Docs Menu

Docs HomeMongoDB Atlas

Update $text Queries with Atlas Search for Improved Search Performance

On this page

  • Atlas Search Feature Advantages
  • Examples
  • Learn More

If your queries rely heavily on the $text aggregation pipeline stage, you can modify these queries to use $search instead to improve both the flexibility and performance of these queries.

The $search aggregation stage provides the following features which are either not available through the $text operator, available but less performant, or available only with significant implementation work by the user:

The examples in the following sections use queries against the sample_mflix.movies collection in the sample data to illustrate the improvements to flexibility and performance that Atlas Search offers over $text. You can run the queries from both examples using the following indexes:

Text Index
Atlas Search Index
db.movies.createIndex(
{
genres: "text",
plot: "text",
year: -1
}
)
{
"mappings": {
"dynamic": false,
"fields": {
"genres": {
"type": "string"
},
"plot": {
"type": "string"
},
"year": {
"type": "number"
}
}
}
}

Either index definition will index the genres and plot fields as text, and the year field as numeric. For instructions on creating $text indexes, see Create Text Index. For instructions on creating Atlas Search indexes, see Create an Atlas Search Index.

You can update your $text-based queries to use $search for greater flexibility and convenience. In this example, you will query the sample_mflix.movies collection in the sample data to retrieve entries with the word 'poet' in the plot field, sorted in ascending order by year.

The index definitions laid out in the previous section illustrate one of the flexibility enhancements of $search: in order to create the $text index on sample_mflix.movies, you must first delete any existing text index on the sample data, as MongoDB supports only a single text index per collection.

In contrast, you can create multiple Atlas Search indexes for a single collection, allowing your applications to leverage distinct full text queries in parallel.

The following queries return the five most recent movies with 'poet' in the plot field, showing their titles, genres, plots, and years of release.

Regex Index
Atlas Search Index
db.movies.find(
{
$text: { $search: "poet" }
},
{
_id: 0,
title: 1,
genres: 1,
plot: 1,
year: 1
}
).limit(5)
db.movies.aggregate([
{
"$search": {
"text": {
"path": "plot",
"query": "poet"
}
}
},
{
"$limit": 5
},
{
"$project": {
"_id": 0,
"title": 1,
"genres": 1,
"plot": 1,
"year": 1,
}
}
])

Both of these queries return the following results:

{
plot: `It's the story of the murder of a poet, a man, a great film director: Pier Paolo Pasolini. The story begin with the arrest of "Pelosi", a young man then accused of the murder of the poet. ...`,
genres: [ 'Crime', 'Drama' ],
title: 'Who Killed Pasolini?',
year: 1995
},
{
plot: 'Friendship and betrayal between two poets during the French Revolution.',
genres: [ 'Biography', 'Drama' ],
title: 'Pandaemonium',
year: 2000
},
{
year: 2003,
plot: 'Story of the relationship between the poets Ted Hughes and Sylvia Plath.',
genres: [ 'Biography', 'Drama', 'Romance' ],
title: 'Sylvia'
},
{
year: 2003,
plot: 'Story of the relationship between the poets Ted Hughes and Sylvia Plath.',
genres: [ 'Biography', 'Drama', 'Romance' ],
title: 'Sylvia'
},
{
plot: 'A love-struck Italian poet is stuck in Iraq at the onset of an American invasion.',
genres: [ 'Comedy', 'Drama', 'Romance' ],
title: 'The Tiger and the Snow',
year: 2005
}

Unique to Atlas Search, you can add highlights to the results, displaying matches in the contexts in which they were found. To do so, replace the Atlas Search query above with the following:

1db.movies.aggregate([
2 {
3 "$search": {
4 "text": {
5 "path": "plot",
6 "query": "poet"
7 },
8 "highlight": {
9 "path": "plot"
10 }
11 }
12 },
13 {
14 "$limit": 1
15 },
16 {
17 "$project": {
18 "_id": 0,
19 "title": 1,
20 "genres": 1,
21 "plot": 1,
22 "year": 1,
23 "highlights": { "$meta": "searchHighlights" }
24 }
25 }
26])

The results of the above query include the highlights field, which contains both the context in which all matches occurred, and relevance scores for each. For example, the following shows the highlights field for the first document in the $search results.

{
plot: `It's the story of the murder of a poet, a man, a great film director: Pier Paolo Pasolini. The story begin with the arrest of "Pelosi", a young man then accused of the murder of the poet. ...`,
genres: [ 'Crime', 'Drama' ],
title: 'Who Killed Pasolini?',
year: 1995,
highlights: [
{
score: 1.0902210474014282,
path: 'plot',
texts: [
{ value: "It's the story of the murder of a ", type: 'text' },
{ value: 'poet', type: 'hit' },
{
value: ', a man, a great film director: Pier Paolo Pasolini. ',
type: 'text'
}
]
},
{
score: 1.0202842950820923,
path: 'plot',
texts: [
{
value: 'The story begin with the arrest of "Pelosi", a young man then accused of the murder of the ',
type: 'text'
},
{ value: 'poet', type: 'hit' },
{ value: '. ...', type: 'text' }
]
}
]
}

In addition to greater flexibility and convenience, Atlas Search provides significant performance advantages over analogous $text queries. Consider a query against the sample_mflix.movies collection to retrieve movies released between 2000 and 2010, in the comedy genre, with 'poet' in the plot field.

Run the following queries:

Text Index
Atlas Search Index
db.movies.aggregate([
{
$match: {
year: {$gte: 2000, $lte: 2010},
$text: { $search: "poet" },
genres : { $eq: "Comedy" }
}
},
{ "$sort": { "year": 1 } },
{
"$limit": 3
},
{
"$project": {
"_id": 0,
"title": 1,
"genres": 1,
"plot": 1,
"year": 1
},
}
])
db.movies.aggregate([
{
"$search": {
"compound": {
"filter": [{
"range": {
"gte": 2000,
"lte": 2010,
"path": "year"
}
},
{
"text": {
"path": "plot",
"query": "poet"
}
},
{
"text": {
"path": "genres",
"query": "comedy"
}
}]
}
}
},
{ "$sort": { "year": 1 } },
{
"$limit": 3
},
{
"$project": {
"_id": 0,
"title": 1,
"genres": 1,
"plot": 1,
"year": 1
}
}
])

Both of these queries will return the following three documents.

{
year: 2000,
plot: 'A film poem inspired by the Peruvian poet Cèsar Vallejo. A story about our need for love, our confusion, greatness and smallness and, most of all, our vulnerability. It is a story with many...',
genres: [ 'Comedy', 'Drama' ],
title: 'Songs from the Second Floor'
},
{
plot: 'When his mother, who has sheltered him his entire 40 years, dies, Elling, a sensitive, would-be poet, is sent to live in a state institution. There he meets Kjell Bjarne, a gentle giant and...',
genres: [ 'Comedy', 'Drama' ],
title: 'Elling',
year: 2001
},
{
plot: 'Heart-broken after several affairs, a woman finds herself torn between a Poet and a TV Host.',
genres: [ 'Comedy', 'Romance', 'Drama' ],
title: 'Easy',
year: 2003
}

Although $text is adequate for simple, narrow searches such as this, as the size of the datasets and breadth of your queries increases, the performance advantages of $search will significantly improve the responsiveness of your applications. We recommend that you use an Atlas Search query through the $search aggregation pipeline stage.

←  Monitor Real-Time PerformanceImprove Your Schema →
Share Feedback
© 2023 MongoDB, Inc.

About

  • Careers
  • Investor Relations
  • Legal Notices
  • Privacy Notices
  • Security Information
  • Trust Center
© 2023 MongoDB, Inc.