mysql中数据格式转换的一个示例

背景

工作中遇到的一个需求，把这种存储形式的数据（都是测试数据也不包含什么敏感信息就直接截图了），

以下面这种形式返回，

根据表中put_sell字段判定是投放还是卖出，0-投放 1-卖出

对于每天的数据，上面这种是最全的情况，也可能每个asset_scale只有卖出或投放，也可能都没有，像下面这样

Method 1

最先想到这种思路，简单但有缺陷，直解看代码吧

with t1 as (
	select
		asset_scale,
		(case put_sell when 0 then last_week else 0 end) as last_week_put,
		(case put_sell when 0 then current_week else 0 end) as current_week_put,
		(case put_sell when 1 then last_week else 0 end) as last_week_sell,
		(case put_sell when 1 then current_week else 0 end) as current_week_sell
	from book_management.weekly_asset_put_sell where reportDateId = 20220616
)
select
	asset_scale,
	sum(last_week_put) as last_week_put,
	sum(current_week_put) as current_week_put,
	sum(last_week_sell) as last_week_sell,
	sum(current_week_sell) as current_week_sell	
from t1 group by asset_scale;

执行结果：

什么缺陷到这儿就能看出来了，就是本应该是空值null的地方都返回了0，原因是因为mysql中的sum求和null并不会引起传播，也就是null会被忽略掉。直白的说，sum列中如果部分值为null，则忽略null，返回非null值的求和结果，如果sum列的值全为null，则返回null。

针对这个特点，上面代码只需要改动一个很小的地方就可以符合需求了，

with t1 as (
	select
		asset_scale,
		(case put_sell when 0 then last_week else null end) as last_week_put,
		(case put_sell when 0 then current_week else null end) as current_week_put,
		(case put_sell when 1 then last_week else null end) as last_week_sell,
		(case put_sell when 1 then current_week else null end) as current_week_sell
	from book_management.weekly_asset_put_sell where reportDateId = 20220616
)
select
	asset_scale,
	sum(last_week_put) as last_week_put,
	sum(current_week_put) as current_week_put,
	sum(last_week_sell) as last_week_sell,
	sum(current_week_sell) as current_week_sell	
from t1 group by asset_scale;

也就是将case when语句中else后面的0全部换成null，执行结果：

上面代码还可以再精简一下：

select 
	asset_scale,
	abs(sum((case put_sell when 0 then last_week else null end))) as last_week_put,
	abs(sum((case put_sell when 0 then current_week else null end))) as current_week_put,
	abs(sum((case put_sell when 1 then last_week else null end))) as last_week_sell,
	abs(sum((case put_sell when 1 then current_week else null end))) as current_week_sell
from book_management.weekly_asset_put_sell where reportDateId = 20220616 group by asset_scale;

Method 2

思路是将投放与卖出各自过滤出来单独做成一个表，两个表再根据asset_scale做全连接，只不过实现过程中才知道mysql中暂时还不支持full join这种全连接操作，可以用left join+right join+union的方式实现相同效果。

with t1 as (
	select 
		asset_scale,
		last_week as last_week_put,
		current_week as current_week_put
	from book_management.weekly_asset_put_sell where reportDateId = 20220616 and put_sell = 0
),

t2 as (
	select 
		asset_scale,
		last_week as last_week_sell,
		current_week as current_week_sell
	from book_management.weekly_asset_put_sell where reportDateId = 20220616 and put_sell = 1		
)

select  
	t1.asset_scale,
	last_week_put,
	current_week_put,
	last_week_sell,
	current_week_sell
from t1 left join t2 on t1.asset_scale = t2.asset_scale
union
select  
	t1.asset_scale,
	last_week_put,
	current_week_put,
	last_week_sell,
	current_week_sell
from t2 left join t1 on t1.asset_scale = t2.asset_scale;

执行结果：

感觉这种思路比第一种复杂了一些，主要能理解好全连接这块就没其它问题了。