$sampleRate (aggregation)¶

On this page

Definition
Behavior
Examples

Definition¶

$sampleRate¶

New in version 4.4.2.

Matches a random selection of input documents. The number of documents selected approximates the sample rate expressed as a percentage of the total number of documents.

The $sampleRate operator has the following syntax:

copy

{ $sampleRate: <non-negative float> }

Behavior¶

The selection process uses a uniform random distribution. The sample rate is a floating point number between 0 and 1, inclusive, which represents the probability that a given document will be selected as it passes through the pipeline.

For example, a sample rate of 0.33 selects roughly one document in three.

This expression:

copy

{ $match: { $sampleRate: 0.33 } }

is equivalent to using the $rand operator as follows:

copy

{ $match: { $expr: { $lt: [ { $rand: {} }, 0.33 ] } } }

Repeated runs on the same data will produce different outcomes since the selection process is non-deterministic. In general, smaller datasets will show more variability in the number of documents selected on each run. As collection size increases, the number of documents chosen will approach the expected value for a uniform random distribution.

Note

If an exact number of documents is required from each run, the $sample operator should be used instead of $sampleRate.

Examples¶

This code creates a small collection with 100 documents.

copy

N = 100
bulk = db.collection.initializeUnorderedBulkOp()
for ( i = 0; i < N; i++) { bulk.insert( {_id: i, r: 0} ) }
bulk.execute()

The $sampleRate operator can be used in a pipeline to select random documents from the collection. In this example we use $sampleRate to select about one third of the documents.

copy

db.collection.aggregate(
   [
     { $match: { $sampleRate: 0.33 } },
     { $count: "numMatches" }
   ]
)

This is the output from 5 runs on the sample collection:

{ "numMatches" : 38 }
{ "numMatches" : 36 }
{ "numMatches" : 29 }
{ "numMatches" : 29 }
{ "numMatches" : 28 }