mongo $ sum在进行$ unwind时复合,然后在多个字段上进行$ group

时间:2014-08-08 10:15:53

标签: mongodb mongodb-query aggregation-framework

我有以下文档结构

{
    "app_id": "DHJFK67JDSJjdasj909",
    "date": ISODate("2014-08-07T00:00:00.000Z"),
    "event_count": 100,
    "events": [
        { "type": 0,  "value": 12  },
        { "type": 10, "value": 24 },
        { "type": 20, "value": 36  },
        { "type": 30, "value": 43 }
    ],
    "unique_events": [
        { "type": 0,  "value": 5  },
        { "type": 10, "value": 8 },
        { "type": 20, "value": 12  },
        { "type": 30, "value": 56 }
    ]
}

我正在尝试获取event_counts的总和以及每种类型的unique_events和事件的值。这是我期望的输出类型,其中event_count以及每个事件和unique_events值已按类型求和。

{
    "app_id": "DHJFK67JDSJjdasj909",
    "date": ISODate("2014-08-07T00:00:00.000Z"),
    "event_count": 4345,
    "events": [
        { "type": 0,  "value": 624  },
        { "type": 10, "value": 234 },
        { "type": 20, "value": 353 },
        { "type": 30, "value": 472 }
    ],
    "unique_events": [
        { "type": 0,  "value": 433  },
        { "type": 10, "value": 554 },
        { "type": 20, "value": 645  },
        { "type": 30, "value": 732 }
    ]
}

这是我的查询

db.events.aggregate([
    { "$unwind": "$events" },
    { "$group": {
        "_id": { 
            "app_id": "$app_id",
            "type": "$events.type"
            "unique_type": "$unique_events.type"
        },
        "event_count": { "$sum": "$event_count" },
        "event_value": { "$sum": "$events.value" },
        "unique_event_value": { "$sum": "$unique_events.value" }
    }},
    { "$group": {
        "_id": "$_id.app_id",
        "event_count": { "$sum": "$event_count" },
        "events": { "$push": { "type": "$_id.type", "value": "$event_value" } }
        "unique_events": { "$push": { "type": "$_id.unique_type", "value": "$unique_event_value" } }
    }}
]) 

问题是使用两个$ unwinds然后按事件和unique_events进行分组会导致$ sum被复合并且太大。有什么方法可以使用mongo解决这个问题,或者我必须运行两个查询然后在代码中合并两个结果集。

由于

伊尔凡

2 个答案:

答案 0 :(得分:8)

这真的很简单,为了对每个数组的结果求和,它只是在辨别哪个是哪个和"组合元素"的问题。简而言之,你可能应该在你的文档中这样做,因为从第一个流水线阶段可以看出这一点。

因此,对于MongoDB 2.6及更高版本,有一些辅助方法:

db.events.aggregate([
    { "$project": {
        "app_id": 1,
        "event_count": 1,
        "all_events": {
            "$setUnion": [
                { "$map": {
                    "input": "$events",
                    "as": "el",
                    "in": {
                        "type": "$$el.type",
                        "value": "$$el.value",
                        "class": { "$literal": "A" }
                    }
                }},
                { "$map": {
                    "input": "$unique_events",
                    "as": "el",
                    "in": {
                        "type": "$$el.type",
                        "value": "$$el.value",
                        "class": { "$literal": "B" }
                    }
                }}
            ]
        }
    }},
    { "$unwind": "$all_events" },
    { "$group": {
        "_id": {
            "app_id": "$app_id",
            "class": "$all_events.class",
            "type": "$all_events.type"
        },
        "event_count": { "$sum": "$event_count" },
        "value": { "$sum": "$all_events.value" }
    }},
    { "$group": {
        "_id": "$_id.app_id",
        "event_count": { "$sum": "$event_count" },
        "events": {
            "$push": {
                "$cond": [
                    { "$eq": [ "$_id.class", "A" ] },
                    { "type": "$_id.type", "value": "$value" },
                    false
                ]
            }
        },
        "unique_events": {
            "$push": {
                "$cond": [
                    { "$eq": [ "$_id.class", "B" ] },
                    { "type": "$_id.type", "value": "$value" },
                    false
                ]
            }
        }
    }},
    { "$project": {
        "event_count": 1,
        "events": { "$setDifference": [ "$events", [false] ] },
        "unique_events": {
            "$setDifference": [ "$unique_events", [false] ]
        }
    }}
])

主要位于$setUnion$setDifference运营商中。另一个ccase是$map,它处理数组到位。整个过程就是在不使用$unwind的情况下对数组进行操作。但这些当然可以在以前的版本中完成,只需要做一些工作:

db.events.aggregate([
    { "$unwind": "$events" },
    { "$group": {
        "_id": "$_id",
        "app_id": { "$first": "$app_id" },
        "event_count": { "$first": "$event_count" },
        "events": {
            "$push": {
                "type": "$events.type",
                "value": "$events.value",
                "class": { "$const": "A" }
            }
        },
        "unique_events": { "$first": "$unique_events" }            
    }},
    { "$unwind": "$unique_events" },
    { "$group": {
        "_id": "$_id",
        "app_id": { "$first": "$app_id" },
        "event_count": { "$first": "$event_count" },
        "events": { "$first": "$events" },
        "unique_events": {
            "$push": {
                "type": "$unique_events.type",
                "value": "$unique_events.value",
                "class": { "$const": "B" }
            }
        }
    }},
    { "$project": {
        "app_id": 1,
        "event_count": 1,
        "events": 1,
        "unique_events": 1,
        "type": { "$const": [ "A","B" ] }
    }},
    { "$unwind": "$type" },
    { "$unwind": "$events" },
    { "$unwind": "$unique_events" },
    { "$group": {
        "_id": "$_id",
        "app_id": { "$first": "$app_id" },
        "event_count": { "$first": "$event_count" },
        "all_events": {
            "$addToSet": {
                "$cond": [
                     { "$eq": [ "$events.class", "$type" ] },
                     {
                         "type": "$events.type",
                         "value": "$events.value",
                         "class": "$events.class"
                     },
                     {
                         "type": "$unique_events.type",
                         "value": "$unique_events.value",
                         "class": "$unique_events.class"
                     }
                ]
            }
        }
    }},
    { "$unwind": "$all_events" },
   { "$group": {
        "_id": {
            "app_id": "$app_id",
            "class": "$all_events.class",
            "type": "$all_events.type"
        },
        "event_count": { "$sum": "$event_count" },
        "value": { "$sum": "$all_events.value" }
    }},
    { "$group": {
        "_id": "$_id.app_id",
        "event_count": { "$sum": "$event_count" },
        "events": {
            "$push": {
                "$cond": [
                    { "$eq": [ "$_id.class", "A" ] },
                    { "type": "$_id.type", "value": "$value" },
                    false
                ]
            }
        },
        "unique_events": {
            "$push": {
                "$cond": [
                    { "$eq": [ "$_id.class", "B" ] },
                    { "type": "$_id.type", "value": "$value" },
                    false
                ]
            }
        }
    }},
    { "$unwind": "$events" },
    { "$match": { "events": { "$ne": false } } },
    { "$group": {
        "_id": "$_id",
        "event_count": { "$first": "$event_count" },
        "events": { "$push": "$events" },
        "unique_events": { "$first": "$unique_events" }
    }},
    { "$unwind": "$unique_events" },
    { "$match": { "unique_events": { "$ne": false } } },
    { "$group": {
       "_id": "$_id",
        "event_count": { "$first": "$event_count" },
        "events": { "$first": "$events" },
        "unique_events": { "$push": "$unique_events" }
    }}
])

这可以获得你想要的结果,每个数组都是"总结"一起以及主人" event_count"结果正确。

您可能应该考虑将两个阵列与管道中使用的类似标识符进行组合,如图所示。这部分是工作的一半。另一半考虑您可能应该将预聚合结果存储在某个集合中,以获得最佳应用程序性能。

答案 1 :(得分:2)

您可以执行以下地图缩小:
它不是动态解决方案,我为每个variableevents创建了unique_events 我使用collection函数中的out: "session_stat"在不同的mapReduce中保存了输出。

var mapFunction = function() {
                      var key = this.app_id;
                      var value = {                                 
                                    event_count: this.event_count,
                                    events: this.events,
                                    unique_events: this.unique_events
                                   };

                      emit( key, value );
                  };

var reduceFunction = function(key, values) {

                        var reducedObject = {
                                              app_id: key,
                                              events_wise_total: 0,
                                              unique_events_wise_total:0,
                                              total_event_count:0
                                            };

                        var events_wise_total = [];
                        var events_0_total = { type:0, value :0};
                        var events_10_total = { type:10, value :0};
                        var events_20_total = { type:20, value :0};
                        var events_30_total = { type:30, value :0};

                        var unique_events_wise_total = [];
                        var unique_events_0_total = { type:0, value :0};
                        var unique_events_10_total = { type:10, value :0};
                        var unique_events_20_total = { type:20, value :0};
                        var unique_events_30_total = { type:30, value :0};

                        var total_event_count = 0;
                        values.forEach( function(value) {
                                total_event_count += value.event_count;
                                var events = value.events;

                                events.forEach(function(event){
                                                if(event.type == 0){events_0_total.value += event.value;}
                                                if(event.type == 10){events_10_total.value += event.value;}
                                                if(event.type == 20){events_20_total.value += event.value;}
                                                if(event.type == 30){events_30_total.value += event.value;}
                                        });

                                var unique_events = value.unique_events;

                                unique_events.forEach(function(unique_event){
                                                if(unique_event.type == 0){unique_events_0_total.value += unique_event.value;}
                                                if(unique_event.type == 10){unique_events_10_total.value += unique_event.value;}
                                                if(unique_event.type == 20){unique_events_20_total.value += unique_event.value;}
                                                if(unique_event.type == 30){unique_events_30_total.value += unique_event.value;}
                                        }); 
                            }
                          );
                        events_wise_total.push(events_0_total);
                        events_wise_total.push(events_10_total);
                        events_wise_total.push(events_20_total);
                        events_wise_total.push(events_30_total);

                        unique_events_wise_total.push(unique_events_0_total);
                        unique_events_wise_total.push(unique_events_10_total);
                        unique_events_wise_total.push(unique_events_20_total);
                        unique_events_wise_total.push(unique_events_30_total);

                        reducedObject.events_wise_total = events_wise_total;
                        reducedObject.unique_events_wise_total = unique_events_wise_total;
                        reducedObject.total_event_count = total_event_count;

                        return reducedObject;
                     };

var finalizeFunction = function (key, reducedValue) {
                          return reducedValue;
                       };                    

db.GroupBy.mapReduce(
                       mapFunction,
                       reduceFunction,
                       {
                         out: "session_stat",
                         finalize: finalizeFunction
                       });

希望它可能有用