好久没有写带CURSOR的存储过程了,语法的有点生疏了。今天写了个玩玩。呵呵!该功能是计算数据仓库里的客户的丢失和恢复分析信息的。
/*
Definition:
Customer had sales before, but there is no sales in the last six month.
e.g:
Before or in June has sales
From July to Dec – no sales
Customer lost in Jan
*/
/*
Following is the Solution:
*/
--Create a new VSC_BI_CUSTOMER_STATUS to save the analysis data
USE [DW]
GO
/****** Object: StoredProcedure [dbo].[Setting_Customers_Status] Script Date: 2015/1/30 16:29:25 ******/
SET ANSI_NULLS ON
GO
SET QUOTED_IDENTIFIER ON
GO
--Create Procedure [dbo].[Setting_Customers_Status] --创建存储过程
ALTER Procedure [dbo].[Setting_Customers_Status] --修改存储过程
as
--判断表是否存在
IF NOT EXISTS (SELECT * FROM dbo.SysObjects WHERE ID = object_id(N'VSC_BI_CUSTOMER_STATUS') AND OBJECTPROPERTY(ID, 'IsTable') = 1)
BEGIN
CREATE TABLE [dbo].[VSC_BI_CUSTOMER_STATUS](
--[CUSTOMER_NAME] [nvarchar](100) Collate Chinese_PRC_CI_AS NOT NULL,
[CUSTOMER_NAME] [nvarchar](100) NOT NULL,
[CUSTOMER_ID] [int] NOT NULL,
[YEAR] [nvarchar](4) NOT NULL,
[MONTH] [nvarchar](2) NOT NULL,
[STATUS] [nvarchar](20) NOT NULL,
[INVOICE_DATE] [date] NOT NULL,
[LOST] [int] NOT NULL,
[NORMAL] [int] NOT NULL,
[NEW] [int] NOT NULL,
[RECOVER] [int] NOT NULL
) ON [PRIMARY]
ALTER TABLE [dbo].[VSC_BI_CUSTOMER_STATUS] ADD CONSTRAINT [DF_VSC_BI_CUSTOMERR_STATUS_YEAR] DEFAULT ('') FOR [YEAR]
ALTER TABLE [dbo].[VSC_BI_CUSTOMER_STATUS] ADD CONSTRAINT [DF_VSC_BI_CUSTOMER_STATUS_MONTH] DEFAULT ('') FOR [MONTH]
ALTER TABLE [dbo].[VSC_BI_CUSTOMER_STATUS] ADD CONSTRAINT [DF_VSC_BI_CUSTOMER_STATUS_LOST] DEFAULT ((0)) FOR [LOST]
ALTER TABLE [dbo].[VSC_BI_CUSTOMER_STATUS] ADD CONSTRAINT [DF_VSC_BI_CUSTOMER_STATUS_NORMAL] DEFAULT ((0)) FOR [NORMAL]
ALTER TABLE [dbo].[VSC_BI_CUSTOMER_STATUS] ADD CONSTRAINT [DF_VSC_BI_CUSTOMER_STATUS_NEW] DEFAULT ((0)) FOR [NEW]
ALTER TABLE [dbo].[VSC_BI_CUSTOMER_STATUS] ADD CONSTRAINT [DF_VSC_BI_CUSTOMER_STATUS_RECOVER] DEFAULT ((0)) FOR [RECOVER]
END
--Use the Cursor to get the target data
--测试的时候需要把表里的数据删除了
truncate table VSC_BI_CUSTOMER_STATUS
declare @CUSTOMER_ID int
declare @CUSTOMER_NAME nvarchar(100) --定要定义成NVARCHAR的类型,否则后面有汉字的时候,有可能会出现乱码
declare @INVOICE_DATE date
--记录游标里的下一条记录
declare @CUSTOMER_ID_NEXT int
declare @CUSTOMER_NAME_NEXT nvarchar(100) --定要定义成NVARCHAR的类型,否则后面有汉字的时候,有可能会出现乱码
declare @INVOICE_DATE_NEXT date
--记录丢失客户的那个的月份
declare @INVOICE_DATE_LOST date
--记录循环次数
declare @LOOP_COUNT INT
SET @LOOP_COUNT=0
declare @sql char(5000)
declare table_customer_cursor cursor for
select c.[CUSTOMER_ID],c.CUSTOMER_NAME,h.[INVOICE_DATE]
--COUNT(*) TransTimes
from [dbo].[VSC_BI_CUSTOMER] c,[dbo].[VSC_BI_INV_HEADERS] h
where c.[CUSTOMER_ID]=h.[CUSTOMER_ID] and h.[INVOICE_DATE]>'2012-03-31'
group by c.[CUSTOMER_ID],c.CUSTOMER_NAME,h.[INVOICE_DATE]
--having count(*)>=1
order by c.[CUSTOMER_ID],c.CUSTOMER_NAME,h.[INVOICE_DATE]
open table_customer_cursor
fetch next from table_customer_cursor into @CUSTOMER_ID,@CUSTOMER_NAME,@INVOICE_DATE
WHILE @@FETCH_STATUS = 0
BEGIN
--第一进入循环的时候,@CUSTOMER_ID_NEXT是空的,这个时候直接用上面这个语句取到的值fetch next from table_customer_cursor into @CUSTOMER_ID,@CUSTOMER_NAME,@INVOICE_DATE
if @CUSTOMER_ID_NEXT<>''
BEGIN
set @CUSTOMER_ID=@CUSTOMER_ID_NEXT
set @CUSTOMER_NAME = @CUSTOMER_NAME_NEXT
set @INVOICE_DATE = @INVOICE_DATE_NEXT
END
SET @LOOP_COUNT=@LOOP_COUNT+1
print @LOOP_COUNT
--第一条记录插入到[dbo].[CUSTOMER_STATUS]表里,而且STATUS=NEW
IF @LOOP_COUNT=1
BEGIN
--set @sql='insert into CUSTOMER_STATUS(CUSTOMER_ID,CUSTOMER_NAME,STATUS,INVOICE_DATE) values('+@CUSTOMER_ID+','+''''+@CUSTOMER_NAME+''''+','+'NEW'+','+@INVOICE_DATE+')'
--print @sql
--exec(@sql)
insert into VSC_BI_CUSTOMER_STATUS(CUSTOMER_ID,CUSTOMER_NAME,STATUS,INVOICE_DATE)values(+@CUSTOMER_ID,@CUSTOMER_NAME,'NEW',@INVOICE_DATE)
END
ELSE --不是一个客户的第一条记录
--set @sql='delete from '+@tblname
--exec(@sql)
print @CUSTOMER_ID
print @CUSTOMER_NAME
print @INVOICE_DATE
fetch next from table_customer_cursor into @CUSTOMER_ID_NEXT,@CUSTOMER_NAME_NEXT,@INVOICE_DATE_NEXT
--set @CUSTOMER_ID_NEXT = @CUSTOMER_ID
--set @CUSTOMER_NAME_NEXT = @CUSTOMER_NAME
--set @INVOICE_DATE_NEXT = @INVOICE_DATE
--由于HK的服务器数据库的编码方式导致了汉字显示乱码,所以用一个N'标识,表示加N前缀指定后面的字符串为UNICODE常量,
print N'显示下一行的记录'
print @CUSTOMER_ID_NEXT
print @CUSTOMER_NAME_NEXT
print @INVOICE_DATE_NEXT
--计算逻辑如下:
--如果相邻的两条记录是一个相同的客户,则需要比较两条记录的时间间隔,
--1,如果时间间隔大于6个月,则表示这个客户在上一个时间的基础上加6个月份丢失的,这个时候需要把该客户在那个丢失的月份插入到CUSTOMER_STATUS表里
--同时还要把此期间的下区间记录插入到CUSTOMER_STATUS表,表示此客户在此条记录的月份RECOVERY
--2, 如果时间间隔小于6个月在插入此记录到CUSTOMER_STATUS表里,同时STATUS是NORMAL
--总上所述,除了需要把循环的表记录都要插入CUSTOMER_STATUS表里外,还要插入丢失的那个月份的信息
--如果一个客户循环到最后一条记录,此时要判断该记录距离getdate的时间是否大于6个月,如果是则是丢失LOST(这个时候需要再插入一条丢失的记录,
--即最后一次交易的时间加6个月),否则是正常NORMAL
--Notes:如果客户只有一次交易,则需要特别的考虑,这个时候需要插入两条记录,一条是NEW,一条是丢失的记录
IF (@CUSTOMER_ID=@CUSTOMER_ID_NEXT and @CUSTOMER_NAME=@CUSTOMER_NAME_NEXT) --表示还是同一个客户
--BEGIN
--@INVOICE_DATE_NEXT-@INVOICE_DATE>180
if DATEDIFF(day,@INVOICE_DATE,@INVOICE_DATE_NEXT)>180
begin
--set @INVOICE_DATE_LOST=@INVOICE_DATE+180
set @INVOICE_DATE_LOST=dateadd(day,180,@INVOICE_DATE)
--插入客户丢失的月份信息
insert into VSC_BI_CUSTOMER_STATUS(CUSTOMER_ID,CUSTOMER_NAME,STATUS,INVOICE_DATE)values(+@CUSTOMER_ID_NEXT,@CUSTOMER_NAME_NEXT,'LOST',@INVOICE_DATE_LOST)
--插入恢复的月份信息
insert into VSC_BI_CUSTOMER_STATUS(CUSTOMER_ID,CUSTOMER_NAME,STATUS,INVOICE_DATE)values(+@CUSTOMER_ID_NEXT,@CUSTOMER_NAME_NEXT,'RECOVER',@INVOICE_DATE_NEXT)
end
else
insert into VSC_BI_CUSTOMER_STATUS(CUSTOMER_ID,CUSTOMER_NAME,STATUS,INVOICE_DATE)values(+@CUSTOMER_ID_NEXT,@CUSTOMER_NAME_NEXT,'NORMAL',@INVOICE_DATE_NEXT)
ELSE --不是同一个客户
BEGIN
SET @LOOP_COUNT=0 --是为了把每个客户的第一次交易时,他们会是我们的新客
--此时客户循环到最后一条记录,此时要判断该记录距离getdate的时间是否大于6个月,如果是则是丢失LOST(这个时候需要再插入一条丢失的记录,
--即最后一次交易的时间加6个月),否则是正常NORMAL
if DATEDIFF(day,@INVOICE_DATE,getdate())>180
BEGIN
insert into VSC_BI_CUSTOMER_STATUS(CUSTOMER_ID,CUSTOMER_NAME,STATUS,INVOICE_DATE)values(+@CUSTOMER_ID,@CUSTOMER_NAME,'LOST',dateadd(day,180,@INVOICE_DATE))
--update CUSTOMER_STATUS set STATUS='LOST' where CUSTOMER_ID=@CUSTOMER_ID andCUSTOMER_NAME=@CUSTOMER_NAME and INVOICE_DATE=dateadd(day,180,@INVOICE_DATE)
END
END
END
close table_customer_cursor
deallocate table_customer_cursor
--更新下YEAR和MONTH,其实BI里可以直接用INOVICE_DATE就可以关联到YEAR,MONTH,QUARTER您自己看吧,需要就加上。
update VSC_BI_CUSTOMER_STATUS set YEAR=YEAR(INVOICE_DATE),MONTH=MONTH(INVOICE_DATE)
--更新LOST,RECOVER,NEW,NORMAL数量
update VSC_BI_CUSTOMER_STATUS set LOST=1 WHERE STATUS='LOST'
update VSC_BI_CUSTOMER_STATUS set RECOVER=1 WHERE STATUS='RECOVER'
update VSC_BI_CUSTOMER_STATUS set NEW=1 WHERE STATUS='NEW'
update VSC_BI_CUSTOMER_STATUS set NORMAL=1 WHERE STATUS='NORMAL'
/*调用执行Procedure
--EXEC [dbo].[Setting_Customers_Status]
*/
/* 验证用的SQL
truncate table CUSTOMER_STATUS
select * from CUSTOMER_STATUS
where 1=1
--AND status='RECOVERY'
AND CUSTOMER_ID=3698
order by customEr_id ,invoice_date
SELECT * FROM [dbo].[VSC_BI_CUSTOMER_STATUS] where CUSTOMER_ID=3698
SELECT * FROM [dbo].[VSC_BI_CUSTOMER_STATUS] where STATUS='RECOVER'
TRUNCATE TABLE [dbo].[VSC_BI_CUSTOMER_STATUS]
--汉字乱码
select * from [dbo].[VSC_BI_CUSTOMER_STATUS] WHERE CUSTOMER_ID=12101
select * from [dbo].[VSC_BI_CUSTOMER_STATUS] WHERE CUSTOMER_ID=49531
*/
在写上面的存储过程的时候出现了下面的小故事需要修改:
1,一开始我在创建表的时候,其中一个CUSTOMER_NAME的列我用的是char(100), 由于我们DW服务器的数据库字符集是:Chinese_HongKong_Stroke_90_CI_AS,这个导致我的汉字部分有出现问号乱码的情况出现。
2,直到上面的原因后,我把表的相关子段设计成NVARCHAR类型,但是在执行存储过程后还是发现和上面一样有问号乱码
3,再次查看存储过程,发现我存储过程的@CUSTOMRR_NAME的类型些的是char(100),把类型改成nvarchar(100)后,再次执行存储过程,终于没有了乱码。
Lessons Learned:
如果数据库的数据只是英文和数字,可以用varchar类型,如果数据库需要程序很多国家的语言,尤其像汉字的话,就必须把字段类型设计成nvarchar.
其中N表示是National Language,国家语言。NVARCHAR表示用2个字节存储一个字符,不管是英文字母还是汉字等。