SQL (2) - join window function view

SQL(二) —— joinWindow function view

大家可以关注知乎或微信公众号的share16,我们也会同步更新此文章.

一、Views and subqueries

1.1 什么是视图

   视图是一个虚拟表,不同于直接操作数据表,视图是依据SELECT语句来创建的,所以操作视图时会根据创建视图的SELECT语句生成一张虚拟表,然后在这张虚拟表上做SQL操作.

   The reason the view exists：

• By defining a view,Can be used frequentlySELECT语句保存以提高效率;
• By defining a view,It can make the data seen by the user more clear;
• By defining a view,All fields of the data table may not be exposed to the public,增强数据的保密性;
• By defining a view,Data redundancy can be reduced;

视图与表的区别：是否保存了实际的数据

1.2 视图操作

创建视图
   CREATE VIEW < 视图名 > AS
   < SELECT语句 >

修改视图
   ALTER VIEW < 视图名 > AS
   < SELECT语句 >

删除视图
   DROP VIEW < 视图名 >

   视图归根结底还是从表派生出来的,So if the original table is updated,那么视图中的数据也可以更新;Conversely, if the view has changed,而原表没有进行相应更新的话,就无法保证数据的一致性.

1.3 什么是子查询

   子查询也被称为嵌套查询,指一个查询语句嵌套在另一个查询语句内部的查询;When using subqueries in a query,Subqueries are executed first,The main query is then executed based on the results returned by the subquery.

• 嵌套子查询：Subqueries can also be nestedSELECT语句;
• 标量子查询：Also called a single subquery,即返回表中具体的“某一行的某一列”;
• 关联子查询：are subqueries that rely on information from the main query,This means that the tables in the subquery can be related to the tables in the main query(可理解为：in the main query and subquery,Use the same table);
  
  

二、函数

   所谓函数,类似一个黑盒子,你给它一个输入值,它便按照预设的程序定义给出返回值,输入值称为参数.

• 算术函数：除了+ - * /外,还有 绝对值ABS()、求余MOD()、四舍五入ROUND()等;
• 字符串函数：拼接CONCAT(s1,s2,…)、长度LENGTH(s)、替换REPLACE(s,old,new)、截取(pop即索引,从1开始,python从0开始) SUBSTRING(str FROM pos FOR len)、截取(先按delim对str分割,索引从1开始) SUBSTRING_INDEX(str,delim,count)等;
• 日期函数：now()、curdate()、current_date()、current_time()、获取date的年/时等EXTRACT(unit FROM date)等;
• 转换函数：字符串转数字CAST(‘123’ AS BINARY)、数字转字符串CAST(123 AS CHAR)、Returns the first non-null value from the leftCOALESCE(value1,value2,value3,…)等;
• 谓词：LIKE(%匹配零个或多个字符串,_匹配任意1个字符)、IN、BETWEEN、IS NULL、IS NOT NULL等;
• CASE WHEN表达式：sum(case when…)、count(case when…)等;
  CASE WHEN < 求值表达式 > THEN < 表达式 >
    WHEN < 求值表达式 > THEN < 表达式 > ...
    ELSE < 表达式 > END [AS < 字段名 >]
  
  

三、数据拼接(union/join)

   数据拼接大致可分为两类,That is, splicing up and down(集合运算)and splicing left and right(列连接).

集合运算

• union：取并集且去重,但union allTake the union without deduplication;(The number of fields in the upper and lower tables&The numeric types must be the same;若要order排序,Just write it on the last line);
• intersect：取交集,但目前的mysqlThe version does not support arithmetic;
• except：取差集(The former table minus the latter table,If the order is different,运行结果也不同),但目前的mysqlThe version does not support arithmetic;

列连接

• 内连接(join或inner join)：Data on both sides will be retained,类似于交集;
• 左连接(left join)：The data in the left table will be retained,If there is no matching data in the right table,则返回null;
• 右连接(right join)：The data in the right table will be retained,If there is no matching data in the left table,则返回null;
• 全连接(full join)：Regardless of whether the two sides match,所有数据都会保留,If there is no matching data,则返回null;
• 笛卡尔积(cross join)：Also called cross-bonding,as left table3条记录,右表4条记录,结果有12条记录;
  
  
  

四、窗口函数

   窗口函数也称为OLAP函数,意思是对数据库数据进行实时分析处理.常规的SELECT语句都是对整张表进行查询,The window function allows us to selectively take a certain part of the data for summarization、计算和排序.

<窗口函数> OVER ([PARTITION BY <列名>] ORDER BY <列名>)
partition by：用来分组,类似于group by子句;order by：用来排序,That is to decide which rule to follow(字段)来排序;

• 专用窗口函数：有rank、dense_rank、row_number等;
• Application of aggregate functions：有sum、avg、max等,其结果是一个累计值;
• Add subtotals/总计等：group by子句后,添加WITH ROLLUP;
  

五、练习题

The dataset used for the exercises,点此下载.

01. 连接(join)

   请使用A股上市公司季度营收预测数据集income_statement、company_operating和market_data;以market _data为主表,将三张表中的TICKER_SYMBOL为000545和200550的信息合并在一起(只需要显示以下字段).

select distinct a.TICKER_SYMBOL,a.END_DATE,a.CLOSE_PRICE,
b.INDIC_NAME_EN,b.VALUE,c.T_REVENUE,c.T_COGS,c.N_INCOME
from market_data a 
left join company_operating b 
on a.TICKER_SYMBOL = b.TICKER_SYMBOL and a.END_DATE = b.END_DATE

left join income_statement c  
on a.TICKER_SYMBOL = c.TICKER_SYMBOL and a.END_DATE = c.END_DATE

where a.TICKER_SYMBOL in (000545,200550)
order by a.TICKER_SYMBOL,a.END_DATE

   使用A股上市公司季度营收预测中的数据集Macro_Industry,计算’Depository Securities: Circulation Market Value of Listed Stocks’在2015年用电最高峰是发生在哪月？并且相比去年同期增长/减少了多少个百分比？

select m.*,n.value_2014,round((m.value_2015-n.value_2014)/n.value_2014,2) 'rate'
from (select month(PERIOD_DATE)'month',sum(DATA_VALUE)'value_2015'
			from macro_industry 
			where year(PERIOD_DATE) = 2015
			and name_cn = 'Depository Securities: Circulation Market Value of Listed Stocks'
			group by month ) m 
left join (select month(PERIOD_DATE)'month',sum(DATA_VALUE)'value_2014'
						from macro_industry 
						where year(PERIOD_DATE) = 2014
						and name_cn = 'Depository Securities: Circulation Market Value of Listed Stocks'
						group by month ) n on m.month = n.month
order by m.value_2015 desc 

02. 排序(rank/dense_rank/row_number)

   Please use datasetwinequality-red,找出pH=3.03的所有红葡萄酒,然后对其citric acid进行中式排名(相同排名的下一个名次应该是下一个连续的整数值,In other words, there should be no rankings“间隔”).

select  pH,`citric acid`,
dense_rank() over(order by `citric acid`) '排名'
from `winequality-red`
where pH = 3.03

   Please use datasetwinequality-white,找出pH=3.63的所有白葡萄酒,然后对其residual sugar量进行英式排名(非连续的排名).

select  pH, `residual sugar`,
rank() over(order by  `residual sugar`) '排名'
from `winequality-white`
where pH = 3.63

03. 分割字符串(substring_index)

   使用数据集ccf_offline_stage1_test_revised,to be found separately2016年7月期间,发放优惠券总金额最多和发放The merchant with the most coupons(Only full deductions are considered,Do not consider discounts;Discount_rate：x in [0,1]代表折扣率,x:y表示满x减y).

select Merchant_id,sum(amount),count(Coupon_id)
from (select Merchant_id,Coupon_id,Discount_rate,
      SUBSTRING_INDEX(Discount_rate,':',-1) amount,
      case when (Discount_rate between 0 and 1) then '折扣' else '满减' end as type
      from ccf_offline_stage1_test_revised
			where Date_received >= '2016-07-01' and Date_received <= '2016-07-31') k
where k.type = '满减' 
group by Merchant_id
order by sum(amount) desc  -- The merchant with the most total coupon amount
#order by count(Coupon_id) desc -- The merchant with the most coupons
limit 1 

谢谢大家