Database Manual / Reference / Query Language / Expressions

$sampleRate (aggregation)(聚合)

Definition定义

$sampleRate

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.匹配随机选择的输入文档。所选文档的数量近似于采样率,表示为文档总数的百分比。

The $sampleRate operator has the following syntax:$sampleRate运算符具有以下语法:

{ $sampleRate: <non-negative float> }

Behavior行为

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.选择过程使用均匀随机分布。采样率是一个介于0和1之间的浮点数,表示给定文档在通过管道时被选中的概率。

For example, a sample rate of 0.33 selects roughly one document in three.例如,0.33的采样率大约选择三分之一的文档。

This expression:此表达式:

{ $match: { $sampleRate: 0.33 } }

is equivalent to using the $rand operator as follows:相当于使用$rand运算符,如下所示:

{ $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.由于选择过程是非确定性的,对相同数据的重复运行将产生不同的结果。一般来说,较小的数据集将显示每次运行时所选文档数量的更多可变性。随着集合大小的增加,所选文档的数量将接近均匀随机分布的预期值。

Note

If an exact number of documents is required from each run, the $sample operator should be used instead of $sampleRate.如果每次运行都需要精确数量的文档,则应使用$sample运算符而不是$sampleRate

Examples示例

This code creates a small collection with 100 documents.这段代码创建了一个包含100个文档的小集合。

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 $sampleRate to select about one third of the documents.$sampleRate运算符可以在管道中用于从集合中选择随机文档。在这个例子中,我们使用$sampleRate来选择大约三分之一的文档。

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

This is the output from 5 runs on the sample collection:这是样本采集上5次运行的输出:

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