MongoDB索引
数据库中的索引就是用来提高查询操作的性能，但是会影响插入、更新和删除的效率，因为数据库不仅要执行这些操作，还要负责索引的更新。通过建立索引，影响一部分插入、更新和删除的效率，但是能大大挺高查询的效率，这个还是很值得的。
为了开始后面的操作，首先通过MongoDB shell插入一些测试数据。
1 for(var i=0;i db.school.students.ensureIndex({"name": 1}, {"unique": true})
2 >
查看索引：
1 > db.school.students.getIndexes()
2 [
3 {
4 "v" : 1,
5 "key" : {
6 "_id" : 1
7 },
8 "ns" : "test.school.students",
9 "name" : "_id_"
10 },
11 {
12 "v" : 1,
13 "key" : {
14 "name" : 1
15 },
16 "unique" : true,
17 "ns" : "test.school.students",
18 "name" : "name_1"
19 }
20 ]
21 >
删除索引：
1 > db.school.students.dropIndex("name_1")
2 { "nIndexesWas" : 2, "ok" : 1 }
3 >
索引名称：默认情况下，索引的名称是"键_值_键_值…"的形式，当键的数量很多的时候，索引的名字就会很长。
所以，在创建索引的时候，可以通过"name"参数自定义索引的名字。
1 > db.school.students.ensureIndex({"name": 1}, {"name": "myIndex"})
2 >
explain()和hint()
通过explain()可以得到很多跟find相关的信息，对索引的分析很有帮助。
当有多个可以使用的索引时，MongoDB会自动选择最优索引，但是我们可以通过hint()操作选择我们想要使用的索引。
下面来看看没有索引时explain()的输出：
1 > db.school.students.find({"name": "Will5"}).explain()
2 {
3 "cursor" : "BasicCursor",
4 "isMultiKey" : false,
5 "n" : 1,
6 "nscannedObjects" : 6,
7 "nscanned" : 6,
8 "nscannedObjectsAllPlans" : 6,
9 "nscannedAllPlans" : 6,
10 "scanAndOrder" : false,
11 "indexOnly" : false,
12 "nYields" : 0,
13 "nChunkSkips" : 0,
14 "millis" : 0,
15 "indexBounds" : {
16
17 },
18 "server" : "××××:27017"
19 }
20 >
分析：下面选择了几个我们比较关心的字段
[*]cursor：BasicCursor表示是full Collection scan，即没有索引的全表扫描
[*]n：满足查询条件的文档数量
[*]nscannedObjects：总共扫描的文档的数量
[*]nscanned：总共扫描的索引节点的数量
[*]scanAndOrder：false表示，MongoDB现有索引下文档的顺序来返回排序结果；true表示，MongoDB需要在得到查询结果后重新排序
[*]millis：完成查询需要的毫秒数
添加索引，再次检查explain()的输出：
1 > db.school.students.ensureIndex({"name": 1}, {"unique": true})
2 > db.school.students.find({"name": "Will5"}).explain()
3 {
4 "cursor" : "BtreeCursor name_1",
5 "isMultiKey" : false,
6 "n" : 1,
7 "nscannedObjects" : 1,
8 "nscanned" : 1,
9 "nscannedObjectsAllPlans" : 1,
10 "nscannedAllPlans" : 1,
11 "scanAndOrder" : false,
12 "indexOnly" : false,
13 "nYields" : 0,
14 "nChunkSkips" : 0,
15 "millis" : 0,
16 "indexBounds" : {
17 "name" : [
18 [
19 "Will5",
20 "Will5"
21 ]
22 ]
23 },
24 "server" : "××××:27017"
25 }
26 >
组合索引
单键索引还是比较简单的，当使用组合索引的时候，就要多考虑一些了。自己也不确定能否总结的很好，如果错误，希望大家指出、讨论。
索引建立可能有多种方式，我们的目标就是减少"nscanned"（当然也有特例，请参照"索引和排序"）。
下面分析基于前面生成的数据来分析一下组合索引，假设我们要查询年龄大于等于23的女学生。
[*]
使用"age_1"索引的输出如下
1 > db.school.students.find({"age":{"$gte":23}, "gender":"Female"}).hint("age_1").explain()
2 {
3 "cursor" : "BtreeCursor age_1",
4 "isMultiKey" : false,
5 "n" : 2,
6 "nscannedObjects" : 4,
7 "nscanned" : 4,
8 "nscannedObjectsAllPlans" : 4,
9 "nscannedAllPlans" : 4,
10 "scanAndOrder" : false,
11 "indexOnly" : false,
12 "nYields" : 0,
13 "nChunkSkips" : 0,
14 "millis" : 0,
15 "indexBounds" : {
16 "age" : [
17 [
18 23,
19 1.7976931348623157e+308
20 ]
21 ]
22 },
23 "server" : "××××:27017"
24 }
25 >
索引的分析：
Index
Documents
Result
age:20
{ "name" : "Will1", "gender" : "Female", "age" : 20 }
"n" : 2
age:20
{ "name" : "Will5", "gender" : "Male", "age" : 20 }
"nscannedObjects" : 4
age:20
{ "name" : "Will6", "gender" : "Female", "age" : 20 }
"nscanned" : 4
age:21
{ "name" : "Will4", "gender" : "Male", "age" : 21 }
age:21
{ "name" : "Will8", "gender" : "Male", "age" : 21 }
age:22
{ "name" : "Will0", "gender" : "Female", "age" : 22 }
age:23
{ "name" : "Will3", "gender" : "Male", "age" : 23 }
age:24
{ "name" : "Will2", "gender" : "Male", "age" : 24 }
age:24
{ "name" : "Will7", "gender" : "Female", "age" : 24 }
age:24
{ "name" : "Will9", "gender" : "Female", "age" : 24 }
[*]
使用"age_1_gender_1"索引的输出如下
1 > db.school.students.find({"age":{"$gte":23}, "gender":"Female"}).hint("age_1_gender_1").explain()
2 {
3 "cursor" : "BtreeCursor age_1_gender_1",
4 "isMultiKey" : false,
5 "n" : 2,
6 "nscannedObjects" : 2,
7 "nscanned" : 4,
8 "nscannedObjectsAllPlans" : 2,
9 "nscannedAllPlans" : 4,
10 "scanAndOrder" : false,
11 "indexOnly" : false,
12 "nYields" : 0,
13 "nChunkSkips" : 0,
14 "millis" : 0,
15 "indexBounds" : {
16 "age" : [
17 [
18 23,
19 1.7976931348623157e+308
20 ]
21 ],
22 "gender" : [
23 [
24 "Female",
25 "Female"
26 ]
27 ]
28 },
29 "server" : "××××:27017"
30 }
31 >
索引的分析：
Index
Documents
Result
age:20, gender:Female
{ "name" : "Will1", "gender" : "Female", "age" : 20 }
"n" : 2
age:20, gender:Female
{ "name" : "Will6", "gender" : "Female", "age" : 20 }
"nscannedObjects" : 2
age:20, gender:Male
{ "name" : "Will5", "gender" : "Male", "age" : 20 }
"nscanned" : 4
age:21, gender:Male
{ "name" : "Will4", "gender" : "Male", "age" : 21 }
age:21, gender:Male
{ "name" : "Will8", "gender" : "Male", "age" : 21 }
age:22, gender:Female
{ "name" : "Will0", "gender" : "Female", "age" : 22}
age:23, gender:Male
{ "name" : "Will3", "gender" : "Male", "age" : 23 }
age:24, gender:Female
{ "name" : "Will7", "gender" : "Female", "age" : 24 }
age:24, gender:Female
{ "name" : "Will9", "gender" : "Female", "age" : 24 }
age:24, gender:Male
{ "name" : "Will2", "gender" : "Male", "age" : 24 }
[*]
使用"gender_1_age_1"索引的输出如下
1 > db.school.students.find({"age":{"$gte":23}, "gender":"Female"}).hint("gender_1_age_1").explain()
2 {
3 "cursor" : "BtreeCursor gender_1_age_1",
4 "isMultiKey" : false,
5 "n" : 2,
6 "nscannedObjects" : 2,
7 "nscanned" : 2,
8 "nscannedObjectsAllPlans" : 2,
9 "nscannedAllPlans" : 2,
10 "scanAndOrder" : false,
11 "indexOnly" : false,
12 "nYields" : 0,
13 "nChunkSkips" : 0,
14 "millis" : 0,
15 "indexBounds" : {
16 "gender" : [
17 [
18 "Female",
19 "Female"
20 ]
21 ],
22 "age" : [
23 [
24 23,
25 1.7976931348623157e+308
26 ]
27 ]
28 },
29 "server" : "××××:27017"
30 }
31 >
索引的分析：
Index
Documents
Result
gender:Female, age:20
{ "name" : "Will1", "gender" : "Female", "age" : 20 }
"n" : 2
gender:Female, age:20
{ "name" : "Will6", "gender" : "Female", "age" : 20 }
"nscannedObjects" : 2
gender:Female, age:22
{ "name" : "Will0", "gender" : "Female", "age" : 22 }
"nscanned" : 2
gender:Female, age:24
{ "name" : "Will7", "gender" : "Female", "age" : 24 }
gender:Female, age:24
{ "name" : "Will9", "gender" : "Female", "age" : 24 }
gender:Male, age:20
{ "name" : "Will5", "gender" : "Male", "age" : 20 }
gender:Male, age:21
{ "name" : "Will4", "gender" : "Male", "age" : 21 }
gender:Male, age:21
{ "name" : "Will8", "gender" : "Male", "age" : 21 }
gender:Male, age:23
{ "name" : "Will3", "gender" : "Male", "age" : 23 }
gender:Male, age:24
{ "name" : "Will2", "gender" : "Male", "age" : 24 }
通过上面的例子可以看出，在使用组合索引的时候还是要考虑很多东西的，所以可以结合explain()来进行分析。
索引选择机制
由于我们前面创建了三个索引，下面我们直接使用默认查询。
1 > db.school.students.find({"age":{"$gte":23}, "gender":"Female"}).explain()
2 {
3 "cursor" : "BtreeCursor gender_1_age_1",
4 "isMultiKey" : false,
5 "n" : 2,
6 "nscannedObjects" : 2,
7 "nscanned" : 2,
8 "nscannedObjectsAllPlans" : 2,
9 "nscannedAllPlans" : 2,
10 "scanAndOrder" : false,
11 "indexOnly" : false,
12 "nYields" : 0,
13 "nChunkSkips" : 0,
14 "millis" : 0,
15 "indexBounds" : {
16 "gender" : [
17 [
18 "Female",
19 "Female"
20 ]
21 ],
22 "age" : [
23 [
24 23,
25 1.7976931348623157e+308
26 ]
27 ]
28 },
29 "server" : "××××:27017"
30 }
31 >
存在多条索引的情况下，MongoDB首选nscanned值最低的索引。
索引和排序
基于上面的例子，我们加上对"name"的排序操作。这时，我们可以看到"scanAndOrder"变成了"true"。
1 > db.school.students.find({"age":{"$gte":23}, "gender":"Female"}).sort({"name":1}).explain()
2 {
3 "cursor" : "BtreeCursor gender_1_age_1",
4 "isMultiKey" : false,
5 "n" : 2,
6 "nscannedObjects" : 2,
7 "nscanned" : 2,
8 "nscannedObjectsAllPlans" : 7,
9 "nscannedAllPlans" : 9,
10 "scanAndOrder" : true,
11 "indexOnly" : false,
12 "nYields" : 0,
13 "nChunkSkips" : 0,
14 "millis" : 0,
15 "indexBounds" : {
16 "gender" : [
17 [
18 "Female",
19 "Female"
20 ]
21 ],
22 "age" : [
23 [
24 23,
25 1.7976931348623157e+308
26 ]
27 ]
28 },
29 "server" : "××××:27017"
30 }
在这个例子中，"nscanned"是最小的，所以这个方案是查询效率最高的。但是，我们要注意一下"scanAndOrder"，根据MongoDB文档的解释，查询结果的排序不能利用现有的索引，MongoDB会把find找到的结果放入内存重新排序。这样的话，如果数据量很大，会对性能产生很大的影响。
最好的办法是利用索引来进行排序。
在这种情况下，就要加入一个"name"的索引，同时在find操作时使用hint来指定索引方式，因为默认情况MongoDB会选择"nscanned"最小的方式。
1 > db.school.students.ensureIndex({"gender":1,"name":1})
2 > db.school.students.find({"age":{"$gte":23}, "gender":"Female"}).sort({"name":1}).hint("gender_1_name_1").explain()
3 {
4 "cursor" : "BtreeCursor gender_1_name_1",
5 "isMultiKey" : false,
6 "n" : 2,
7 "nscannedObjects" : 5,
8 "nscanned" : 5,
9 "nscannedObjectsAllPlans" : 5,
10 "nscannedAllPlans" : 5,
11 "scanAndOrder" : false,
12 "indexOnly" : false,
13 "nYields" : 0,
14 "nChunkSkips" : 0,
15 "millis" : 0,
16 "indexBounds" : {
17 "gender" : [
18 [
19 "Female",
20 "Female"
21 ]
22 ],
23 "name" : [
24 [
25 {
26 "$minElement" : 1
27 },
28 {
29 "$maxElement" : 1
30 }
31 ]
32 ]
33 },
34 "server" : "xxxx:27017"
35 }
36 >
通过这种方式，就可以利用索引的排序来避免"scanAndOrder"为true的情况。但是再看看上面的方式，似乎可以进一步优化，虽然不能减少"nscanned"，但是可以减少"nscannedObjects"。
1 > db.school.students.ensureIndex({"gender":1,"name":1,"age":1})
2 > db.school.students.find({"age":{"$gte":23}, "gender":"Female"}).sort({"name":1}).hint("gender_1_name_1_age_1").explain()
3 {
4 "cursor" : "BtreeCursor gender_1_name_1_age_1",
5 "isMultiKey" : false,
6 "n" : 2,
7 "nscannedObjects" : 2,
8 "nscanned" : 5,
9 "nscannedObjectsAllPlans" : 2,
10 "nscannedAllPlans" : 5,
11 "scanAndOrder" : false,
12 "indexOnly" : false,
13 "nYields" : 0,
14 "nChunkSkips" : 0,
15 "millis" : 0,
16 "indexBounds" : {
17 "gender" : [
18 [
19 "Female",
20 "Female"
21 ]
22 ],
23 "name" : [
24 [
25 {
26 "$minElement" : 1
27 },
28 {
29 "$maxElement" : 1
30 }
31 ]
32 ],
33 "age" : [
34 [
35 23,
36 1.7976931348623157e+308
37 ]
38 ]
39 },
40 "server" : "xxxx:27017"
41 }
42 >
总结
MongoDB中，索引还有很多东西，本文只是通过一些例子来介绍了索引的使用，以及组合索引的简单分析
Ps: 本文中所有例子中的命令都可以参考以下链接
http://files.iyunv.com/wilber2013/index.js
页:
[1]