{ "_id" : ObjectId("62939a37b3a0d806d251ddb2"), "name" : "Samuel Patterson" }
{ "_id" : ObjectId("62939a37b3a0d806d251ddb3"), "name" : "Salvatore Callahan" }
{ "_id" : ObjectId("62939a37b3a0d806d251ddb4"), "name" : "Mikaela Christensen" }

$regexPerforming fuzzy search in MongoDB using the operator

Sample code:

> db.collection_one.find({"name": /m/})

Output:

{ "_id" : ObjectId("62939a37b3a0d806d251ddb2"), "name" : "Samuel Patterson" }

In this code, we nameperform a fuzzy search on the field and retrieve all documents namewhere the field contains the letters .m

As you can see, we have only one mrecord containing the letter , but there are two more Mdocuments starting with (uppercase letter). To handle this, we can use i a modifier like the following , which performs a case-insensitive search.

Sample code:

> db.collection_one.find({"name": /m/i})

Output:

{ "_id" : ObjectId("62939a37b3a0d806d251ddae"), "name" : "Mehvish Ashiq" }
{ "_id" : ObjectId("62939a37b3a0d806d251ddb2"), "name" : "Samuel Patterson" }
{ "_id" : ObjectId("62939a37b3a0d806d251ddb4"), "name" : "Mikaela Christensen" }

It shows that properly designed regular expressions are very important; otherwise, we may get misleading results. We can also do the same thing in the following way.

Sample code (case insensitive search):

> db.collection_one.find({'name': {'$regex': 'm','$options': 'i'}})

Output:

{ "_id" : ObjectId("62939a37b3a0d806d251ddae"), "name" : "Mehvish Ashiq" }
{ "_id" : ObjectId("62939a37b3a0d806d251ddb2"), "name" : "Samuel Patterson" }
{ "_id" : ObjectId("62939a37b3a0d806d251ddb4"), "name" : "Mikaela Christensen" }

Similarly, we can get all namedocuments ending with a two-letter combination as on.

Sample code:

> db.collection_one.find({name:{'$regex' : 'on$', '$options' : 'i'}})

Output:

{ "_id" : ObjectId("62939a37b3a0d806d251ddaf"), "name" : "Jennifer Johnson" }
{ "_id" : ObjectId("62939a37b3a0d806d251ddb0"), "name" : "Natalie Robinson" }
{ "_id" : ObjectId("62939a37b3a0d806d251ddb1"), "name" : "John Ferguson" }
{ "_id" : ObjectId("62939a37b3a0d806d251ddb2"), "name" : "Samuel Patterson" }

$textPerforming fuzzy searches in MongoDB using query

$textThe query does not apply to our collection_oneexample collection named because it does not have a text index. Therefore, we create an index as follows.

Sample code:

> db.collection_one.createIndex({name:"text"});

If the specified collection does not exist, the above statement will also create it. Remember that we can create an index on one or more fields separated by commas.

See the following example.

db.collection_name.createIndex({name:"text", description:"text"});

After creating the index, we can do a fuzzy search as shown below.

Sample code:

> db.collection_one.find({ $text: { $search: "Mehvish" } } )

Output:

{ "_id" : ObjectId("62939a37b3a0d806d251ddae"), "name" : "Mehvish Ashiq" }

Fuse.jsPerform fuzzy search in MongoDB using JavaScript 's

Sample code ( fuzzysearch.jsfile code):

const Fuse = require('fuse.js')
var MongoClient = require('mongodb').MongoClient;
var url = "mongodb://localhost:27017/";

MongoClient.connect(url, function(err, db) {
    if (err) throw err;
    var dbo = db.db("FuseFuzzySearch");

    var personObj = [
        { name : 'Mehvish Ashiq'},
        { name : 'Jennifer Johnson'},
        { name : 'Natalie Robinson'},
        { name : 'John Ferguson'},
        { name : 'Samuel Patterson'},
        { name : 'Salvatore Callahan'},
        { name : 'Mikaela Christensen'}
    ];

    dbo.collection("person").insertMany(personObj, function(err, res) {
        if (err) throw err;
    });

    const options = {
        includeScore: true,
        keys: ['name']
    }

    const fuse = new Fuse(personObj, options);
    const result = fuse.search('jahson');
    console.log(result);
    db.close();
});

Output:

[
  {
    item: { name: 'Jennifer Johnson', _id: 6293aa0340aa3b21483d9885 },
    refIndex: 1,
    score: 0.5445835311565898
  },
  {
    item: { name: 'John Ferguson', _id: 6293aa0340aa3b21483d9887 },
    refIndex: 3,
    score: 0.612592665952338
  },
  {
    item: { name: 'Natalie Robinson', _id: 6293aa0340aa3b21483d9886 },
    refIndex: 2,
    score: 0.6968718698752637
  },
  {
    item: { name: 'Samuel Patterson', _id: 6293aa0340aa3b21483d9888 },
    refIndex: 4,
    score: 0.6968718698752637
  }
]

In this code example, we first imported fuse.jsthe library. Next, we connected to MongoDB.

If not connected for any reason, throw an error. Otherwise, create a FuseFussySearchdatabase named .

We then create an personObjobject called that contains personall the documents we want to insert into the collection. If there are any problems inserting the data, errors will be generated.

Create Fusean object of , pass an array of objects keysand with and to perform a fuzzy search and get the results, as shown above.includeScorepersonObjoptions

Here, keysspecify the field on which the search will be performed. includeScoreis optional, but it is better to have it because it tells the match score.

If it is 0, the program found a perfect match, while 1a score of indicates a complete mismatch. You can find all the options here.

Finally, don't forget to close the connection.