使用group-by子句优化MySQL查询

Question

我想根据国家、合作伙伴和零售商来计算总点击量和唯一点击量。我已经达到了预期的结果，但我认为这不是一个最佳解决方案，对于更长的数据集，这将需要更长的时间。如何改进此查询？这是我的测试表，设计的查询和预期的输出：

"country_id","partner","retailer","id_customer","id_clickout"
"1","A","B","100","XX"
"1","A","B","100","XX"
"2","A","B","100","XX"
"2","A","B","100","GG"
"2","A","B","100","XX"
"2","A","B","101","XX"
DROP TABLE IF EXISTS x;
CREATE TEMPORARY TABLE x AS
SELECT test1.country_id, test1.partner,test1.retailer, test1.id_customer, 
SUM(CASE WHEN test1.id_clickout IS NULL THEN 0 ELSE 1 END) AS clicks,
CASE WHEN test1.id_clickout IS NULL THEN 0 ELSE 1 END AS unique_clicks
FROM test1
GROUP BY 1,2,3,4
;
SELECT country_id,partner,retailer, SUM(clicks), SUM(unique_clicks)
FROM x
GROUP BY 1,2,3

输出：

"country_id","partner","retailer","SUM(clicks)","SUM(unique_clicks)"
"1","A","B","2","1"
"2","A","B","4","2"

这是DDL和输入数据：

CREATE TABLE test (
 country_id INT(11) DEFAULT NULL,
 partner VARCHAR(256) CHARACTER SET utf8 DEFAULT NULL,
 retailer VARCHAR(256) CHARACTER SET utf8 DEFAULT NULL,
 id_customer BIGINT(20) DEFAULT NULL,
 id_clickout VARCHAR(256) CHARACTER SET utf8 DEFAULT NULL)
  ENGINE=InnoDB DEFAULT CHARSET=utf8;

INSERT INTO test VALUES(1,'A','B','100','XX'),(1,'A','B','100','XX'),
            (2,'A','B','100','XX'),(2,'A','B','100','GG'),
            (2,'A','B','100','XX'),(2,'A','B','101','xx')

Answer 1

SELECT
  country_id,
  partner,
  retailer,
  COUNT(id_clickout)   AS clicks,
  COUNT(DISTINCT CASE WHEN id_clickout IS NOT NULL THEN id_customer END) AS unique_clicks
FROM
  test1
GROUP BY
  1,2,3
;

COUNT(a_field)不会计算任何NULL值。

因此，COUNT(id_clickout)将只计算它是NOT NULL的次数。

同样，unique_clicks中的CASE WHEN语句只为单击的记录返回id_customer，否则返回NULL。这意味着COUNT(DISTINCT CASE)只计算不同的客户，并且只在他们点击时计算。

编辑：

我刚刚意识到，它可能比这更简单。。。

SELECT
  country_id,
  partner,
  retailer,
  COUNT(*)                    AS clicks,
  COUNT(DISTINCT id_customer) AS unique_clicks
FROM
  test1
WHERe
  id_clickout IS NOT NULL
GROUP BY
  1,2,3
;

结果中唯一的实质性差异是，以前与0 clicks一起出现的任何country_id, partner, retailed现在都不会出现在结果中。

但是，如果在country_id, partner, retailed, id_clickout, id_customer或country_id, partner, retailed, id_customer, id_clickout上使用INDEX，则此查询应该会快得多。

Answer 2

我想这就是你想要的：

SELECT country_id,partner,retailer,COUNT(retailer) as `sum(clicks)`,count(distinct id_clickout) as `SUM(unique_clicks)`
FROM test1
GROUP BY country_id,partner,retailer

结果：

COUNTRY_ID  PARTNER  RETAILER  SUM(CLICKS)  SUM(UNIQUE_CLICKS)
1           A        B         2            1
2           A        B         4            2

请参阅SQL Fiddle中的结果。