SQL面试题练习 —— 查询每个产品每年总销售额

1 题目

已知有表 t_product_sales 如下，记录了每个产品id、产品名称、产品销售开始日期、产品销售结束日期以及产品日均销售金额，请计算出每个产品每年的销售金额。

样例数据

+-------------+---------------+----------------------+----------------------+----------------------+
| product_id  | product_name  |     period_start     |      period_end      | average_daily_sales  |
+-------------+---------------+----------------------+----------------------+----------------------+
| 1           | LC Phone      | 2019-01-25 00:00:00  | 2019-02-28 00:00:00  | 100                  |
| 2           | LC T-Shirt    | 2018-12-01 00:00:00  | 2020-01-01 00:00:00  | 10                   |
| 3           | LC Keychain   | 2019-12-01 00:00:00  | 2020-01-31 00:00:00  | 1                    |
+-------------+---------------+----------------------+----------------------+----------------------+
1
2
3
4
5
6
7

期望结果

+------------+--------------+-------------+--------------+
| product_id | product_name | report_year | total_amount |
+------------+--------------+-------------+--------------+
| 1          | LC Phone     |    2019     | 3500         |
| 2          | LC T-Shirt   |    2018     | 310          |
| 2          | LC T-Shirt   |    2019     | 3650         |
| 2          | LC T-Shirt   |    2020     | 10           |
| 3          | LC Keychain  |    2019     | 31           |
| 3          | LC Keychain  |    2020     | 31           |
+------------+--------------+-------------+--------------+
1
2
3
4
5
6
7
8
9
10

2 建表语句

--建表语句
create table if not exists t_product_sales
(
    product_id          bigint,
    product_name        string,
    period_start        string,
    period_end          string,
    average_daily_sales bigint
)
    ROW FORMAT DELIMITED
        FIELDS TERMINATED BY ','
    STORED AS orc;

--插入数据

insert into t_product_sales(product_id, product_name, period_start, period_end, average_daily_sales)
values (1, 'LC Phone', '2019-01-25 00:00:00', '2019-02-28 00:00:00', 100),
       (2, 'LC T-Shirt', '2018-12-01 00:00:00', '2020-01-01 00:00:00', 10),
       (3, 'LC Keychain', '2019-12-01 00:00:00', '2020-01-31 00:00:00', 1);
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

3 题解

（1）构建年份维表

with dim_year as (select '2018' as year, '2018-01-01' as year_first_day, '2018-12-31' as year_end_day
                  union all
                  select '2019' as year, '2019-01-01' as year_first_day, '2019-12-31' as year_end_day
                  union all
                  select '2020' as year, '2020-01-01' as year_first_day, '2020-12-31' as year_end_day)
select year,year_first_day,year_end_day from dim_year
1
2
3
4
5
6

执行结果

+-------+-----------------+---------------+
| year  | year_first_day  | year_end_day  |
+-------+-----------------+---------------+
| 2018  | 2018-01-01      | 2018-12-31    |
| 2019  | 2019-01-01      | 2019-12-31    |
| 2020  | 2020-01-01      | 2020-12-31    |
+-------+-----------------+---------------+
1
2
3
4
5
6
7

（2）年份维表与原始数据进行关联

把原始数据与年份维表进行笛卡尔积，得到每年与原始数据的一个交叉值。

# 这一设置的含义是关闭Hive中的笛卡尔积严格检查。
# 具体来说，Hive默认情况下不允许生成笛卡尔积（即两个没有连接条件的表的笛卡尔积），
# 因为笛卡尔积通常会生成非常大的结果集，可能会导致性能问题或资源耗尽。
# 通过将这个设置设为false，你可以禁用这种严格检查，从而允许在查询中生成笛卡尔积。

set hive.strict.checks.cartesian.product = false;
with dim_year as (select '2018' as year, '2018-01-01' as year_first_day, '2018-12-31' as year_end_day
                  union all
                  select '2019' as year, '2019-01-01' as year_first_day, '2019-12-31' as year_end_day
                  union all
                  select '2020' as year, '2020-01-01' as year_first_day, '2020-12-31' as year_end_day)
select
    product_id,
    product_name,
    period_start,
    period_end,
    average_daily_sales,
    year,
    year_first_day,
    year_end_day
from t_product_sales
left join dim_year
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22

执行结果

+-------------+---------------+----------------------+----------------------+----------------------+-------+-----------------+---------------+
| product_id  | product_name  |     period_start     |      period_end      | average_daily_sales  | year  | year_first_day  | year_end_day  |
+-------------+---------------+----------------------+----------------------+----------------------+-------+-----------------+---------------+
| 1           | LC Phone      | 2019-01-25 00:00:00  | 2019-02-28 00:00:00  | 100                  | 2019  | 2019-01-01      | 2019-12-31    |
| 1           | LC Phone      | 2019-01-25 00:00:00  | 2019-02-28 00:00:00  | 100                  | 2020  | 2020-01-01      | 2020-12-31    |
| 1           | LC Phone      | 2019-01-25 00:00:00  | 2019-02-28 00:00:00  | 100                  | 2018  | 2018-01-01      | 2018-12-31    |
| 2           | LC T-Shirt    | 2018-12-01 00:00:00  | 2020-01-01 00:00:00  | 10                   | 2019  | 2019-01-01      | 2019-12-31    |
| 2           | LC T-Shirt    | 2018-12-01 00:00:00  | 2020-01-01 00:00:00  | 10                   | 2020  | 2020-01-01      | 2020-12-31    |
| 2           | LC T-Shirt    | 2018-12-01 00:00:00  | 2020-01-01 00:00:00  | 10                   | 2018  | 2018-01-01      | 2018-12-31    |
| 3           | LC Keychain   | 2019-12-01 00:00:00  | 2020-01-31 00:00:00  | 1                    | 2019  | 2019-01-01      | 2019-12-31    |
| 3           | LC Keychain   | 2019-12-01 00:00:00  | 2020-01-31 00:00:00  | 1                    | 2020  | 2020-01-01      | 2020-12-31    |
| 3           | LC Keychain   | 2019-12-01 00:00:00  | 2020-01-31 00:00:00  | 1                    | 2018  | 2018-01-01      | 2018-12-31    |
+-------------+---------------+----------------------+----------------------+----------------------+-------+-----------------+---------------+
1
2
3
4
5
6
7
8
9
10
11
12
13

（3）计算每年每个产品在售天数，计算年销售额

我们先观察2中的结果，可以看到原始记录中每行数据都与所有年都有一行记录。我们从销售日期和每年开始日期中取较大日期 得到一个开始时间，然后从销售截止日期和每年的结束日期取较小日期 得到一个结束日期，然后用结束日期减去开始日期。

with dim_year as (select '2018' as year, '2018-01-01' as year_first_day, '2018-12-31' as year_end_day
                  union all
                  select '2019' as year, '2019-01-01' as year_first_day, '2019-12-31' as year_end_day
                  union all
                  select '2020' as year, '2020-01-01' as year_first_day, '2020-12-31' as year_end_day),
     tmp as (select product_id
                  , product_name
                  , period_start
                  , period_end
                  , average_daily_sales
                  , year
                  , year_first_day
                  , year_end_day
                  , datediff(if(to_date(period_end)
                                    > to_date(year_end_day)
                                 , to_date(year_end_day)
                                 , to_date(period_end))
             , if(to_date(period_start)
                      > to_date(year_first_day)
                                 , to_date(period_start)
                                 , to_date(year_first_day))) as date_diff
             from t_product_sales
                      left join dim_year)
select product_id,
       product_name,
       year,
       (date_diff + 1) * average_daily_sales as total_amount
from tmp
where date_diff >= 0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

执行结果

+-------------+---------------+-------+---------------+
| product_id  | product_name  | year  | total_amount  |
+-------------+---------------+-------+---------------+
| 1           | LC Phone      | 2019  | 3500          |
| 2           | LC T-Shirt    | 2018  | 310           |
| 2           | LC T-Shirt    | 2019  | 3650          |
| 2           | LC T-Shirt    | 2020  | 10            |
| 3           | LC Keychain   | 2019  | 31            |
| 3           | LC Keychain   | 2020  | 31            |
+-------------+---------------+-------+---------------+
1
2
3
4
5
6
7
8
9
10

只有在当年有销售时间的数据date_diff >=0，如果在当年没有销售时间，则date_diff为负。但是这个date_diff 的值比预期小1，这是因为我们算了日期差，所以我们在结果上+1即可。