有时候复杂的逻辑处理可以交给SQL来处理，自认为是比java处理要快点。
举个例子：如何统计每个品牌的的偏好用户数？
当时我做的处理是在java中处理这些品牌，若是品牌相同，就加一，但是这种方法很慢，用户体验很不好。
后来才发现sql处理速度会很快，就用UNION ALL 来处理。在这里简单说一下UNION和UNION ALL区别，这两个都是用于合并select的结果集。只不过UNION ALL可以允许有重复值。一条sql就可以解决问题。如下：
SELECT
T.BRANDNAME AS brandName,
SUM(T.NUMB) AS userNum
FROM
(
SELECT
PREFER_BRAND1 AS BRANDNAME,
COUNT(PREFER_BRAND1) AS NUMB
FROM
TAS_TAS.TERM_CONT_RECOM_ITEM_201605
GROUP BY
PREFER_BRAND1
UNION ALL
SELECT
PREFER_BRAND2 AS BRANDNAME,
COUNT(PREFER_BRAND2) AS NUMB
FROM
TAS_TAS.TERM_CONT_RECOM_ITEM_201605
GROUP BY
PREFER_BRAND2
UNION ALL
SELECT
PREFER_BRAND3 AS BRANDNAME,
COUNT(PREFER_BRAND3) AS NUMB
FROM
TAS_TAS.TERM_CONT_RECOM_ITEM_201605
GROUP BY
PREFER_BRAND3 ) T
WHERE
T.BRANDNAME!='-9'
GROUP BY
T.BRANDNAME
ORDER BY
SUM(T.NUMB) DESC
这样就可以知道每个品牌的偏好人数了。