比较两个表后创建标志列

Question

我有两个不同的报告 table，其中包含日期时间和报告所有者。我想 select 那些至少写过一次报告的人。我还需要一个计算字段来显示他们编写的报告编号。报告 1 优先，因此如果有人在任何时候撰写了报告 1，新的 report_number 列应该显示 1，否则为 2（针对报告 2）。

'people' table
| person_id | full_name
--------------------------
| 1         | John L Smith
| 2         | Carl M Selt
| 3         | Another Person

'report_1' table
| report_1_id | author_person_id | date_entered | other_columns
---------------------------------------------------------------
| 1           | 1                | 2018-01-12   | foo
| 2           | 1                | 2018-02-18   | foo foo

'report_2' table
| report_2_id | author_person_id | date_entered | other_columns
---------------------------------------------------------------
| 1           | 1                | 2018-03-21   | bar
| 2           | 1                | 2018-03-28   | bar bar
| 3           | 2                | 2018-04-16   | baz
| 4           | 2                | 2018-04-30   | baz baz

想要的结果：

| full_name    | report_number
---------------------------
| John L Smith | 1
| Carl M Smelt | 2

请注意，约翰的 report_number 是 1，尽管他还撰写了报告 2。

报告 1 和报告 2 具有不同的附加列，即使它们在上面看起来相同。

我试过的：

    /* Get people from both reports */
WITH report_1_people AS (
    SELECT P.full_name
    FROM report_1 R1
    INNER JOIN people P ON R1.author_person_id = P.person_id
    WHERE P.full_name IS NOT NULL 
    AND P.full_name <> ''
), report_2_people AS (
    SELECT P2.full_name
    FROM report_2 R2
    INNER JOIN people P2 ON R2.author_person_id = P2.person_id
    WHERE P2.full_name IS NOT NULL 
    AND P2.full_name <> ''
)
SELECT 
    P.full_name,
    CASE WHEN P.full_name IN ( /* Check if in report 1 */
                    SELECT full_name
                    FROM report_1)
                    THEN 1
            ELSE 2
            END AS report_number
FROM people P
WHERE P.full_name IS NOT NULL AND P.full_name <> ''
/* Eliminate duplicate names */
GROUP BY P.full_name 
/* Filter only who either authored report 1 or report 2 */
HAVING P.full_name IN (SELECT full_name
                       FROM report_1_people)
OR P.full_name IN (SELECT full_name
                   FROM report_2_people)

注：GROUP BY 人 table 因为某些原因有重复条目。

查询时间太长以至于与数据库断开连接（超过 24 小时），所以我认为我做错了什么。有没有更好的方法来完成这个基于两个 table 的标志计算列？ SQL 相对较新，所以我想知道是否还有另一种思考方式，我正在考虑 SQL 逻辑。

Answer 1

您可以使用 OUTER APPLY:

SELECT person_id, full_name, COALESCE(ca1.report_num, ca2.report_num)
FROM people
OUTER APPLY (SELECT TOP (1) 1 FROM report_1 WHERE author_person_id = people.person_id) AS ca1(report_num)
OUTER APPLY (SELECT TOP (1) 2 FROM report_2 WHERE author_person_id = people.person_id) AS ca2(report_num)

Demo on db<>fiddle

Answer 2

CTE 上的 IN 很可能会杀死它。

另一种方法是使用 EXISTS 来检查某人是否写过报告。 CASE 表达式可以处理优先级。

SELECT p.full_name,
       CASE
         WHEN EXISTS (SELECT *
                             FROM report_1 r1
                             WHERE r1.author_person_id = p.person_id) THEN
           1
         WHEN EXISTS (SELECT *
                             FROM report_2 r2
                             WHERE r2.author_person_id = p.person_id) THEN
           2
       END report_number
       FROM people p
       WHERE EXISTS (SELECT *
                            FROM report_1 r1
                            WHERE r1.author_person_id = p.person_id)
              OR EXISTS (SELECT *
                                FROM report_2 r2
                                WHERE r2.author_person_id = p.person_id);

为了提高性能，请尝试在 report_1 (author_person_id) 和 report_2 (author_person_id) 上建立索引。对于 people，您可以在 person_id（可能已经存在）上尝试一个索引，或者在 person_id 和 full_name.

上尝试一个复合索引

Answer 3

这只是获取结果的另一种方式。

SELECT 
    P.full_name,
    MIN( R.Report_Number) AS report_number
FROM people P
OUTER APPLY (SELECT 1 WHERE EXISTS(SELECT * FROM report_1 R1 WHERE R1.author_person_id = P.person_id)
             UNION ALL
             SELECT 2 WHERE EXISTS(SELECT * FROM report_2 R2 WHERE R2.author_person_id = P.person_id)) AS R(Report_Number)
WHERE P.full_name IS NOT NULL AND P.full_name <> ''
/* Eliminate duplicate names */
GROUP BY P.full_name;

比较两个表后创建标志列

Create flag column after comparing two tables

sql

tsql

sql-server

performance

sql-server-2017