WHERE用于过滤行，如果想过滤分组，需要使用HAVING。
实例：
列出具有两个以上产品且价格大于等于4的供应商。
SELECT vend_id,COUNT(*) AS num_prods FROM Products
WHERE prod_price>=4
GROUP BY vend_id
HAVING COUNT(*)>=2
面试容易被坑：FROM子句一定必须使用吗？ 不是的。SELECT 1+1也是合法的。
爱出的面试题：
1.找出表中重复数据。
SELECT D_ADDRESS,COUNT(D_ADDRESS) AS ADDRESS_COUNT FROM T_DEMO
GROUP BY D_ADDRESS HAVING COUNT(D_ADDRESS)>1
注意下面这种情况：
SELECT D_ADDRESS,COUNT(1) AS ADDRESS_COUNT FROM T_DEMO
GROUP BY D_ADDRESS HAVING COUNT(1)>1
值为NULL的也会被抓出来，具体需要哪种看具体情况。
2.删除重复数据。（注意：这里有坑，不是让你全删了，其实是希望你删掉重复的，并保留一条数据。）
2.1 如果只有一个字段重复：D_ADDRESS
SELECT * FROM T_DEMO
WHERE D_ADDRESS IN
(SELECT D_ADDRESS FROM T_DEMO GROUP BY D_ADDRESS HAVING COUNT(D_ADDRESS)>1)
AND D_NAME NOT IN
(SELECT MIN(D_NAME) FROM T_DEMO GROUP BY D_ADDRESS HAVING COUNT(D_ADDRESS)>1)
把SELECT * 换成DELETE就行了。
注意：建议在做任何Insert,Update,Delete操作前，都用SELECT查一下，“预览”一下结果。
当然也可以用Transaction来检查，如果结果不对就rollback，但是多个人同时操作一张表时，transaction可能会锁死表（因为某个人没提交）。
另一种写法：
2.2 把IN换成EXISTS:
SELECT * FROM T_DEMO AS T
WHERE EXISTS
(SELECT 1 FROM T_DEMO WHERE T.D_ADDRESS=D_ADDRESS GROUP BY D_ADDRESS HAVING COUNT(D_ADDRESS)>1)
AND NOT EXISTS
(SELECT 1 FROM T_DEMO GROUP BY D_ADDRESS HAVING COUNT(D_ADDRESS)>1 AND T.D_NAME=MIN(D_NAME))
2.3 如果有多个字段重复：D_ADDRESS 和 D_REMARKS
SELECT * FROM T_DEMO AS T WHERE EXISTS
(SELECT 1 FROM T_DEMO WHERE T.D_REMARKS=D_REMARKS AND T.D_ADDRESS=D_ADDRESS GROUP BY D_ADDRESS,D_REMARKS HAVING COUNT(1)>1)
AND NOT EXISTS
(SELECT 1 FROM T_DEMO GROUP BY D_ADDRESS,D_REMARKS HAVING COUNT(1)>1 AND T.D_NAME=MIN(D_NAME))
注意：这里已经不能使用IN了