N1ql 计算不同的关联文档

Question

我有一个存储桶（Couchbase 社区版 6.5），其中包含以下文档：

employee {
    type: "Employee"
}
X {
    type: "X",
    employeeId: string,
    date: string
}
Y {
    type: "Y",
    employeeId: string,
    date: string
}
Z {
    type: "Z",
    employeeId: string,
    date: string
}

我需要获取两个日期之间与每个员工关联的文档总数 (X,Y,Z)。

我编写了以下有效但执行时间较慢的查询：

CREATE INDEX `index_X` ON `bucket`(`type`,`date`, `employeeId`)
WHERE type = "X"
CREATE INDEX `index_Y` ON `bucket`(`type`,`date`, `employeeId`)
WHERE type = "Y"
CREATE INDEX `index_Z` ON `bucket`(`type`,`date`, `employeeId`)
WHERE type = "Z"

SELECT META(employee).id,
       x.totalX,
       y.totalY,
       z.totalZ,
FROM `bucket` employee 
    LEFT JOIN (
    SELECT obj.employeeId,
           COUNT(obj.employeeId) AS totalX
    FROM `bucket` obj
    WHERE obj.type = "X"
        AND obj.date BETWEEN "startDate" AND "endDate"
    GROUP BY obj.employeeId) x ON x.employeeId = META(employee).id 
    LEFT JOIN (
    SELECT obj.employeeId,
           COUNT(obj.employeeId) AS totalY
    FROM `bucket` obj
    WHERE obj.type = "Y"
        AND obj.date BETWEEN "startDate" AND "endDate"
    GROUP BY obj.employeeId) y ON y.employeeId = META(employee).id 
    LEFT JOIN (
    SELECT obj.employeeId,
           COUNT(obj.employeeId) AS totalZ
    FROM `bucket` obj
    WHERE obj.type = "Z"
        AND obj.date BETWEEN "startDate" AND "endDate"
    GROUP BY obj.employeeId) z ON z.employeeId = META(employee).id 
WHERE employee.type = "Employee"

我也尝试了以下方法，但此查询完全超时！

CREATE INDEX `index_X` ON `bucket`(`type`,`date`, `employeeId`)
WHERE type = "X"
CREATE INDEX `index_Y` ON `bucket`(`type`,`date`, `employeeId`)
WHERE type = "Y"
CREATE INDEX `index_Z` ON `bucket`(`type`,`date`, `employeeId`)
WHERE type = "Z"

SELECT META(employee).id,
       COUNT(x.employeeId),
       COUNT(y.employeeId),
       COUNT(z.employeeId)
FROM `bucket` employee
    LEFT JOIN `bucket` x ON x.employeeId = META(employee).id
    AND x.type = "X"
    AND x.date BETWEEN "startDate" AND "endDate"
    LEFT JOIN `bucket` y ON y.employeeId = META(employee).id
    AND y.type = "Y"
    AND y.date BETWEEN "startDate" AND "endDate"
    LEFT JOIN `bucket` z ON z.employeeId = META(employee).id
    AND z.type = "Z"
    AND z.date BETWEEN "startDate" AND "endDate"
WHERE employee.type = "Employee"
GROUP BY META(employee).id

任何人都可以建议更优化的路线吗？

通过这两个查询，我可以看到我的索引正在被使用，但我也可以在我的查询计划中看到，对于每个连接，"NestedLoopJoin" 都被链接到前一个。这可能是问题所在吗？

我对编写 n1ql 查询和尝试找出最有效的方法仍然很陌生，因此欢迎任何建议。

Answer 1

CREATE INDEX `index_1` ON `agrigistics_dev`(`employeeId`, type, `date`) WHERE type IN ["X", "Y", "Z"];

SELECT META(employee).id,
       SUM(CASE x.type = "X" THEN 1 ELSE 0 END) xcount,
       SUM(CASE x.type = "Y" THEN 1 ELSE 0 END) ycount,
       SUM(CASE x.type = "Z" THEN 1 ELSE 0 END) zcount
FROM `bucket` employee
LEFT JOIN `bucket` x
           ON x.employeeId = META(employee).id AND x.type IN ["X", "Y", "Z"] AND x.date BETWEEN "startDate" AND "endDate"
WHERE employee.type = "Employee"
GROUP BY META(employee).id;

为了避免 case-cade 嵌套循环连接或连接爆炸，在这种情况下使用 CTE(6.5)

CREATE INDEX `index_1` ON `agrigistics_dev`(type, date, `employeeId`) WHERE type IN ["X", "Y", "Z"];

WITH etype AS (SELECT x.employeeId,
               SUM(CASE x.type = "X" THEN 1 ELSE 0 END) xcount,
               SUM(CASE x.type = "Y" THEN 1 ELSE 0 END) ycount,
               SUM(CASE x.type = "Z" THEN 1 ELSE 0 END) zcount
               FROM `bucket` x
               WHERE x.type IN ["X", "Y", "Z"] AND x.date BETWEEN "startDate" AND "endDate"
               GROUP BY x.employeeId)
SELECT META(employee).id,
       SUM(y.xcount) xcount,
       SUM(y.ycount) ycount,
       SUM(y.zcount) zcount
FROM `bucket` AS employee
LEFT JOIN etype AS y ON y.employeeId = META(employee).id
WHERE employee.type = "Employee"
GROUP BY META(employee).id;

https://index-advisor.couchbase.com/indexadvisor/#1

https://blog.couchbase.com/create-right-index-get-right-performance/

N1ql 计算不同的关联文档

N1ql count distinct associated documents

performance

join

count

couchbase

n1ql