MongoDB新的數(shù)據(jù)統(tǒng)計(jì)框架介紹

目前的MongoDB在進(jìn)行復(fù)雜的數(shù)據(jù)統(tǒng)計(jì)計(jì)算時(shí)都需要寫(xiě)MapReduce來(lái)實(shí)現(xiàn)，包括在SQL中比較常用的group by查詢(xún)也需要寫(xiě)一個(gè)reduce才能實(shí)現(xiàn)，這是比較麻煩的。在MongoDB2.1中，將會(huì)引入一套全新的數(shù)據(jù)統(tǒng)計(jì)計(jì)算框架，讓用戶(hù)更方便的進(jìn)行統(tǒng)計(jì)操作。
-
 

下面我們就來(lái)看看幾個(gè)新的操作符：

$match
$match的作用是過(guò)濾數(shù)據(jù)，通過(guò)設(shè)置一個(gè)條件，將數(shù)據(jù)進(jìn)行篩選過(guò)濾，例子：

db.runCommand({ aggregate : "article", pipeline : [
    { $match : { author : "dave" } }
]});這相當(dāng)于將article這個(gè)collection中的記錄進(jìn)行篩選，篩選條件是author屬性值為dave，其作用其實(shí)相當(dāng)于普通的find命令，如：

> db.article.find({ author : "dave" });
所以，那這個(gè)命令有什么用呢？與find不同，find的結(jié)果是直接作為最終數(shù)據(jù)返回，而$match只是pipeline中的一環(huán)，它篩選的結(jié)果數(shù)據(jù)可以再進(jìn)行下一級(jí)的統(tǒng)計(jì)操作。

$project
$project命令用于設(shè)定數(shù)據(jù)的篩選字段，就像我們SQL中select需要的字段一樣。例子：

db.runCommand({ aggregate : "article", pipeline : [
    { $match : { author : "dave" } },
    { $project : {
        _id : 0,
 author : 1,
        tags : 1
    }}
]});上面就是將所有author為dave的記錄的author和tags兩個(gè)字段取出來(lái)。（_id:0 表示去掉默認(rèn)會(huì)返回的_id字段）

其實(shí)上面這個(gè)功能也能用我們平時(shí)用的find命令來(lái)實(shí)現(xiàn)，如：

> db.article.find({ author : "dave" }, { _id : 0, author : 1, tags : 1);
$unwind
$unwind命令很神奇，他可以將某一個(gè)為array類(lèi)型字段的數(shù)據(jù)拆分成多條，每一條包含array中的一個(gè)屬性。
比如你使用下面命令添加一條記錄：

db.article.save( {
    title : "this is your title" ,
    author : "dave" ,
    posted : new Date(4121381470000) ,
    pageViews : 7 ,
    tags : [ "fun" , "nasty" ] ,
    comments : [
        { author :"barbara" , text : "this is interesting" } ,
        { author :"jenny" , text : "i like to play pinball", votes: 10 }
    ],
    other : { bar : 14 }
});這里面tags字段就是一個(gè)array。下面我們?cè)谶@個(gè)字段上應(yīng)用$unwind操作

db.runCommand({ aggregate : "article", pipeline : [
    { $unwind : "$tags" }
]});上面命令的意思就是按tags字段來(lái)拆分，此命令執(zhí)行的結(jié)果如下：

{
        "result" : [
                {
                        "_id" : ObjectId("4eeeb5fef09a7c9170df094b"),
                        "title" : "this is your title",
                        "author" : "dave",
                        "posted" : ISODate("2100-08-08T04:11:10Z"),
                        "pageViews" : 7,
                        "tags" : "fun",
                        "comments" : [
                                {
                                        "author" : "barbara",
                                        "text" : "this is interesting"
                                },
                                {
                                        "author" : "jenny",
                                        "text" : "i like to play pinball",
                                        "votes" : 10
                                }
                        ],
                        "other" : {
                                "bar" : 14
                        }
                },
                {
                        "_id" : ObjectId("4eeeb5fef09a7c9170df094b"),
                        "title" : "this is your title",
                        "author" : "dave",
                        "posted" : ISODate("2100-08-08T04:11:10Z"),
                        "pageViews" : 7,
                        "tags" : "nasty",
                        "comments" : [
                                {
                                        "author" : "barbara",
                                        "text" : "this is interesting"
                                },
                                {
                                        "author" : "jenny",
                                        "text" : "i like to play pinball",
                                        "votes" : 10
                                }
                        ],
                        "other" : {
                                "bar" : 14
                        }
                }
        ],
        "ok" : 1
}我們可以看到，原來(lái)的tags字段是一個(gè)包含兩個(gè)元素的數(shù)組，通過(guò)$unwind命令后，被拆分成兩條記錄，每一條記錄的tags字段是原來(lái)數(shù)組中的一個(gè)元素。

$group
$group命令比較好理解，功能就是按某一個(gè)key將key值相同的多條數(shù)據(jù)組織成一條。
比如我們使用下面命令再往article這個(gè)collection中寫(xiě)入一條記錄，這時(shí)候我們就有兩條記錄了：

db.article.save( {
    title : "this is some other title" ,
    author : "jane" ,
    posted : new Date(978239834000) ,
    pageViews : 6 ,
    tags : [ "nasty" , "filthy" ] ,
    comments : [
        { author :"will" , text : "i don't like the color" } ,
        { author :"jenny" , text : "can i get that in green?" }
    ],
    other : { bar : 14 }
});我們可以先用上面的$unwind按tags將記錄拆成多條，然后再將記錄按tags字段重新組織，將同一個(gè)tag對(duì)應(yīng)的所有author放在一個(gè)array中。只需要像下面這樣寫(xiě)：

db.runCommand({ aggregate : "article", pipeline : [
    { $unwind : "$tags" },
    { $group : {
 _id : "$tags",
        count : { $sum : 1 },
 authors : { $addToSet : "$author" }
    }}
]});這時(shí)候你就能得到如下結(jié)果了

{
        "result" : [
                {
                        "_id" : "filthy",
                        "count" : 1,
                        "authors" : [
                                "jane"
                        ]
                },
                {
                        "_id" : "fun",
                        "count" : 1,
                        "authors" : [
                                "dave"
                        ]
                },
                {
                        "_id" : "nasty",
                        "count" : 2,
                        "authors" : [
                                "jane",
                                "dave"
                        ]
                }
        ],
        "ok" : 1
}上面是2.1版本將會(huì)推出的一些新的統(tǒng)計(jì)類(lèi)命令的介紹，在易用性方面它們提供給我們很多便利，但是MongoDB MapReduce的最大硬傷，單個(gè)mongod中無(wú)法并行執(zhí)行，貌似還是沒(méi)有解決。雖然其命令中采用了pipeline 的組織模式，但是貌似還是完全串行且分降段完成的。

服務(wù)器租用/服務(wù)器托管中國(guó)五強(qiáng)！虛擬主機(jī)域名注冊(cè)頂級(jí)提供商！15年品質(zhì)保障！--億恩科技[ENKJ.COM]