【发布时间】：2020-03-04 02:30:22
【问题描述】：
我有一个水果车集合，数百万行（更改名称以保护有罪者。）
这些购物车中的每一个都有一个所有者、一个品牌和一个型号，以及关于购物车是否包含某些类型的水果的几个 Y 或 N 字段。
我想生成一个模型列表，并计算每个模型有多少个装有特定类型水果的推车。
所以数据会是这样的：
fruit.cart.insert(
{ 'owner' => 'Fred', 'make' => 'Toshiba', 'model' => 'fruitmaster 9000', 'have_apples' => 'Y', 'have_grapes' => 'N', 'have_peaches' => 'Y', ... },
{ 'owner' => 'Wilma', 'make' => 'Toshiba', 'model' => 'fruitmaster 9000', 'have_apples' => 'Y', 'have_grapes' => 'N', 'have_peaches' => 'N', ... },
{ 'owner' => 'Betty', 'make' => 'Toshiba', 'model' => 'fruitmaster 9000', 'have_apples' => 'N', 'have_grapes' => 'Y', 'have_peaches' => 'Y', ... },
{ 'owner' => 'Barney', 'make' => 'Toshiba', 'model' => 'fruitmaster X5', 'have_apples' => 'N', 'have_grapes' => 'N', 'have_peaches' => 'Y', ... },
{ 'owner' => 'Fred', 'make' => 'Honda', 'model' => 'T-1000', 'have_apples' => 'Y', 'have_grapes' => 'Y', 'have_peaches' => 'Y', ... },
{ 'owner' => 'Wilma', 'make' => 'Honda', 'model' => 'T-1000', 'have_apples' => 'N', 'have_grapes' => 'N', 'have_peaches' => 'N', ... },
我希望输出是：
{ 'make' => 'Toshiba', 'model' => 'fruitmaster 9000', 'count' => 3, 'apples_count' => 2, 'grapes_count' => 1, 'peaches_count' => 2, ... },
{ 'make' => 'Toshiba', 'model' => 'fruitmaster X5', 'count' => 1, 'apples_count' => 0, 'grapes_count' => 0, 'peaches_count' => 1, ... },
{ 'make' => 'Honda', 'model' => 'T-1000', 'count' => 2, 'apples_count' => 1, 'grapes_count' => 1, 'peaches_count' => 1, ... },
这是我的汇总查询：
{ '$group' =>
{
_id => { 'model' => '$model', },
model => { '$first' => '$model' },
make => { '$first' => '$make' },
count => { '$sum' => 1 },
oranges_count => { '$sum' => { '$cond' => [ { '$eq' => [ '$have_oranges', 'Y' ] }, 1, 0 ] }, },
grapes_count => { '$sum' => { '$cond' => [ { '$eq' => [ '$have_grapes', 'Y' ] }, 1, 0 ] }, },
peaches_count => { '$sum' => { '$cond' => [ { '$eq' => [ '$have_peaches', 'Y' ] }, 1, 0 ] }, },
apples_count => { '$sum' => { '$cond' => [ { '$eq' => [ '$have_apples', 'Y' ] }, 1, 0 ] }, },
pears_count => { '$sum' => { '$cond' => [ { '$eq' => [ '$have_pears', 'Y' ] }, 1, 0 ] }, },
},
},
带索引：
db.cart.createIndex (
{
'have_pears' : 1,
'have_apples' : 1,
'have_oranges' : 1,
'have_grapes' : 1,
'have_peaches' : 1,
'make' : 1,
'model' : 1,
}, { background : true, name : 'fruit_counts' } );
如果我在不提供索引提示的情况下运行它，那么它就不会使用它并且查询需要永远：
...
'queryPlanner' => {
'indexFilterSet' => bless( do{\(my $o = 0)}, 'boolean' ),
'namespace' => 'fruit.cart',
'parsedQuery' => {},
'plannerVersion' => 1,
'rejectedPlans' => [],
'winningPlan' => {
'direction' => 'forward',
'stage' => 'COLLSCAN'
}
}
...
提示很快：
{
'ok' => '1',
'stages' => [
{
'$cursor' => {
'fields' => {
'_id' => 0,
'have_peaches' => 1,
'have_apples' => 1,
'have_oranges' => 1,
'have_grapes' => 1,
'have_pears' => 1,
'model' => 1,
'make' => 1
},
'query' => {},
'queryPlanner' => {
'indexFilterSet' => bless( do{\(my $o = 0)}, 'boolean' ),
'namespace' => 'fruit.cart',
'parsedQuery' => {},
'plannerVersion' => 1,
'rejectedPlans' => [],
'winningPlan' => {
'inputStage' => {
'direction' => 'forward',
'indexBounds' => {
'have_peaches' => [ '[MinKey, MaxKey]' ],
'have_apples' => [ '[MinKey, MaxKey]' ],
'have_oranges' => [ '[MinKey, MaxKey]' ],
'have_grapes' => [ '[MinKey, MaxKey]' ],
'have_pears' => [ '[MinKey, MaxKey]' ],
'model' => [ '[MinKey, MaxKey]' ],
'make' => [ '[MinKey, MaxKey]' ]
},
'indexName' => 'fruit_counts',
'indexVersion' => 2,
'isMultiKey' => $VAR1->[0]{'stages'}[0]{'$cursor'}{'queryPlanner'}{'indexFilterSet'},
'isPartial' => $VAR1->[0]{'stages'}[0]{'$cursor'}{'queryPlanner'}{'indexFilterSet'},
'isSparse' => $VAR1->[0]{'stages'}[0]{'$cursor'}{'queryPlanner'}{'indexFilterSet'},
'isUnique' => $VAR1->[0]{'stages'}[0]{'$cursor'}{'queryPlanner'}{'indexFilterSet'},
'keyPattern' => {
'have_peaches' => '1',
'have_apples' => '1',
'have_oranges' => '1',
'have_grapes' => '1',
'have_pears' => '1',
'model' => '1',
'make' => '1'
},
'multiKeyPaths' => {
'have_peaches' => [],
'have_apples' => [],
'have_oranges' => [],
'have_grapes' => [],
'have_pears' => [],
'model' => [],
'make' => []
},
'stage' => 'IXSCAN'
},
'stage' => 'PROJECTION',
'transformBy' => {
'_id' => 0,
'have_peaches' => 1,
'have_apples' => 1,
'have_oranges' => 1,
'have_grapes' => 1,
'have_pears' => 1,
'model' => 1,
'make' => 1
}
}
}
}
},
...
没有提示的documentdb：
'queryPlanner' => {
'namespace' => 'fruit.cart',
'plannerVersion' => 1,
'winningPlan' => {
'inputStage' => {
'inputStage' => {
'stage' => 'COLLSCAN'
},
'stage' => 'SORT'
},
'stage' => 'SORT_AGGREGATE'
}
},
带有提示的文档数据库：
MongoDB::DatabaseError: Cannot use Hint for this Query. Index is multi key index or sparse index and query is not optimized to use this index.
DOH！！！
那么我该怎么做才能让它甚至不需要提示，然后让它在 DocumentDB 中工作？
【问题讨论】：
-
DocumentDB 有一个来自 MongoDB 的 different server implementation，因此某些功能和行为会有所不同。由于您的最终目标是让此查询在 DocumentDB 中工作，因此 MongoDB 服务器行为不会成为有用的指南。
-
以
$group阶段开始的聚合查询将需要完整的集合扫描，但看起来您正在尝试为$group中使用的字段创建覆盖索引。 DocumentDB 显然不喜欢提示多键索引，因此索引字段之一必须是（或曾经是）数组。您可以尝试删除并重新创建索引以查看是否清除了多键标志。
标签： indexing aggregate aws-documentdb