我正在尝试使用 MongoDB 的 B-tree 索引构建自定义地理空间索引，因为我发现 Mongo 的本机实现限制了我自己的情况。为了完成使用复合索引有效搜索 Mongo 的地理空间查询，我需要按location.locIndexKey具有多个范围的字段以及其他字段进行过滤。
到目前为止，我能想出的支持这种查询的唯一解决方案是使用 Mongo 的$or运算符。然而，这表现得很糟糕，因为它是一个or查询，Mongo 必须一次又一次地检查索引上的相同键。为了克服这种低效率，我需要一种方法让 Mongo 在该字段上使用多个索引边界，而不是使用or短语为查询中定义的每个边界复制查询。
这是我的查询：
db.users.find({
"gender":2,
"preferences.feed.gender":1,
"age":{"$gte":18,"$lte":55},
"feedPrefChangeDay":{"$gte":1553461200,"$lte":1554066000},
"$or":[{"location.locIndexKey":{"$gte":NumberLong(1493233547543052300),"$lte":NumberLong(1493242343636074500)}},{"location.locIndexKey":{"$gte":NumberLong(1493242343636074500),"$lte":NumberLong(1493251139729096700)}},{"location.locIndexKey":{"$gte":NumberLong(1493287011295953000),"$lte":NumberLong(1493287148734906400)}}]
}).limit(20);
如您所见，为了在 field 上表达多个范围location.locIndexKey，我不得不使用$or运算符。这是查询计划器执行统计信息的缩短版本：
{
"executionSuccess" : true,
"nReturned" : 0,
"executionTimeMillis" : 17762,
"totalKeysExamined" : 196192,
"totalDocsExamined" : 0,
"executionStages" : {
"stage" : "LIMIT",
"nReturned" : 0,
"executionTimeMillisEstimate" : 351,
"works" : 196193,
"advanced" : 0,
"needTime" : 196191,
"needYield" : 0,
"saveState" : 19944,
"restoreState" : 19944,
"isEOF" : 1,
"invalidates" : 0,
"limitAmount" : 20,
"inputStage" : {
"stage" : "FETCH",
"nReturned" : 0,
"executionTimeMillisEstimate" : 351,
"works" : 196193,
"advanced" : 0,
"needTime" : 196191,
"needYield" : 0,
"saveState" : 19944,
"restoreState" : 19944,
"isEOF" : 1,
"invalidates" : 0,
"docsExamined" : 0,
"alreadyHasObj" : 0,
"inputStage" : {
"stage" : "OR",
"nReturned" : 0,
"executionTimeMillisEstimate" : 351,
"works" : 196192,
"advanced" : 0,
"needTime" : 196191,
"needYield" : 0,
"saveState" : 19944,
"restoreState" : 19944,
"isEOF" : 1,
"invalidates" : 0,
"dupsTested" : 0,
"dupsDropped" : 0,
"recordIdsForgotten" : 0,
"inputStages" : [
{
"stage" : "IXSCAN",
"nReturned" : 0,
"executionTimeMillisEstimate" : 10,
"works" : 4534,
"advanced" : 0,
"needTime" : 4533,
"needYield" : 0,
"saveState" : 19944,
"restoreState" : 19944,
"isEOF" : 1,
"invalidates" : 0,
"keyPattern" : {
"gender" : 1.0,
"preferences.feed.gender" : 1.0,
"age" : 1.0,
"feedPrefChangeDay" : 1.0,
"location.locIndexKey" : 1.0
},
"indexName" : "gender_1_preferences.feed.gender_1_age_1_feedPrefChangeDay_1_location.locIndexKey_1",
"isMultiKey" : false,
"multiKeyPaths" : {
"gender" : [],
"preferences.feed.gender" : [],
"age" : [],
"feedPrefChangeDay" : [],
"location.locIndexKey" : []
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"gender" : [
"[2.0, 2.0]"
],
"preferences.feed.gender" : [
"[1.0, 1.0]"
],
"age" : [
"[18.0, 55.0]"
],
"feedPrefChangeDay" : [
"[1553461200.0, 1554066000.0]"
],
"location.locIndexKey" : [
"[1493569998101151700, 1493572197124407300]"
]
},
"keysExamined" : 4534,
"seeks" : 4534,
"dupsTested" : 0,
"dupsDropped" : 0,
"seenInvalidated" : 0
},
{
"stage" : "IXSCAN",
"nReturned" : 0,
"executionTimeMillisEstimate" : 0,
"works" : 4534,
"advanced" : 0,
"needTime" : 4533,
"needYield" : 0,
"saveState" : 19944,
"restoreState" : 19944,
"isEOF" : 1,
"invalidates" : 0,
"keyPattern" : {
"gender" : 1.0,
"preferences.feed.gender" : 1.0,
"age" : 1.0,
"feedPrefChangeDay" : 1.0,
"location.locIndexKey" : 1.0
},
"indexName" : "gender_1_preferences.feed.gender_1_age_1_feedPrefChangeDay_1_location.locIndexKey_1",
"isMultiKey" : false,
"multiKeyPaths" : {
"gender" : [],
"preferences.feed.gender" : [],
"age" : [],
"feedPrefChangeDay" : [],
"location.locIndexKey" : []
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"gender" : [
"[2.0, 2.0]"
],
"preferences.feed.gender" : [
"[1.0, 1.0]"
],
"age" : [
"[18.0, 55.0]"
],
"feedPrefChangeDay" : [
"[1553461200.0, 1554066000.0]"
],
"location.locIndexKey" : [
"[1493587581697261600, 1493587590287196200]"
]
},
"keysExamined" : 4534,
"seeks" : 4534,
"dupsTested" : 0,
"dupsDropped" : 0,
"seenInvalidated" : 0
},
{
"stage" : "IXSCAN",
"nReturned" : 0,
"executionTimeMillisEstimate" : 0,
"works" : 4534,
"advanced" : 0,
"needTime" : 4533,
"needYield" : 0,
"saveState" : 19944,
"restoreState" : 19944,
"isEOF" : 1,
"invalidates" : 0,
"keyPattern" : {
"gender" : 1.0,
"preferences.feed.gender" : 1.0,
"age" : 1.0,
"feedPrefChangeDay" : 1.0,
"location.locIndexKey" : 1.0
},
"indexName" : "gender_1_preferences.feed.gender_1_age_1_feedPrefChangeDay_1_location.locIndexKey_1",
"isMultiKey" : false,
"multiKeyPaths" : {
"gender" : [],
"preferences.feed.gender" : [],
"age" : [],
"feedPrefChangeDay" : [],
"location.locIndexKey" : []
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"gender" : [
"[2.0, 2.0]"
],
"preferences.feed.gender" : [
"[1.0, 1.0]"
],
"age" : [
"[18.0, 55.0]"
],
"feedPrefChangeDay" : [
"[1553461200.0, 1554066000.0]"
],
"location.locIndexKey" : [
"[1493981215449940000, 1493990011542962200]"
]
},
"keysExamined" : 4534,
"seeks" : 4534,
"dupsTested" : 0,
"dupsDropped" : 0,
"seenInvalidated" : 0
}
如果您indexBounds在上面进行检查，您将看到每个范围 forlocation.locIndexKey都应用于单个查询并与or. 但是，如果我选择使用 Mongo 的本地地理空间运算符运行相同的查询$geoWithin：
db.users.find({
"gender" : 2.0,
"preferences.feed.gender" : 1.0,
"age" : {
"$gte" : 18.0,
"$lte" : 55.0
},
"feedPrefChangeDay" : {
"$gte" : 1553461200.0,
"$lte" : 1554066000.0
},
"location.loc" : {
"$geoWithin" : {
"$centerSphere" : [
[
0.0,
0.0
],
0.00784806152880239
]
}
}
}).limit(20);
我从查询规划器得到以下响应：
{
"executionSuccess" : true,
"nReturned" : 0,
"executionTimeMillis" : 7,
"totalKeysExamined" : 4506,
"totalDocsExamined" : 0,
"executionStages" : {
"stage" : "LIMIT",
"nReturned" : 0,
"executionTimeMillisEstimate" : 0,
"works" : 4506,
"advanced" : 0,
"needTime" : 4505,
"needYield" : 0,
"saveState" : 35,
"restoreState" : 35,
"isEOF" : 1,
"invalidates" : 0,
"limitAmount" : 20,
"inputStage" : {
"stage" : "FETCH",
"filter" : {
"location.loc" : {
"$geoWithin" : {
"$centerSphere" : [
[
0.0,
0.0
],
0.00784806152880239
]
}
}
},
"nReturned" : 0,
"executionTimeMillisEstimate" : 0,
"works" : 4506,
"advanced" : 0,
"needTime" : 4505,
"needYield" : 0,
"saveState" : 35,
"restoreState" : 35,
"isEOF" : 1,
"invalidates" : 0,
"docsExamined" : 0,
"alreadyHasObj" : 0,
"inputStage" : {
"stage" : "IXSCAN",
"nReturned" : 0,
"executionTimeMillisEstimate" : 0,
"works" : 4506,
"advanced" : 0,
"needTime" : 4505,
"needYield" : 0,
"saveState" : 35,
"restoreState" : 35,
"isEOF" : 1,
"invalidates" : 0,
"keyPattern" : {
"gender" : 1.0,
"preferences.feed.gender" : 1.0,
"age" : 1.0,
"feedPrefChangeDay" : 1.0,
"location.loc" : "2dsphere"
},
"indexName" : "gender_1_preferences.feed.gender_1_age_1_feedPrefChangeDay_1_location.loc_2dsphere",
"isMultiKey" : false,
"multiKeyPaths" : {
"gender" : [],
"preferences.feed.gender" : [],
"age" : [],
"feedPrefChangeDay" : [],
"location.loc" : []
},
"isUnique" : false,
"isSparse" : false,
"isPartial" : false,
"indexVersion" : 2,
"direction" : "forward",
"indexBounds" : {
"gender" : [
"[2.0, 2.0]"
],
"preferences.feed.gender" : [
"[1.0, 1.0]"
],
"age" : [
"[18.0, 55.0]"
],
"feedPrefChangeDay" : [
"[1553461200.0, 1554066000.0]"
],
"location.loc" : [
"[360287970189639680, 360287970189639680]",
"[378302368699121664, 378302368699121664]",
"[382805968326492160, 382805968326492160]",
"[383931868233334784, 383931868233334784]",
"[384213343210045440, 384213343210045440]",
"[384230935396089856, 384230935396089856]",
"[384235333442600960, 384235333442600960]",
"[384236432954228736, 384236432954228736]",
"[384236432954228737, 384236982710042623]",
"[384266119768178688, 384266119768178688]",
"[384266119768178689, 384274915861200895]",
"[384274915861200897, 384283711954223103]",
"[384283711954223104, 384283711954223104]",
"[384283711954223105, 384318896326311935]",
"[384318896326311937, 384354080698400767]",
"[1080863910568919040, 1080863910568919040]",
"[1134907106097364992, 1134907106097364992]",
"[1148417904979476480, 1148417904979476480]",
"[1151795604700004352, 1151795604700004352]",
"[1152640029630136320, 1152640029630136320]",
"[1152789563211513857, 1152798359304536063]",
"[1152798359304536064, 1152798359304536064]",
"[1152798359304536065, 1152807155397558271]",
"[1152833543676624896, 1152833543676624896]",
"[1152846737816158208, 1152846737816158208]",
"[1152850036351041536, 1152850036351041536]",
"[1152850586106855425, 1152851135862669311]",
"[1152851135862669312, 1152851135862669312]",
"[1152851135862669313, 1152859931955691519]",
"[1152868728048713728, 1152868728048713728]",
"[1152877524141735937, 1152886320234758143]",
"[1152886320234758145, 1152921504606846975]",
"[1152921504606846977, 1152956688978935807]",
"[1152956688978935809, 1152991873351024639]",
"[1152991873351024640, 1152991873351024640]",
"[1152991873351024641, 1152992423106838527]",
"[1152992972862652416, 1152992972862652416]",
"[1152996271397535744, 1152996271397535744]",
"[1153009465537069056, 1153009465537069056]",
"[1153035853816135681, 1153044649909157887]",
"[1153044649909157888, 1153044649909157888]",
"[1153044649909157889, 1153053446002180095]",
"[1153202979583557632, 1153202979583557632]",
"[1154047404513689600, 1154047404513689600]",
"[1157425104234217472, 1157425104234217472]",
"[1170935903116328960, 1170935903116328960]",
"[1224979098644774912, 1224979098644774912]",
"[1921488928515293185, 1921524112887382015]",
"[1921524112887382017, 1921559297259470847]",
"[1921559297259470848, 1921559297259470848]",
"[1921559297259470849, 1921594481631559679]",
"[1921606026503651329, 1921606576259465215]",
"[1921606576259465216, 1921606576259465216]",
"[1921607675771092992, 1921607675771092992]",
"[1921612073817604096, 1921612073817604096]",
"[1921629666003648512, 1921629666003648512]",
"[1921911140980359168, 1921911140980359168]",
"[1923037040887201792, 1923037040887201792]",
"[1927540640514572288, 1927540640514572288]",
"[1945555039024054272, 1945555039024054272]"
]
},
"keysExamined" : 4506,
"seeks" : 4506,
"dupsTested" : 0,
"dupsDropped" : 0,
"seenInvalidated" : 0
}
}
}
}
如您所见，Mongo 利用多个索引边界来完成此查询，并且更加有效。
我相信原始查询效率低下的原因是 Mongo 的查询计划器不检查$or表达式中的内容。我认为理解表达式中只有一个具有多个范围的字段并使用多个索引边界构建查询应该更聪明。可悲的是，事实并非如此。
我的问题：有什么方法可以强制 Mongo 为我的查询使用多个索引范围，以便它与本机地理空间查询一样高效？
任何帮助，将不胜感激。
谢谢！