Map-Reduce转换到聚合管道

从4.4版本开始,MongoDB添加了$accumulator$function aggregation运算符。这些运算符为用户提供了定义自定义聚合表达式的能力。使用这些操作,可以大致重写map-reduce表达式,如下表所示。

注意

可以使用聚合管道操作符(如$group、$merge等)重写各种map-reduce表达式,而不需要自定义函数。

例如,请参见map-reduce示例。

Map-Reduce到聚合管道转换表

这张表只是粗略的翻译。例如,该表显示了使用$projectmapFunction的近似转换。

  • 然而,mapFunction逻辑可能需要额外的阶段,例如,如果逻辑包括对数组的迭代:

    function() {
       this.items.forEach(function(item){ emit(item.sku, 1); });
    }
    

    然后,聚合管道包括一个$unwind和一个$project:

    { $unwind: "$items "},
    { $project: { emits: { key: { "$items.sku" }, value: 1 } } },
    
  • $project中的emit字段可以被命名为其他名称。为了进行可视化比较,选择了字段名称emit。

Map-Reduce Aggregation Pipeline
db.collection.mapReduce(
<mapFunction>,
<reduceFunction>,
{
query: <queryFilter>,
sort: <sortOrder>,
limit: <number>,
finalize: <finalizeFunction>,
out: <collection>
} )
db.collection.aggregate( [
{ $match: <queryFilter> },
{ $sort: <sortOrder> },
{ $limit: <number> },
{ $project: { emits: { k: <expression>, v: <expression> } } },
{ $unwind: “$emits” },
{ $group: {
_id: “$emits.k”},
value: { $accumulator: {
init: <initCode>,
accumulate: <reduceFunction>,
accumulateArgs: [ “$emit.v”],
merge: <reduceFunction>,
finalize: <finalizeFunction>,
lang: “js” }}
} },
{ $out: <collection> }
] )
db.collection.mapReduce(
<mapFunction>,
<reduceFunction>,
{
query: <queryFilter>,
sort: <sortOrder>,
limit: <number>,
finalize: <finalizeFunction>,
out: { merge: <collection>, db: <db> }
}
)
db.collection.aggregate( [
{ $match: <queryFilter> },
{ $sort: <sortOrder> },
{ $limit: <number> },
{ $project: { emits: { k: <expression>, v: <expression> } } },
{ $unwind: “$emits” },
{ $group: {
_id: “$emits.k”},
value: { $accumulator: {
init: <initCode>,
accumulate: <reduceFunction>,
accumulateArgs: [ “$emit.v”],
merge: <reduceFunction>,
finalize: <finalizeFunction>,
lang: “js” }}
} },
{ $out: { db: <db>, coll: <collection> } }
] )
db.collection.mapReduce(
<mapFunction>,
<reduceFunction>,
{
query: <queryFilter>,
sort: <sortOrder>,
limit: <number>,
finalize: <finalizeFunction>,
out: { merge: <collection>, db: <db> }
}
)
db.collection.aggregate( [
{ $match: <queryFilter> },
{ $sort: <sortOrder> },
{ $limit: <number> },
{ $project: { emits: { k: <expression>, v: <expression> } } },
{ $unwind: “$emits” },
{ $group: {
_id: “$emits.k”},
value: { $accumulator: {
init: <initCode>,
accumulate: <reduceFunction>,
accumulateArgs: [ “$emit.v”],
merge: <reduceFunction>,
finalize: <finalizeFunction>,
lang: “js” }}
} },
{ $merge: {
into: { db: <db>, coll: <collection>},
on: “_id”
whenMatched: “replace”,
whenNotMatched: “insert”
} },
] )
db.collection.mapReduce(
<mapFunction>,
<reduceFunction>,
{
query: <queryFilter>,
sort: <sortOrder>,
limit: <number>,
finalize: <finalizeFunction>,
out: { merge: <collection>, db: <db> }
}
)
db.collection.aggregate( [
{ $match: <queryFilter> },
{ $sort: <sortOrder> },
{ $limit: <number> },
{ $project: { emits: { k: <expression>, v: <expression> } } },
{ $unwind: “$emits” },
{ $group: {
_id: “$emits.k”},
value: { $accumulator: {
init: <initCode>,
accumulate: <reduceFunction>,
accumulateArgs: [ “$emit.v”],
merge: <reduceFunction>,
finalize: <finalizeFunction>,
lang: “js” }}
} },
{ $merge: {
into: { db: <db>, coll: <collection> },
on: “_id”
whenMatched: [
{ $project: {
value: { $function: {
body: <reduceFunction>,
args: [
“$_id”,
[ “$value”, “$$new.value” ]
],
lang: “js”
} }
} }
]
whenNotMatched: “insert”
} },
] )
db.collection.mapReduce(
<mapFunction>,
<reduceFunction>,
{
query: <queryFilter>,
sort: <sortOrder>,
limit: <number>,
finalize: <finalizeFunction>,
out: { inline: 1 }
}
)
db.collection.aggregate( [
{ $match: <queryFilter> },
{ $sort: <sortOrder> },
{ $limit: <number> },
{ $project: { emits: { k: <expression>, v: <expression> } } },
{ $unwind: “$emits” },
{ $group: {
_id: “$emits.k”},
value: { $accumulator: {
init: <initCode>,
accumulate: <reduceFunction>,
accumulateArgs: [ “$emit.v”],
merge: <reduceFunction>,
finalize: <finalizeFunction>,
lang: “js” }}
} }
] )

例子

可以使用聚合管道操作符(如$group$merge等)重写各种map-reduce表达式,而不需要自定义函数。但是,为了说明目的,下面的例子提供了两种选择。

示例1

通过cust_id对订单集合组执行以下map-reduce操作,并计算每个cust_id的价格总和:

var mapFunction1 = function() {
   emit(this.cust_id, this.price);
};

var reduceFunction1 = function(keyCustId, valuesPrices) {
   return Array.sum(valuesPrices);
};

db.orders.mapReduce(
   mapFunction1,
   reduceFunction1,
   { out: "map_reduce_example" }
)

备选方案1:(推荐)您可以重写操作到聚合管道,而不将map-reduce函数转换为等效的管道阶段:

db.orders.aggregate([
   { $group: { _id: "$cust_id", value: { $sum: "$price" } } },
   { $out: "agg_alternative_1" }
])

备选方案2:(仅为说明目的)下面的聚合管道提供了各种map-reduce函数的转换,使用$accumulator定义自定义函数:

db.orders.aggregate( [
   { $project: { emit: { key: "$cust_id", value: "$price" } } },  // equivalent to the map function
   { $group: {                                                    // equivalent to the reduce function
        _id: "$emit.key",
        valuesPrices: { $accumulator: {
                    init: function() { return 0; },
                    initArgs: [],
                    accumulate: function(state, value) { return state + value; },
                    accumulateArgs: [ "$emit.value" ],
                    merge: function(state1, state2) { return state1 + state2; },
                    lang: "js"
        } }
   } },
   { $out: "agg_alternative_2" }
] )
  1. 首先,$project阶段输出带有emit字段的文档。emit字段是一个包含以下字段的文档:

    • key包含cust_id文档的值
    • value包含price文档的值
    { "_id" : 1, "emit" : { "key" : "Ant O. Knee", "value" : 25 } }
    { "_id" : 2, "emit" : { "key" : "Ant O. Knee", "value" : 70 } }
    { "_id" : 3, "emit" : { "key" : "Busby Bee", "value" : 50 } }
    { "_id" : 4, "emit" : { "key" : "Busby Bee", "value" : 25 } }
    { "_id" : 5, "emit" : { "key" : "Busby Bee", "value" : 50 } }
    { "_id" : 6, "emit" : { "key" : "Cam Elot", "value" : 35 } }
    { "_id" : 7, "emit" : { "key" : "Cam Elot", "value" : 25 } }
    { "_id" : 8, "emit" : { "key" : "Don Quis", "value" : 75 } }
    { "_id" : 9, "emit" : { "key" : "Don Quis", "value" : 55 } }
    { "_id" : 10, "emit" : { "key" : "Don Quis", "value" : 25 } }
    
  2. 然后,$group使用$accumulator操作符来添加发出的值:

    { "_id" : "Don Quis", "valuesPrices" : 155 }
    { "_id" : "Cam Elot", "valuesPrices" : 60 }
    { "_id" : "Ant O. Knee", "valuesPrices" : 95 }
    { "_id" : "Busby Bee", "valuesPrices" : 125 }
    
  3. 最后,$out将输出写入集合agg_alternative_2。或者,您可以使用$merge而不是$out

    示例2

以下字段对orders集合组的map-reduce操作,item.sku并计算每个sku的订单数量和总订购量。然后,该操作将为每个sku值计算每个订单的平均数量,并将结果合并到输出集合中。

var mapFunction2 = function() {
    for (var idx = 0; idx < this.items.length; idx++) {
       var key = this.items[idx].sku;
       var value = { count: 1, qty: this.items[idx].qty };

       emit(key, value);
    }
};

var reduceFunction2 = function(keySKU, countObjVals) {
   reducedVal = { count: 0, qty: 0 };

   for (var idx = 0; idx < countObjVals.length; idx++) {
       reducedVal.count += countObjVals[idx].count;
       reducedVal.qty += countObjVals[idx].qty;
   }

   return reducedVal;
};

var finalizeFunction2 = function (key, reducedVal) {
  reducedVal.avg = reducedVal.qty/reducedVal.count;
  return reducedVal;
};

db.orders.mapReduce(
   mapFunction2,
   reduceFunction2,
   {
     out: { merge: "map_reduce_example2" },
     query: { ord_date: { $gte: new Date("2020-03-01") } },
     finalize: finalizeFunction2
   }
 );

备选方案1:(推荐)您可以重写操作到聚合管道,而不将map-reduce函数转换为等效的管道阶段:

db.orders.aggregate( [
   { $match: { ord_date: { $gte: new Date("2020-03-01") } } },
   { $unwind: "$items" },
   { $group: { _id: "$items.sku", qty: { $sum: "$items.qty" }, orders_ids: { $addToSet: "$_id" } }  },
   { $project: { value: { count: { $size: "$orders_ids" }, qty: "$qty", avg: { $divide: [ "$qty", { $size: "$orders_ids" } ] } } } },
   { $merge: { into: "agg_alternative_3", on: "_id", whenMatched: "replace",  whenNotMatched: "insert" } }
] )

备选方案2:(仅为说明目的)下面的聚合管道提供了各种map-reduce函数的转换,使用$accumulator定义自定义函数:

db.orders.aggregate( [
    { $match: { ord_date: {$gte: new Date("2020-03-01") } } },
    { $unwind: "$items" },
    { $project: { emit: { key: "$items.sku", value: { count: { $literal: 1 }, qty: "$items.qty" } } } },
    { $group: {
           _id: "$emit.key",
           value: { $accumulator: {
             init: function() { return { count: 0, qty: 0 }; },
             initArgs: [],
             accumulate: function(state, value) {
                  state.count += value.count;
                  state.qty += value.qty;
                  return state;
             },
             accumulateArgs: [ "$emit.value" ],
             merge: function(state1, state2) {
                return { count: state1.count + state2.count, qty: state1.qty + state2.qty };
             },
             finalize: function(state) {
                state.avg = state.qty / state.count;
                return state;
             },
             lang: "js"}
          }
    } },
    { $merge: {
       into: "agg_alternative_4",
       on: "_id",
       whenMatched: "replace",
       whenNotMatched: "insert"
    } }
] )
  1. $match阶段只选择那些ord_date大于或等于new Date("2020-03-01")的文档。

  2. $unwinds阶段按items数组字段分解文档,为每个数组元素输出一个文档。例如:

    { "_id" : 1, "cust_id" : "Ant O. Knee", "ord_date" : ISODate("2020-03-01T00:00:00Z"), "price" : 25, "items" : { "sku" : "oranges", "qty" : 5, "price" : 2.5 }, "status" : "A" }
    { "_id" : 1, "cust_id" : "Ant O. Knee", "ord_date" : ISODate("2020-03-01T00:00:00Z"), "price" : 25, "items" : { "sku" : "apples", "qty" : 5, "price" : 2.5 }, "status" : "A" }
    { "_id" : 2, "cust_id" : "Ant O. Knee", "ord_date" : ISODate("2020-03-08T00:00:00Z"), "price" : 70, "items" : { "sku" : "oranges", "qty" : 8, "price" : 2.5 }, "status" : "A" }
    { "_id" : 2, "cust_id" : "Ant O. Knee", "ord_date" : ISODate("2020-03-08T00:00:00Z"), "price" : 70, "items" : { "sku" : "chocolates", "qty" : 5, "price" : 10 }, "status" : "A" }
    { "_id" : 3, "cust_id" : "Busby Bee", "ord_date" : ISODate("2020-03-08T00:00:00Z"), "price" : 50, "items" : { "sku" : "oranges", "qty" : 10, "price" : 2.5 }, "status" : "A" }
    { "_id" : 3, "cust_id" : "Busby Bee", "ord_date" : ISODate("2020-03-08T00:00:00Z"), "price" : 50, "items" : { "sku" : "pears", "qty" : 10, "price" : 2.5 }, "status" : "A" }
    { "_id" : 4, "cust_id" : "Busby Bee", "ord_date" : ISODate("2020-03-18T00:00:00Z"), "price" : 25, "items" : { "sku" : "oranges", "qty" : 10, "price" : 2.5 }, "status" : "A" }
    { "_id" : 5, "cust_id" : "Busby Bee", "ord_date" : ISODate("2020-03-19T00:00:00Z"), "price" : 50, "items" : { "sku" : "chocolates", "qty" : 5, "price" : 10 }, "status" : "A" }
    ...
    
  3. $project阶段输出带有emit字段的文档。emit字段是一个包含以下字段的文档:

    • key包含items.sku
    • value包含具有qty值和count值的文档
    { "_id" : 1, "emit" : { "key" : "oranges", "value" : { "count" : 1, "qty" : 5 } } }
    { "_id" : 1, "emit" : { "key" : "apples", "value" : { "count" : 1, "qty" : 5 } } }
    { "_id" : 2, "emit" : { "key" : "oranges", "value" : { "count" : 1, "qty" : 8 } } }
    { "_id" : 2, "emit" : { "key" : "chocolates", "value" : { "count" : 1, "qty" : 5 } } }
    { "_id" : 3, "emit" : { "key" : "oranges", "value" : { "count" : 1, "qty" : 10 } } }
    { "_id" : 3, "emit" : { "key" : "pears", "value" : { "count" : 1, "qty" : 10 } } }
    { "_id" : 4, "emit" : { "key" : "oranges", "value" : { "count" : 1, "qty" : 10 } } }
    { "_id" : 5, "emit" : { "key" : "chocolates", "value" : { "count" : 1, "qty" : 5 } } }
    ...
    
  4. $group使用$accumulator操作符来添加发出的计数和数量,并计算avg字段:

    { "_id" : "chocolates", "value" : { "count" : 3, "qty" : 15, "avg" : 5 } }
    { "_id" : "oranges", "value" : { "count" : 7, "qty" : 63, "avg" : 9 } }
    { "_id" : "carrots", "value" : { "count" : 2, "qty" : 15, "avg" : 7.5 } }
    { "_id" : "apples", "value" : { "count" : 4, "qty" : 35, "avg" : 8.75 } }
    { "_id" : "pears", "value" : { "count" : 1, "qty" : 10, "avg" : 10 } }
    
  5. 最后,$merge将输出写入集合agg_alternative_4。如果现有文档具有与新结果相同的键_id,则操作将覆盖现有文档。如果没有具有相同密钥的现有文档,操作将插入该文档。

也可以看看
聚合命令比较

译者:李冠飞

校对:

参见

原文 - Map-Reduce to Aggregation Pipeline

Copyright © 上海锦木信息技术有限公司 all right reserved,powered by Gitbook文件修订时间: 2020-12-18 11:34:57

results matching ""

    No results matching ""