mongodb aggregate and mapReduce

标签：aggregate 聚合 mapreduce

Aggregate

MongoDB中聚合(aggregate)主要用于处理数据(诸如统计平均值,求和等)，并返回计算后的数据结果。有点类似sql语句中的 count(*)

语法如下:

db.collection.aggregate()

db.collection.aggregate(pipeline,options)

db.runCommand({

aggregate: "<collection>",

pipeline: [ <stage>, <...> ],

explain: <boolean>,

allowDiskUse: <boolean>,

cursor: <document>

})

在使用aggregate实现聚合操作之前,我们首先来认识下几个常用的聚合操作符。 

$project:：可以对结果集中的键 重命名，控制键是否显示,对列进行计算。 

$match:  过滤结果集 

$group:  分组，聚合，求和，平均数，等 

$skip:  在显示结果的时候跳过前几行 

$sort:  对即将显示的结果集排序 

$limit:  控制结果集的大小

例：

db.createCollection("emp")

db.emp.insert({_id:1,"ename":"tom","age":25,"department":"Sales","salary":6000})

db.emp.insert({_id:2,"ename":"eric","age":24,"department":"HR","salary":4500})

db.emp.insert({_id:3,"ename":"robin","age":30,"department":"Sales","salary":8000})

db.emp.insert({_id:4,"ename":"jack","age":28,"department":"Development","salary":8000})

db.emp.insert({_id:5,"ename":"Mark","age":22,"department":"Development","salary":6500})

db.emp.insert({_id:6,"ename":"marry","age":23,"department":"Planning","salary":5000})

db.emp.insert({_id:7,"ename":"hellen","age":32,"department":"HR","salary":6000})

db.emp.insert({_id:8,"ename":"sarah","age":24,"department":"Development","salary":7000})

> use company

switched to db company

> db.emp.aggregate(

... {$group:{_id:"$department",dpct:{$sum:1}}}

... )

{ "_id" : "Development", "dpct" : 3 }

{ "_id" : "HR", "dpct" : 2 }

{ "_id" : "Planning", "dpct" : 1 }

{ "_id" : "Sales", "dpct" : 2 }

> db.emp.aggregate(

... {$group:{_id:"$department",salct:{$sum:"$salary"},salavg:{$avg:"$salary"}}}

... )

{ "_id" : "Development", "salct" : 21500, "salavg" : 7166.666666666667 }

{ "_id" : "HR", "salct" : 10500, "salavg" : 5250 }

{ "_id" : "Planning", "salct" : 5000, "salavg" : 5000 }

{ "_id" : "Sales", "salct" : 14000, "salavg" : 7000 }

> db.emp.aggregate(

... {$match:{age:{$lt:25}}}

... )

{ "_id" : 2, "ename" : "eric", "age" : 24, "department" : "HR", "salary" : 4500 }

{ "_id" : 5, "ename" : "Mark", "age" : 22, "department" : "Development", "salary" : 6500 }

{ "_id" : 6, "ename" : "marry", "age" : 23, "department" : "Planning", "salary" : 5000 }

{ "_id" : 8, "ename" : "sarah", "age" : 24, "department" : "Development", "salary" : 7000 }

> db.emp.aggregate(

... {$match:{age:{$gt:25}}},

... {$group:{_id:"$department",salct:{$sum:"$salary"},salavg:{$avg:"$salary"}}}

... )

{ "_id" : "HR", "salct" : 6000, "salavg" : 6000 }

{ "_id" : "Development", "salct" : 8000, "salavg" : 8000 }

{ "_id" : "Sales", "salct" : 8000, "salavg" : 8000 }

> db.emp.aggregate(

... {$group:{_id:"$department",salct:{$sum:"$salary"},salavg:{$avg:"$salary"}}},

... {$match:{salavg:{$gt:6000}}}

... )

{ "_id" : "Development", "salct" : 21500, "salavg" : 7166.666666666667 }

{ "_id" : "Sales", "salct" : 14000, "salavg" : 7000 }

>

> db.emp.aggregate(

... {$sort:{age:1}},{$limit:3}

... )

{ "_id" : 5, "ename" : "Mark", "age" : 22, "department" : "Development", "salary" : 6500 }

{ "_id" : 6, "ename" : "marry", "age" : 23, "department" : "Planning", "salary" : 5000 }

{ "_id" : 2, "ename" : "eric", "age" : 24, "department" : "HR", "salary" : 4500 }

> db.emp.aggregate( {$sort:{age:-1}},{$limit:3} )

{ "_id" : 7, "ename" : "hellen", "age" : 32, "department" : "HR", "salary" : 6000 }

{ "_id" : 3, "ename" : "robin", "age" : 30, "department" : "Sales", "salary" : 8000 }

{ "_id" : 4, "ename" : "jack", "age" : 28, "department" : "Development", "salary" : 8000 }

> db.emp.aggregate( {$sort:{age:-1}},{$skip:4} )

{ "_id" : 2, "ename" : "eric", "age" : 24, "department" : "HR", "salary" : 4500 }

{ "_id" : 8, "ename" : "sarah", "age" : 24, "department" : "Development", "salary" : 7000 }

{ "_id" : 6, "ename" : "marry", "age" : 23, "department" : "Planning", "salary" : 5000 }

{ "_id" : 5, "ename" : "Mark", "age" : 22, "department" : "Development", "salary" : 6500 }

>

> db.emp.aggregate( {$project:{"姓名":"$ename","年龄":"$age","部门":"$department","工资":"$salary",_id:0}})

{ "姓名" : "tom", "年龄" : 25, "部门" : "Sales", "工资" : 6000 }

{ "姓名" : "eric", "年龄" : 24, "部门" : "HR", "工资" : 4500 }

{ "姓名" : "robin", "年龄" : 30, "部门" : "Sales", "工资" : 8000 }

{ "姓名" : "jack", "年龄" : 28, "部门" : "Development", "工资" : 8000 }

{ "姓名" : "Mark", "年龄" : 22, "部门" : "Development", "工资" : 6500 }

{ "姓名" : "marry", "年龄" : 23, "部门" : "Planning", "工资" : 5000 }

{ "姓名" : "hellen", "年龄" : 32, "部门" : "HR", "工资" : 6000 }

{ "姓名" : "sarah", "年龄" : 24, "部门" : "Development", "工资" : 7000 }

> db.emp.aggregate( {$project:{"姓名":"$ename","年龄":"$age","部门":"$department","工资":"$salary",_id:0}},{$match:{"工资":{$gt:6000}}})

{ "姓名" : "robin", "年龄" : 30, "部门" : "Sales", "工资" : 8000 }

{ "姓名" : "jack", "年龄" : 28, "部门" : "Development", "工资" : 8000 }

{ "姓名" : "Mark", "年龄" : 22, "部门" : "Development", "工资" : 6500 }

{ "姓名" : "sarah", "年龄" : 24, "部门" : "Development", "工资" : 7000 }

>

Map Reduce

Map-Reduce是一种计算模型，简单的说就是将大批量的工作（数据）分解（MAP）执行，然后再将结果合并成最终结果（REDUCE）。

MongoDB提供的Map-Reduce非常灵活，对于大规模数据分析也相当实用。

以下是MapReduce的基本语法：

>db.collection.mapReduce(

   function() {emit(key,value);},  //map 函数

   function(key,values) {return reduceFunction},   //reduce 函数

   {

      out: collection,

      query: document,

      sort: document,

      limit: number

   }

)

使用 MapReduce 要实现两个函数 Map 函数和 Reduce 函数,Map 函数调用 emit(key, value), 遍历 collection 中所有的记录, 将key 与 value 传递给 Reduce 函数进行处理。

Map 函数必须调用 emit(key, value) 返回键值对。

参数说明:

map ：映射函数 (生成键值对序列,作为 reduce 函数参数)。

reduce 统计函数，reduce函数的任务就是将key-values变成key-value，也就是把values数组变成一个单一的值value。。

out 统计结果存放集合 (不指定则使用临时集合,在客户端断开后自动删除)。

query 一个筛选条件，只有满足条件的文档才会调用map函数。（query。limit，sort可以随意组合）

sort 和limit结合的sort排序参数（也是在发往map函数前给文档排序），可以优化分组机制

limit 发往map函数的文档数量的上限（要是没有limit，单独使用sort的用处不大）

> db.emp.mapReduce( function() { emit(this.department,1); }, function(key,values) { return Array.sum(values) }, { out:"depart_summary" } ).find()

{ "_id" : "Development", "value" : 3 }

{ "_id" : "HR", "value" : 2 }

{ "_id" : "Planning", "value" : 1 }

{ "_id" : "Sales", "value" : 2 }

    利用内置的sum函数返回每个部门的人数

> db.emp.mapReduce( function() { emit(this.department,this.salary); }, function(key,values) {  return Array.avg(values) }, { out:"depart_summary" } ).find()

{ "_id" : "Development", "value" : 7166.666666666667 }

{ "_id" : "HR", "value" : 5250 }

{ "_id" : "Planning", "value" : 5000 }

{ "_id" : "Sales", "value" : 7000 }

    利用内置的avg函数返回每个部门的工资平均数

> db.emp.mapReduce( function() { emit(this.department,this.salary); }, function(key,values) {  return Array.avg(values).toFixed(2) }, { out:"depart_summary" } ).find()

{ "_id" : "Development", "value" : "7166.67" }

{ "_id" : "HR", "value" : "5250.00" }

{ "_id" : "Planning", "value" : 5000 }

{ "_id" : "Sales", "value" : "7000.00" }

>    保留两位小数

> db.emp.mapReduce( function() { emit(this.department,this.salary); }, function(key,values) {  return Array.sum(values) }, { out:"depart_summary" } ).find()

{ "_id" : "Development", "value" : 21500 }

{ "_id" : "HR", "value" : 10500 }

{ "_id" : "Planning", "value" : 5000 }

{ "_id" : "Sales", "value" : 14000 }

>  利用内置的sum函数返回每个部门的工资总和

> db.emp.mapReduce( function() { emit(this.department,{count:1}); }, function(key,values) { var sum=0; values.forEach(function(val){sum+=val.count}); return sum; }, { out:"depart_summary" } ).find()

{ "_id" : "Development", "value" : 3 }

{ "_id" : "HR", "value" : 2 }

{ "_id" : "Planning", "value" : { "count" : 1 } }

{ "_id" : "Sales", "value" : 2 }

>  手工计算每个部门的员工总数

> db.emp.mapReduce( function() { emit(this.department,{salct:this.salary,count:1}); }, function(key,values) { var res={salct:0,sum:0}; values.forEach(function(val){res.sum+=val.count;res.salct+=val.salct}); return res; }, { out:"depart_summary" } ).find()

{ "_id" : "Development", "value" : { "salct" : 21500, "sum" : 3 } }

{ "_id" : "HR", "value" : { "salct" : 10500, "sum" : 2 } }

{ "_id" : "Planning", "value" : { "salct" : 5000, "count" : 1 } }

{ "_id" : "Sales", "value" : { "salct" : 14000, "sum" : 2 } }

>  手工计算每个部门的员工总数和工资总数

> db.emp.mapReduce( function() { emit(this.department,{salct:this.salary,count:1}); }, function(key,values) { var res={salct:0,sum:0}; values.forEach(function(val){res.sum+=val.count;res.salct+=val.salct}); return res.salct/res.sum; }, { out:"depart_summary" } ).find()

{ "_id" : "Development", "value" : 7166.666666666667 }

{ "_id" : "HR", "value" : 5250 }

{ "_id" : "Planning", "value" : { "salct" : 5000, "count" : 1 } }

{ "_id" : "Sales", "value" : 7000 }

>  手工计算每个部门的工资平均值

> db.emp.mapReduce( function() { emit(this.department,this.salary); }, function(key,values) {  return Array.avg(values) }, { out:"depart_summary" } ).find({value:{$gt:5000}})

{ "_id" : "Development", "value" : 7166.666666666667 }

{ "_id" : "HR", "value" : 5250 }

{ "_id" : "Sales", "value" : 7000 }

    将分组计算后的值进行过滤显示，只显示工资平均数大于5000的部门

> db.emp.mapReduce( function() { emit(this.department,this.salary); }, function(key,values) {  return Array.avg(values) }, { out:"depart_summary" } ).find({value:{$gt:5000}}).sort({value:1})

{ "_id" : "HR", "value" : 5250 }

{ "_id" : "Sales", "value" : 7000 }

{ "_id" : "Development", "value" : 7166.666666666667 }

     将分组计算后的值进行排序，默认为升序

> db.emp.mapReduce( function() { emit(this.department,this.salary); }, function(key,values) {  return Array.avg(values) }, { out:"depart_summary" } ).find({value:{$gt:5000}}).sort({value:-1})

{ "_id" : "Development", "value" : 7166.666666666667 }

{ "_id" : "Sales", "value" : 7000 }

{ "_id" : "HR", "value" : 5250 }

>    将分组计算后的值进行排序，手工指定降序

> db.emp.mapReduce( function() { emit(this.department,this.salary); }, function(key,values) {  return Array.avg(values) }, { out:"depart_summary" } ).find({value:{$gt:5000}}).sort({value:-1}).limit(2)

{ "_id" : "Development", "value" : 7166.666666666667 }

{ "_id" : "Sales", "value" : 7000 }

>    将分组计算后的值进行降序排序后，取其中的两个值 

> db.emp.mapReduce( function() { emit(this.department,{count:1}); }, function(key,values) { var sum=0; values.forEach(function(val){sum+=val.count}); return sum; }, { out:"depart_summary",query:{age:{$gt:25}} } ).find()

{ "_id" : "Development", "value" : { "count" : 1 } }

{ "_id" : "HR", "value" : { "count" : 1 } }

{ "_id" : "Sales", "value" : { "count" : 1 } }

>    分组前过滤数据，然后再分组计算

> db.emp.mapReduce( function() { emit(this.department,{count:1}); }, function(key,values) { var sum=0; values.forEach(function(val){sum+=val.count}); return sum; }, { out:"depart_summary",query:{age:{$gt:22}},sort:{age:1} } ).find()

{ "_id" : "Development", "value" : 2 }

{ "_id" : "HR", "value" : 2 }

{ "_id" : "Planning", "value" : { "count" : 1 } }

{ "_id" : "Sales", "value" : 2 }

>   分组前过滤数据，并排序，然后再分组计算 (本示例无意义)

本文出自 “11462293” 博客，请务必保留此出处http://11472293.blog.51cto.com/11462293/1836526

mongodb aggregate and mapReduce

标签：aggregate 聚合 mapreduce

原文地址：http://11472293.blog.51cto.com/11462293/1836526