MongoDB没有使用multikey索引

时间:2015-05-26 21:50:58

标签: java mongodb indexing multikey

我有一个MongoDB 3.0.3版集合,其中包含2个字段的文档:

  • 数字长ID
  • 60个数字双打数组。

我在数组上构建了一个多键索引,并使用getIndexInfo()方法确认索引存在并且是多键的。但是,当我查询数组中两个特定字段范围的交集时,Mongo不使用此索引,尽管集合中有1,000,000个文档。我可以从explain()方法的输出中看到这一点。更奇怪的是,当我指定使用带有hint的索引时,Mongo遍历所有1,000,000个文档和60,000,000个索引条目,我在explain()的输出中看到。

我正在使用以下代码构建查询:

BasicDBObject q2 = new BasicDBObject("array.0",new BasicDBObject("$lt",1000.0));
BasicDBObject q1 = new BasicDBObject("array.1",new BasicDBObject("$gte",800.0));
BasicDBObject q_and = new BasicDBObject("$and",Arrays.asList(q1,q2));
dbo = collection.find(q_and).explain();

有什么想法吗?

提前感谢您的帮助。

回复解释输出请求:

如果没有指定提示,则解释输出如下:

{ "queryPlanner" : 
{ "plannerVersion" : 1 , "namespace" : "local.TestArrays" , "indexFilterSet" : false , "parsedQuery" : 
{ "$and" : [ 
{ "array.0" : 
{ "$lt" : 1000.0
}
} , 
{ "array.1" : 
{ "$gte" : 800.0
}
}]
} , "winningPlan" : 
{ "stage" : "COLLSCAN" , "filter" : 
{ "$and" : [ 
{ "array.0" : 
{ "$lt" : 1000.0
}
} , 
{ "array.1" : 
{ "$gte" : 800.0
}
}]
} , "direction" : "forward"
} , "rejectedPlans" : [ ]
} , "executionStats" : 
{ "executionSuccess" : true , "nReturned" : 2 , "executionTimeMillis" : 2248 , "totalKeysExamined" : 0 , "totalDocsExamined" : 1000000 , "executionStages" : 
{ "stage" : "COLLSCAN" , "filter" : 
{ "$and" : [ 
{ "array.0" : 
{ "$lt" : 1000.0
}
} , 
{ "array.1" : 
{ "$gte" : 800.0
}
}]
} , "nReturned" : 2 , "executionTimeMillisEstimate" : 2190 , "works" : 1000002 , "advanced" : 2 , "needTime" : 999999 , "needFetch" : 0 , "saveState" : 7812 , "restoreState" : 7812 , "isEOF" : 1 , "invalidates" : 0 , "direction" : "forward" , "docsExamined" : 1000000
} , "allPlansExecution" : [ ]
} , "serverInfo" : 
{ "host" : "NYDEVWS0005052" , "port" : 27017 , "version" : "3.0.3" , "gitVersion" : "b40106b36eecd1b4407eb1ad1af6bc60593c6105"
}
}

指定要与以下代码一起使用的索引时:

collection.find(q_and).hint( “array_1”)解释();

解释输出如下:

{ "queryPlanner" : 
{ "plannerVersion" : 1 , "namespace" : "local.TestArrays" , "indexFilterSet" : false , "parsedQuery" : 
{ "$and" : [ 
{ "array.0" : 
{ "$lt" : 1000.0
}
} , 
{ "array.1" : 
{ "$gte" : 800.0
}
}]
} , "winningPlan" : 
{ "stage" : "KEEP_MUTATIONS" , "inputStage" : 
{ "stage" : "FETCH" , "filter" : 
{ "$and" : [ 
{ "array.0" : 
{ "$lt" : 1000.0
}
} , 
{ "array.1" : 
{ "$gte" : 800.0
}
}]
} , "inputStage" : 
{ "stage" : "IXSCAN" , "keyPattern" : 
{ "array" : 1
} , "indexName" : "array_1" , "isMultiKey" : true , "direction" : "forward" , "indexBounds" : 
{ "array" : [ "[MinKey, MaxKey]"]
}
}
}
} , "rejectedPlans" : [ ]
} , "executionStats" : 
{ "executionSuccess" : true , "nReturned" : 2 , "executionTimeMillis" : 61401 , "totalKeysExamined" : 60000000 , "totalDocsExamined" : 1000000 , "executionStages" : 
{ "stage" : "KEEP_MUTATIONS" , "nReturned" : 2 , "executionTimeMillisEstimate" : 56570 , "works" : 60001744 , "advanced" : 2 , "needTime" : 59999998 , "needFetch" : 1743 , "saveState" : 470130 , "restoreState" : 470130 , "isEOF" : 1 , "invalidates" : 0 , "inputStage" : 
{ "stage" : "FETCH" , "filter" : 
{ "$and" : [ 
{ "array.0" : 
{ "$lt" : 1000.0
}
} , 
{ "array.1" : 
{ "$gte" : 800.0
}
}]
} , "nReturned" : 2 , "executionTimeMillisEstimate" : 55620 , "works" : 60001744 , "advanced" : 2 , "needTime" : 59999998 , "needFetch" : 1743 , "saveState" : 470130 , "restoreState" : 470130 , "isEOF" : 1 , "invalidates" : 0 , "docsExamined" : 1000000 , "alreadyHasObj" : 0 , "inputStage" : 
{ "stage" : "IXSCAN" , "nReturned" : 1000000 , "executionTimeMillisEstimate" : 50820 , "works" : 60000000 , "advanced" : 1000000 , "needTime" : 59000000 , "needFetch" : 0 , "saveState" : 470130 , "restoreState" : 470130 , "isEOF" : 1 , "invalidates" : 0 , "keyPattern" : 
{ "array" : 1
} , "indexName" : "array_1" , "isMultiKey" : true , "direction" : "forward" , "indexBounds" : 
{ "array" : [ "[MinKey, MaxKey]"]
} , "keysExamined" : 60000000 , "dupsTested" : 60000000 , "dupsDropped" : 59000000 , "seenInvalidated" : 0 , "matchTested" : 0
}
}
} , "allPlansExecution" : [ ]
} , "serverInfo" : 
{ "host" : "NYDEVWS0005052" , "port" : 27017 , "version" : "3.0.3" , "gitVersion" : "b40106b36eecd1b4407eb1ad1af6bc60593c6105"
}
}

2 个答案:

答案 0 :(得分:1)

我误解了多键索引的含义。我想为数组中的每个位置创建了一个单独的索引。但是,在进一步阅读时,我看到为数组创建了一个SINGLE索引,并且文档的所有条目都输入到该单个索引中,而不是保留数组位置。我接受了n9code的建议,并在每个字段上创建了单独的字段和索引。这符合预期。

答案 1 :(得分:0)

我担心您的问题是您正在查询数组的特定元素。因此,即使您在该数组上有索引,查询也必须显式扫描整个集合,从0中获取元素1array并执行匹配。

如果您需要在特定元素上精确地运行匹配,则值得将它们从数组中拉出并将它们作为单独的字段,然后在它们上创建索引。所以考虑一下:

{
   ...
   array: [...],
   ex_element_0: "value0",
   ex_element_1: "value0",
   ...
}