Query a document and all of its subdocuments that

问题:

I have a MongoDB storing data from different sensors. It has the following structure:

 {
     "_id" : 1,
     "sensorName" : "Heart Rate",
     "samplePeriod" : 1000,
     "data" : [
             {
                 "timestamp" : NumberLong("1483537204046"),
                 "dataPoints" : [ 68 70 ]
             },
             {
                 "timestamp" : NumberLong("1483537206046"),
                 "dataPoints" : [ 68 70 ]
             }
     ]
}
{
    "_id" : 2,
    "sensorName" : "Ambient Light",
    "samplePeriod" : 500,
    "data" : [
            {
                "timestamp" : NumberLong("1483537204058"),
                "dataPoints" : [ 56, 54, 54, 54 ]
            },
            {
                "timestamp" : NumberLong("1483537206058"),
                "dataPoints" : [ 56, 54, 54, 54 ]
            }
    ]
}

Now for example i need the "Heart Rate" - document with all of its fields and those of its "data" - subdocuments matching the condition "timestamp between 1483537204000 and 1483537214000".

I already got the answer on how to do this in the mongo shell in another Question. See this code:

aggregate([{
    $match: {
        "_id": 1
    }
}, {
    "$project": {
        "_id": 1,
        "sensorName": 1,
        "samplePeriod": 1,
        "data": {
            "$filter": {
                "input": "$data",
                "as": "result",
                "cond": {
                    $and: [{
                        $gte: ["$$result.timestamp", 1483537204000]
                    }, {
                        $lte: ["$$result.timestamp", 1483537214000]
                    }]
                }
            }
        }
    }
}])

But how do I do this in java spring-data? It seems there is nothing like $filter in spring-data. Is there a workaround?

How efficient is $filter anyway? Can you think of a more efficient/practical way of structuring this kind of data in mongodb?

Thanks in advance!

回答1:

You'll need to make use of MongoTemplate provided in the spring mongo data dependency. There is no out of box support for $filter in the current release version. Make use of AggressionExpression. Include below projection in project. Use 1.8.5 spring mongo data version.

Aggregation aggregation = newAggregation(
        match(Criteria.where("_id").is(1)),
        project( "_id", "sensorName", "samplePeriod").and(new AggregationExpression() {
            @Override
            public DBObject toDbObject(AggregationOperationContext aggregationOperationContext) {
                DBObject filter = new BasicDBObject("input", "$data").append("as", "result").append("cond",
                        new BasicDBObject("$and", Arrays.<Object> asList(new BasicDBObject("$gte", Arrays.<Object> asList("$$result.timestamp", 1483537204000L)),
                                new BasicDBObject("$lte", Arrays.<Object> asList("$$result.timestamp", 1483537214000L)))));
                return new BasicDBObject("$filter", filter);
            }
        }).as("data")
);

List<BasicDBObject> dbObjects = monoTemplate.aggregate(aggregation, "collectionname", BasicDBObject.class).getMappedResults();

回答2:

I think the same can be achieved by the use of unwind and an extra match. Spring mongo driver does provide support for unwind and it looks bit cleaner.

aggregate([{
 $match: {
    "_id": 1
   }
 }, {
  $unwind : "$data"
 },{
   $match : {'data.timestamp' : {$gte : 1483537204000, $lte : 1483537214000}}
 }, {
  $group : {
      _id : $_id,
      data : {$push:$data}
  }
 }])

回答3:

The Spring Data MongoDB 1.10 RC1 (Ingalls), 2.0 M2 (Kay) releases (as of writing) have added support for $filter and this can be implemented as follows:

Aggregation.newAggregation(Entity.class,
    match(Criteria.where("_id").is(1)),
    project("sensorName", "samplePeriod")
        .and(
            filter("data")
                .as("item")
                .by(
                    GTE.of(field("item.timestamp"), 1483537204000)
                    .LTE.of(field("item.timestamp"), 1483537214000)
                )
        ).as("data")
    )