MongoDB:4. Index
优化python文档indexmongodb
摘要:MongoDB 提供了多样性的索引支持。 for ( var i=0;i30;i++){ ...u={name: user +i, ...age:20+i, ...contact:{ ...address:[ address1_ +i, address2_ +i], ...postcode:100000+i, ...} ...}; ...db.users.insert(u); ...} 索引信息被保存在 system
MongoDB 提供了多样性的索引支持。
> for (var i = 0; i < 30; i++) {
... u = { name : "user" + i,
... age : 20 + i,
... contact : {
... address : ["address1_" + i, "address2_" + i],
... postcode : 100000 + i,
... }
... };
... db.users.insert(u);
... }
索引信息被保存在 system.indexes 中,且默认总是为 _id 创建索引。
> show collections
system.indexes
users
> db.system.indexes.find()
{ "name" : "_id_", "ns" : "blog.users", "key" : { "_id" : 1 } }
1. ensureIndex / dropIndex / reIndex
使用 ensureIndex 创建索引,dropIndex() 删除索引,dropIndexes() 删除全部索引(不包括 _id 等系统索引)。
> db.users.ensureIndex({name:1})
> db.users.ensureIndex({age:1})
> db.system.indexes.find()
{ "name" : "_id_", "ns" : "blog.users", "key" : { "_id" : 1 } }
{ "_id" : ObjectId("4c4a...b798"), "ns" : "blog.users", "key" : { "name" : 1 }, "name" : "name_1" }
{ "_id" : ObjectId("4c4a...b799"), "ns" : "blog.users", "key" : { "age" : 1 }, "name" : "age_1" }
> db.users.dropIndex({age:1})
{ "nIndexesWas" : 3, "ok" : true }
> db.users.dropIndexes()
{
"nIndexesWas" : 2,
"msg" : "non-_id indexes dropped for collection",
"ok" : true
}
> db.system.indexes.find()
{ "name" : "_id_", "ns" : "blog.users", "key" : { "_id" : 1 } }
reIndex 则是重建索引。
> db.users.ensureIndex({name:1})
> db.users.ensureIndex({age:1})
> db.system.indexes.find()
{ "name" : "_id_", "ns" : "blog.users", "key" : { "_id" : 1 } }
{ "_id" : ObjectId("4c4a...b82a"), "ns" : "blog.users", "key" : { "name" : 1 }, "name" : "name_1" }
{ "_id" : ObjectId("4c4a...b82b"), "ns" : "blog.users", "key" : { "age" : 1 }, "name" : "age_1" }
> db.users.reIndex()
{
"nIndexesWas" : 3,
"msg" : "indexes dropped for collection",
"ok" : 1,
"nIndexes" : 3,
"indexes" : [
{
"name" : "_id_",
"ns" : "blog.users",
"key" : {
"_id" : 1
}
},
{
"_id" : ObjectId("4c4a...b82a"),
"ns" : "blog.users",
"key" : {
"name" : 1
},
"name" : "name_1"
},
{
"_id" : ObjectId("4c4a...b82b"),
"ns" : "blog.users",
"key" : {
"age" : 1
},
"name" : "age_1"
}
],
"ok" : 1
}
当系统已有大量数据时,创建索引就是个非常耗时的活,我们可以在后台执行。
> db.users.dropIndexes()
{
"nIndexesWas" : 3,
"msg" : "non-_id indexes dropped for collection",
"ok" : true
}
> db.users.ensureIndex({name:1}, {backgroud:true})
> db.users.reIndex({backgroud:true})
{
"nIndexesWas" : 2,
"msg" : "indexes dropped for collection",
"ok" : 1,
"nIndexes" : 2,
"indexes" : [
{
"name" : "_id_",
"ns" : "blog.users",
"key" : {
"_id" : 1
}
},
{
"_id" : ObjectId("4c4a...b79c"),
"ns" : "blog.users",
"key" : {
"name" : 1
},
"name" : "name_1",
"backgroud" : true
}
],
"ok" : 1
}
2. explain
MongoDB 提供了一个 explain 命令让我们获知系统如何处理查询请求。
> db.users.ensureIndex({name:1})
> db.users.ensureIndex({age:1})
> db.users.find({age:{$gt:45}}, {name:1, age:1})
{ "_id" : ObjectId("4c4a8edeeb257107735eb826"), "name" : "user26", "age" : 46 }
{ "_id" : ObjectId("4c4a8edeeb257107735eb827"), "name" : "user27", "age" : 47 }
{ "_id" : ObjectId("4c4a8edeeb257107735eb828"), "name" : "user28", "age" : 48 }
{ "_id" : ObjectId("4c4a8edeeb257107735eb829"), "name" : "user29", "age" : 49 }
> db.users.find({age:{$gt:45}}, {name:1, age:1}).explain()
{
"cursor" : "BtreeCursor age_1",
"nscanned" : 5,
"nscannedObjects" : 4,
"n" : 4,
"millis" : 0,
"indexBounds" : [
[
{
"age" : 45
},
{
"age" : 1.7976931348623157e+308
}
]
]
}
返回结果信息包括:
•cursor: 返回游标类型(BasicCursor 或 BtreeCursor)。
•nscanned: 被扫描的文档数量。
•n: 返回的文档数量。
•millis: 耗时(毫秒)。
•indexBounds: 所使用的索引。
利用 explain 命令,我们可以很好地观察系统如何使用索引来加快检索,同时可以针对性优化索引。
3. Embedded Keys Index
我们可以创建深层索引,甚至直接用文档(sub-document)作为索引键。
> db.users.ensureIndex({"contact.postcode":1})
> db.users.find({"contact.postcode":{$lt:100009}}, {name:1, "contact.postcode":1})
{ "_id" : ObjectId("4c4a8edeeb257107735eb80c"), "name" : "user0", "contact" : { "postcode" : 100000 } }
{ "_id" : ObjectId("4c4a8edeeb257107735eb80d"), "name" : "user1", "contact" : { "postcode" : 100001 } }
{ "_id" : ObjectId("4c4a8edeeb257107735eb80e"), "name" : "user2", "contact" : { "postcode" : 100002 } }
{ "_id" : ObjectId("4c4a8edeeb257107735eb80f"), "name" : "user3", "contact" : { "postcode" : 100003 } }
{ "_id" : ObjectId("4c4a8edeeb257107735eb810"), "name" : "user4", "contact" : { "postcode" : 100004 } }
{ "_id" : ObjectId("4c4a8edeeb257107735eb811"), "name" : "user5", "contact" : { "postcode" : 100005 } }
{ "_id" : ObjectId("4c4a8edeeb257107735eb812"), "name" : "user6", "contact" : { "postcode" : 100006 } }
{ "_id" : ObjectId("4c4a8edeeb257107735eb813"), "name" : "user7", "contact" : { "postcode" : 100007 } }
{ "_id" : ObjectId("4c4a8edeeb257107735eb814"), "name" : "user8", "contact" : { "postcode" : 100008 } }
> db.users.find({"contact.postcode":{$lt:100009}}, {name:1, "contact.postcode":1}).explain()
{
"cursor" : "BtreeCursor contact.postcode_1",
"nscanned" : 10,
"nscannedObjects" : 9,
"n" : 9,
"millis" : 0,
"indexBounds" : [
[
{
"contact.postcode" : -1.7976931348623157e+308
},
{
"contact.postcode" : 100009
}
]
]
}
我们直接用 contact 创建索引,查找其下属性时可使用该索引,但需注意语法。
(附注: 在 1.5.4 mongo 中,一直无法使用 contact:{postcode:xxx} 这样的 SubObject 语法查询数据,只能用 "contact.postcode" DotNotation 语法)
> db.users.dropIndex({"contact.postcode":1})
{ "nIndexesWas" : 4, "ok" : true }
> db.users.ensureIndex({contact:1})
> db.system.indexes.find()
{ "name" : "_id_", "ns" : "blog.users", "key" : { "_id" : 1 } }
{ "_id" : ObjectId("4c4a...b82a"), "ns" : "blog.users", "key" : { "name" : 1 }, "name" : "name_1" }
{ "_id" : ObjectId("4c4a...b82b"), "ns" : "blog.users", "key" : { "age" : 1 }, "name" : "age_1" }
{ "_id" : ObjectId("4c4a...b82d"), "ns" : "blog.users", "key" : { "contact" : 1 }, "name" : "contact_1" }
> db.users.find({contact:{postcode:{$lt:100009}}}).explain()
{
"cursor" : "BtreeCursor contact_1",
"nscanned" : 0,
"nscannedObjects" : 0,
"n" : 0,
"millis" : 0,
"indexBounds" : [
[
{
"contact" : {
"postcode" : {
"$lt" : 100009
}
}
},
{
"contact" : {
"postcode" : {
"$lt" : 100009
}
}
}
]
]
}
> db.users.find({"contact.postcode":{$lt:100009}}).explain() // 无法使用索引
{
"cursor" : "BasicCursor",
"nscanned" : 30,
"nscannedObjects" : 30,
"n" : 9,
"millis" : 0,
"indexBounds" : [ ]
}
> db.users.find({contact:{address:"address2_23"}}).explain()
{
"cursor" : "BtreeCursor contact_1",
"nscanned" : 0,
"nscannedObjects" : 0,
"n" : 0,
"millis" : 0,
"indexBounds" : [
[
{
"contact" : {
"address" : "address2_23"
}
},
{
"contact" : {
"address" : "address2_23"
}
}
]
]
}
> db.users.find({"contact.address":"address2_23"}).explain() // 无法使用索引
{
"cursor" : "BasicCursor",
"nscanned" : 30,
"nscannedObjects" : 30,
"n" : 1,
"millis" : 0,
"indexBounds" : [ ]
}
同样的语法问题在 Python 中一样生效。
>>> db.users.find({"contact":{"postcode":{"$lt":100009}}}).explain()
{u'allPlans': [{u'cursor': u'BtreeCursor contact_1',
u'indexBounds': [[{u'contact': {u'postcode': {u'$lt': 100009}}},
{u'contact': {u'postcode': {u'$lt': 100009}}}]]}],
u'cursor': u'BtreeCursor contact_1',
u'indexBounds': [[{u'contact': {u'postcode': {u'$lt': 100009}}},
{u'contact': {u'postcode': {u'$lt': 100009}}}]],
u'millis': 0,
u'n': 0,
u'nscanned': 0,
u'nscannedObjects': 0,
u'oldPlan': {u'cursor': u'BtreeCursor contact_1',
u'indexBounds': [[{u'contact': {u'postcode': {u'$lt': 100009}}},
{u'contact': {u'postcode': {u'$lt': 100009}}}]]}}
>>> db.users.find({"contact.postcode":{"$lt":100009}}).explain()
{u'allPlans': [{u'cursor': u'BasicCursor', u'indexBounds': []}],
u'cursor': u'BasicCursor',
u'indexBounds': [],
u'millis': 0,
u'n': 8,
u'nscanned': 30,
u'nscannedObjects': 30,
u'oldPlan': {u'cursor': u'BasicCursor', u'indexBounds': []}}
4. Compound Keys Index
创建复合索引也很简单 (1: ascending; -1: descending)。
> db.users.ensureIndex({name:1, age:-1})
> db.system.indexes.find()
{ "name" : "_id_", "ns" : "blog.users", "key" : { "_id" : 1 } }
{ "_id" : ..., "ns" : "blog.users", "key" : { "name" : 1 }, "name" : "name_1" }
{ "_id" : ..., "ns" : "blog.users", "key" : { "age" : 1 }, "name" : "age_1" }
{ "_id" : ..., "ns" : "blog.users", "key" : { "contact" : 1 }, "name" : "contact_1" }
{ "_id" : ..., "ns" : "blog.users", "key" : { "name" : 1, "age" : -1 }, "name" : "name_1_age_-1" }
> db.users.find({age:{$lt:25}, name:"user2"}, {name:1, age:1})
{ "_id" : ObjectId("4c4a8edeeb257107735eb80e"), "name" : "user2", "age" : 22 }
> db.users.find({age:{$lt:25}, name:"user2"}, {name:1, age:1}).explain()
{
"cursor" : "BtreeCursor name_1_age_-1",
"nscanned" : 1,
"nscannedObjects" : 1,
"n" : 1,
"millis" : 0,
"indexBounds" : [
[
{
"name" : "user2",
"age" : 25
},
{
"name" : "user2",
"age" : -1.7976931348623157e+308
}
]
]
}
复合索引同样可用于局部属性的搜索,但必须依照索引字段顺序。比如创建索引字段顺序 "a,b,c",那么仅对 "a,b,c"、"a,b"、"a" 查询有效,而对 "b,c" 之类的组合无效。
> db.users.dropIndex({name:1})
{ "nIndexesWas" : 5, "ok" : true }
> db.users.dropIndex({age:1})
{ "nIndexesWas" : 4, "ok" : true }
> db.users.dropIndex({contact:1})
{ "nIndexesWas" : 3, "ok" : true }
> db.system.indexes.find()
{ "name" : "_id_", "ns" : "blog.users", "key" : { "_id" : 1 } }
{ "_id" : ObjectId("4c4a9...b82e"), "ns" : "blog.users", "key" : { "name" : 1, "age" : -1 }, "name" : "name_1_age_-1" }
> db.users.find({name:"user12"}).explain() // 索引有效
{
"cursor" : "BtreeCursor name_1_age_-1",
"nscanned" : 1,
"nscannedObjects" : 1,
"n" : 1,
"millis" : 0,
"indexBounds" : [
[
{
"name" : "user12",
"age" : {
"$maxElement" : 1
}
},
{
"name" : "user12",
"age" : {
"$minElement" : 1
}
}
]
]
}
> db.users.find({age:18}).explain() // 索引无效
{
"cursor" : "BasicCursor",
"nscanned" : 30,
"nscannedObjects" : 30,
"n" : 0,
"millis" : 0,
"indexBounds" : [ ]
}
5. Unique Index
只需在 ensureIndex 命令中指定 unique 即可创建唯一索引。
> db.users.ensureIndex({name:1}, {unique:true})
> db.system.indexes.find()
{ "name" : "_id_", "ns" : "blog.users", "key" : { "_id" : 1 } }
{ "_id" : ..., "ns" : "blog.users", "key" : { "name" : 1, "age" : -1 }, "name" : "name_1_age_-1" }
{ "_id" : ..., "ns" : "blog.users", "key" : { "name" : 1 }, "name" : "name_1", "unique" : true }
> db.users.insert({name:"user1"})
E11000 duplicate key error index: blog.users.$name_1 dup key: { : "user1" }
如果创建唯一索引前已经有重复文档,那么可以用 dropDups 删除多余的数据。
> db.users.dropIndexes()
{
"nIndexesWas" : 3,
"msg" : "non-_id indexes dropped for collection",
"ok" : true
}
> db.users.insert({name:"user1"})
> db.users.find({name:"user1"}, {name:1})
{ "_id" : ObjectId("4c4a9573eb257107735eb831"), "name" : "user1" }
{ "_id" : ObjectId("4c4a8edeeb257107735eb80d"), "name" : "user1" }
> db.users.ensureIndex({name:1}, {unique:true, dropDups:true})
E11000 duplicate key error index: blog.users.$name_1 dup key: { : "user1" }
> db.users.find({name:"user1"}, {name:1})
{ "_id" : ObjectId("4c4a9573eb257107735eb831"), "name" : "user1" }
6. Multikeys
对于数组类型属性,会自动索引全部数组元素。
> db.users.dropIndexes()
{
"nIndexesWas" : 1,
"msg" : "non-_id indexes dropped for collection",
"ok" : true
}
> db.users.ensureIndex({"contact.address":1})
> db.users.find({"contact.address":"address2_13"}).explain()
{
"cursor" : "BtreeCursor contact.address_1",
"nscanned" : 1,
"nscannedObjects" : 1,
"n" : 1,
"millis" : 0,
"indexBounds" : [
[
{
"contact.address" : "address2_13"
},
{
"contact.address" : "address2_13"
}
]
]
}
7. hint
hint 命令可以强制使用某个索引。
> db.users.dropIndexes()
{
"nIndexesWas" : 2,
"msg" : "non-_id indexes dropped for collection",
"ok" : true
}
> db.users.ensureIndex({name:1, age:1})
> db.users.find({age:{$lt:30}}).explain()
{
"cursor" : "BasicCursor",
"nscanned" : 30,
"nscannedObjects" : 30,
"n" : 9,
"millis" : 0,
"indexBounds" : [ ]
}
> db.users.find({age:{$lt:30}}).hint({name:1, age:1}).explain()
{
"cursor" : "BtreeCursor name_1_age_1",
"nscanned" : 30,
"nscannedObjects" : 9,
"n" : 9,
"millis" : 0,
"indexBounds" : [
[
{
"name" : {
"$minElement" : 1
},
"age" : -1.7976931348623157e+308
},
{
"name" : {
"$maxElement" : 1
},
"age" : 30
}
]
]
}
注意 Python 代码中写法。
>>> db.users.find({"age":{"$lt":30}}).hint([("name", ASCENDING), ("age", ASCENDING)]).explain()
{u'cursor': u'BtreeCursor name_1_age_1',
u'indexBounds': [[{u'age': -1.7976931348623157e+308,
u'name': {u'$minElement': 1}},
{u'age': 30, u'name': {u'$maxElement': 1}}]],
u'millis': 0,
u'n': 9,
u'nscanned': 30,
u'nscannedObjects': 9}
8. totalIndexSize
MongoDB 会将索引数据载入内存,以提高查询速度。我们可以用 totalIndexSize 获取全部索引数据大小。
> db.users.totalIndexSize()
16384