sql partition by和rank的统计用法

淩亂°似流年 2023-10-10 23:47 6阅读 0赞

### 问题背景 ###

最近公司有个项目需要用到某种特殊的统计用法， 例如从所有的数据中找出每个账号最新的一条余额信息（根据某个关键信息进行排序并获取排序值最高的记录）。  
当时用的是非常普通的语句，也就是多个子查询嵌套select出每个账号第一条的id，然后left join会主表获取

### 数据准备 ###

![在这里插入图片描述][2b4bef6fcd7a44aaaa27d65898c70035.png]  
![在这里插入图片描述][b53a2b1f64854c9c8c946767df1fc1d4.png]  
执行以下sql来初始化表格和数据

-- ----------------------------
    -- Table structure for test_cust_acct
    -- ----------------------------
    DROP TABLE IF EXISTS "public"."test_cust_acct";
    CREATE TABLE "public"."test_cust_acct" (
      "account_number" varchar(32) COLLATE "pg_catalog"."default",
      "customer_id" varchar(32) COLLATE "pg_catalog"."default"
    )
    ;
    
    -- ----------------------------
    -- Records of test_cust_acct
    -- ----------------------------
    INSERT INTO "public"."test_cust_acct" VALUES ('18522222', 'CNC_10010');
    INSERT INTO "public"."test_cust_acct" VALUES ('804461110', 'WOM_10086');
    
    
    /*
     Navicat Premium Data Transfer
     by https://zhengkai.blog.csdn.net/
    */
    
    
    -- ----------------------------
    -- Table structure for test_intraday
    -- ----------------------------
    DROP TABLE IF EXISTS "public"."test_intraday";
    CREATE TABLE "public"."test_intraday" (
      "_id" int4 NOT NULL,
      "account_number" varchar(32) COLLATE "pg_catalog"."default",
      "amount" numeric(12,2),
      "currency" varchar(12) COLLATE "pg_catalog"."default",
      "balance_date" date
    )
    ;
    
    -- ----------------------------
    -- Records of test_intraday
    -- ----------------------------
    INSERT INTO "public"."test_intraday" VALUES (1, '804461110', 12315.00, 'CNY', '2023-05-27');
    INSERT INTO "public"."test_intraday" VALUES (2, '18522222', 45611.11, 'HKD', '2023-05-27');
    INSERT INTO "public"."test_intraday" VALUES (3, '18522222', 45622.22, 'HKD', '2023-05-28');
    INSERT INTO "public"."test_intraday" VALUES (4, '18522222', 45633.33, 'HKD', '2023-06-01');
    INSERT INTO "public"."test_intraday" VALUES (5, '804461110', -54321.00, 'CNY', '2023-06-01');
    
    -- ----------------------------
    -- Primary Key structure for table test_intraday
    -- ----------------------------
    ALTER TABLE "public"."test_intraday" ADD CONSTRAINT "test_intraday_pkey" PRIMARY KEY ("_id");

### 补习班 ###

`什么是partition by 和 rank ？`

1.  一旦各种开窗函数使用partition by后，需要明确的是，函数会对每个分区独立处理。  
    avg(page) over(partition by book)  
    假如整体是对阿利波特七部曲，则该函数是求出每一部哈利波特的平均页码数。
2.  rank是T\_sql 函数，rank函数返回结果集分区内指定字段的值的排名，指定字段的值的排名是相关行之前的排名加一。

### 解决方案与SQL实战 ###

1.  第一步，先查出所有的记录  
    ![在这里插入图片描述][c1799b91b6e347098e2cd06d91e71d10.png]

-- START
    with myresult as (
    select ca.*,i.* from test_cust_acct ca 
    INNER JOIN test_intraday i on ca.account_number=i.account_number
    where ca.customer_id='CNC_10010'
    )
    
    select * FROM myresult;

让我们explain一下  
![在这里插入图片描述][52d135e71fbc4df3a917eb3595eec4f8.png]

1.  第二步，加上partition by 和 rank 获取排序

![在这里插入图片描述][56a82e5ea2604283b7f34b9a1558af7e.png]

-- AFTER
    with myresult as (
    select i.*,rank() over (partition by i.account_number ORDER BY i.account_number,i.balance_date desc) from test_cust_acct ca 
    INNER JOIN test_intraday i on ca.account_number=i.account_number
    where ca.customer_id in ('CNC_10010','WOM_10086')
    )
    
    select * FROM myresult;

![在这里插入图片描述][4f18389a9b6e4b36a2b47bf5e5aa5f9e.png]  
让我们继续explain一下，  
![在这里插入图片描述][a8985bf0c6074d72b4778fa78e6e2d3c.png]  
3. 第三步，之找到rank为1的也就是所有account最新一条的数据。

-- FINAL
    with myresult as (
    select i.*,rank() over (partition by i.account_number ORDER BY i.account_number,i.balance_date desc) from test_cust_acct ca 
    INNER JOIN test_intraday i on ca.account_number=i.account_number
    where ca.customer_id in ('CNC_10010','WOM_10086')
    )
    
    select * FROM myresult where rank=1;

![在这里插入图片描述][22c61370bc0d411fb7e156a45654402a.png]

### 对比 ###

--old
    select ii.* from (
    select max(myresut._id) as _id ,myresut.account_number,max(myresut.balance_date) as balance_date from (select ca.customer_id,i.* from test_cust_acct ca 
    INNER JOIN test_intraday i on ca.account_number=i.account_number 
    where ca.customer_id in ('CNC_10010','WOM_10086')) myresut
    GROUP BY myresut.account_number
    ) finalresult
    left JOIN test_intraday ii on finalresult.account_number=ii.account_number and ii._id=finalresult._id;

![在这里插入图片描述][3ff5ebc385ae4d268650a3e3204bf552.png]  
![在这里插入图片描述][72b80161cf4f439994dcb35689bdd732.png]

[2b4bef6fcd7a44aaaa27d65898c70035.png]: https://img-blog.csdnimg.cn/2b4bef6fcd7a44aaaa27d65898c70035.png
[b53a2b1f64854c9c8c946767df1fc1d4.png]: https://img-blog.csdnimg.cn/b53a2b1f64854c9c8c946767df1fc1d4.png
[c1799b91b6e347098e2cd06d91e71d10.png]: https://img-blog.csdnimg.cn/c1799b91b6e347098e2cd06d91e71d10.png
[52d135e71fbc4df3a917eb3595eec4f8.png]: https://img-blog.csdnimg.cn/52d135e71fbc4df3a917eb3595eec4f8.png
[56a82e5ea2604283b7f34b9a1558af7e.png]: https://img-blog.csdnimg.cn/56a82e5ea2604283b7f34b9a1558af7e.png
[4f18389a9b6e4b36a2b47bf5e5aa5f9e.png]: https://img-blog.csdnimg.cn/4f18389a9b6e4b36a2b47bf5e5aa5f9e.png
[a8985bf0c6074d72b4778fa78e6e2d3c.png]: https://img-blog.csdnimg.cn/a8985bf0c6074d72b4778fa78e6e2d3c.png
[22c61370bc0d411fb7e156a45654402a.png]: https://img-blog.csdnimg.cn/22c61370bc0d411fb7e156a45654402a.png
[3ff5ebc385ae4d268650a3e3204bf552.png]: https://img-blog.csdnimg.cn/3ff5ebc385ae4d268650a3e3204bf552.png
[72b80161cf4f439994dcb35689bdd732.png]: https://img-blog.csdnimg.cn/72b80161cf4f439994dcb35689bdd732.png