需要对不完全匹配名称的查询结果进行聚合

Question

我有一个查询，用于计算数据库中每个合作伙伴的提及次数。

查询如下所示：

SELECT partner, COUNT(*) AS Total
FROM database.data
GROUP BY 1;

输出如下所示：

partner  | total
---------+------
X        | 5
Y        | 3
Z        | 2
X aa     | 6
aa X     | 7
Y aa     | 1

我需要做的是像X ， X aa ， aa X这样的伙伴一起计算在一行中（ Y和Y aa相同。我最后尝试添加HAVING功能，但无法使其工作并且不确定它实际上是否适合使用。

将不胜感激任何帮助！ 谢谢！

Answer 1

无优化：

WITH RECURSIVE
-- rank by length
cte1 AS ( SELECT partner, total, DENSE_RANK() OVER (ORDER BY LENGTH(partner)) rnk
          FROM  data ),
-- find pairs where partner is a substring of anoter partner
cte2 AS ( SELECT partner, total, rnk, partner short
          FROM cte1
          WHERE rnk = 1
          UNION ALL
          SELECT cte1.partner, 
                 cte1.total, 
                 cte1.rnk, 
                 CASE WHEN LOCATE(cte2.partner, cte1.partner)
                      THEN cte2.partner
                      ELSE cte1.partner
                      END
          FROM cte1, cte2
          WHERE cte1.rnk = cte2.rnk + 1 ),
-- select shortest
cte3 AS ( SELECT partner, 
                 total, 
                 rnk, short, 
                 ROW_NUMBER() OVER (PARTITION BY partner ORDER BY LENGTH(short)) rn
          FROM cte2 )
-- get needed data
SELECT short partner, SUM(total) total
FROM cte3 
WHERE rn = 1
GROUP BY short
ORDER BY partner

小提琴

Answer 2

我认为应该这样做：

SELECT d1.partner, COUNT(*) AS Total
FROM (SELECT * FROM database.data d1
    WHERE CHAR_LENGTH(d1.partner) = 1
    GROUP BY d1.partner
    ) d1
-- i
LEFT JOIN database.data d2 ON
        -- get x in aa x and x aa
        d2.partner LIKE CONCAT('%', d1.partner, '%')
-- remoove x aa and aa x from the count table
GROUP BY d1.partner;

---

使用模拟数据：

CREATE TABLE IF NOT EXISTS tmp_mock_data
SELECT * FROM (
                  SELECT 'X' partner
                  UNION ALL
                  SELECT 'Y'
                  UNION ALL
                  SELECT 'Z'
                  UNION ALL
                  SELECT 'Y aa'
                  UNION ALL
                  SELECT 'Z aa'
                  UNION ALL
                  SELECT 'X aa'
                  UNION ALL
                  SELECT 'aa Z'
                  UNION ALL
                  SELECT 'aa Y'
                  UNION ALL
                  SELECT 'aa X'
              ) A
CROSS JOIN (SELECT NULL UNION ALL SELECT NULL) B;
SELECT d1.partner, COUNT(*) AS Total
FROM (SELECT * FROM tmp_mock_data d1
    WHERE CHAR_LENGTH(d1.partner) = 1
    GROUP BY d1.partner
    ) d1
-- i
LEFT JOIN tmp_mock_data d2 ON
        -- get x in aa x and x aa
        d2.partner LIKE CONCAT('%', d1.partner, '%')
-- remoove x aa and aa x from the count table
GROUP BY d1.partner;
DROP TABLE tmp_mock_data;

Answer 3

SELECT partner,COUNT(*) AS Total

FROM(

SELECT (CASE WHEN ASCII(LEFT(partner,1)) BETWEEN 65 AND 90 THEN  LEFT(partner,1) ELSE RIGHT(partner,1) END)  AS partner

FROM database.data)

GROUP BY 1;