sql总结
- 场景1分组后失败的成功数据带入
- 场景2完全性质的一对一匹配
- 场景3虚拟户的特殊匹配
- 场景4多对多匹配
- 场景5一对一匹配
- [场景6 一对多匹配](#场景6 一对多匹配)
场景1分组后失败的成功数据带入
现有一批交易表的数据，根据户名，日期，金额分组，存在TRADE_STATUS 有成功和失败的情况，
而失败的FUNDFLOW_DETAILS_ID字段已和资金表的FUNDFLOW_DETAILS_ID匹配。
现需要将成功的凭证号和期间根据";"拼接到一起，存储到与该组失败数据匹配的资金表的数据中，存储到JOURNAL_VOUCHER_NUMBER和ACCOUNT_DATE 字段
使用
HAVING
sum(CASE WHEN a.TRADE_STATUS = '失败' THEN 1 ELSE 0 END) >=1
AND
sum(CASE WHEN a.TRADE_STATUS = '成功' THEN 1 ELSE 0 END) >=1
判断该分组下是否至少有一条成功的数据和一条失败的数据
sql
DECLARE
CURSOR res_data IS
SELECT a.CUSTOMER_ACCOUNT_NAME_ORIGIN,
TRUNC(a.TRADE_DATE_BAK) AS ts_date,
a.TRADE_AMOUNT
FROM RC_TRADEFLOW_DETAILS_NAPR a
WHERE 1=1
AND a.OURS_ACCOUNT_NUMBER LIKE '%02222'
GROUP BY a.CUSTOMER_ACCOUNT_NAME_ORIGIN,TRUNC(a.TRADE_DATE_BAK),a.TRADE_AMOUNT
HAVING
sum(CASE WHEN a.TRADE_STATUS = '失败' THEN 1 ELSE 0 END) >=1
AND
sum(CASE WHEN a.TRADE_STATUS = '成功' THEN 1 ELSE 0 END) >=1
;
a_vouchers varchar2(4000);
a_periods varchar2(4000);
BEGIN
FOR s IN res_data LOOP
SELECT EXTRACT(XMLAGG(XMLELEMENT(e, pz, ';') ORDER BY pz),'/E/text()').GETCLOBVAL() INTO a_vouchers
FROM (
SELECT DISTINCT a.JOURNAL_VOUCHER_NUMBER as pz
FROM RC_TRADEFLOW_DETAILS_NAPR a
WHERE 1=1
AND a.OURS_ACCOUNT_NUMBER LIKE '%2222'
AND a.CUSTOMER_ACCOUNT_NAME_ORIGIN = s.CUSTOMER_ACCOUNT_NAME_ORIGIN
AND TRUNC(a.TRADE_DATE_BAK) = s.ts_date
AND a.TRADE_AMOUNT = s.TRADE_AMOUNT
AND a.TRADE_STATUS = '成功'
);
SELECT EXTRACT(XMLAGG(XMLELEMENT(e, qj, ';') ORDER BY qj),'/E/text()').GETCLOBVAL() INTO a_periods
FROM (
SELECT DISTINCT a.ACCOUNT_DATE as qj
FROM RC_TRADEFLOW_DETAILS_NAPR a
WHERE 1=1
AND a.OURS_ACCOUNT_NUMBER LIKE '%02222'
AND a.CUSTOMER_ACCOUNT_NAME_ORIGIN = s.CUSTOMER_ACCOUNT_NAME_ORIGIN
AND TRUNC(a.TRADE_DATE_BAK) = s.ts_date
AND a.TRADE_AMOUNT = s.TRADE_AMOUNT
AND a.TRADE_STATUS = '成功'
);
UPDATE RC_FUNDFLOW_DETAILS_NAPR A
SET a.JOURNAL_VOUCHER_NUMBER = a_vouchers,
a.ACCOUNT_DATE = a_periods,
a.RULER_NAME_TRADEFLOW = '失败和成功同一批次，失败已经和资金表的资金流水编号匹配,将成功的凭证号拼接带入那个资金'
WHERE 1=1
AND a.OURS_ACCOUNT_NUMBER LIKE '%02222'
AND a.FUNDFLOW_DETAILS_ID IN (
SELECT DISTINCT b.FUNDFLOW_DETAILS_ID
FROM RC_TRADEFLOW_DETAILS_NAPR b
WHERE 1=1
AND b.OURS_ACCOUNT_NUMBER LIKE '%02222'
AND b.CUSTOMER_ACCOUNT_NAME_ORIGIN = s.CUSTOMER_ACCOUNT_NAME_ORIGIN
AND TRUNC(b.TRADE_DATE_BAK) = s.ts_date
AND b.TRADE_AMOUNT = s.TRADE_AMOUNT
AND b.TRADE_STATUS = '失败'
)
;
END LOOP;
COMMIT;
END;场景2完全性质的一对一匹配
现有一批数据，企业表和交易表已经匹配大量数据，但还有剩余，剩余数据匹配规则如下：
1.两批数据必须在账户，户名，日期，金额分组下仅有一条，防止匹配错乱，
2.按照账户，户名，日期，金额进行匹配
3.将企业表的凭证号和期间带入资金。
使用窗口函数，在不影响性能的情况下，快速判断该数据在该分区下该条数据是否仅有一条
sql
BEGIN
FOR s in (
SELECT
qy.DOC_SEQUENCE_VALUE,qy.PERIOD_NAME,zj.URID
from(
SELECT
ls.mydate,
ls.CUSTOMER_NAME,
ls.money,
ls.DOC_SEQUENCE_VALUE,
ls.PERIOD_NAME
FROM(SELECT
TO_CHAR(DEFAULT_EFFECTIVE_DATE, 'YYYY-MM-dd') AS mydate,
CUSTOMER_NAME,
(CASE WHEN ENTERED_DR <> 0 THEN ENTERED_DR ELSE -ENTERED_CR end) AS money,
DOC_SEQUENCE_VALUE,
PERIOD_NAME,
count(*) OVER(PARTITION BY TO_CHAR(DEFAULT_EFFECTIVE_DATE, 'YYYY-MM-dd'),CUSTOMER_NAME,ENTERED_DR,ENTERED_CR) AS partition_count
FROM rc_fin_transactions
WHERE 1=1 AND BANKACCNO_SID = (select sid from RC_BANKACCDEF where BANKACCNO like '%11111'))ls
WHERE ls.partition_count = 1
) qy
INNER JOIN (
SELECT
ls.mydate,
ls.DF_ACCOUNTNAME,
ls.AMOUNT,
ls.URID
FROM(SELECT
TO_CHAR(POSTDATETIME, 'YYYY-MM-dd') AS mydate,
DF_ACCOUNTNAME,
AMOUNT,
URID,
COUNT(*) OVER (PARTITION BY TO_CHAR(POSTDATETIME, 'YYYY-MM-dd'), DF_ACCOUNTNAME, AMOUNT) AS partition_count
FROM ISSUE202412_RECON_ZJLSB
WHERE wf_account LIKE '%11111' AND 企业账凭证号 IS NULL) ls
WHERE ls.partition_count = 1
) zj
on qy.mydate = TO_CHAR(TO_DATE(zj.mydate,'YYYY-MM-DD')+1,'YYYY-MM-DD') AND qy.CUSTOMER_NAME = zj.DF_ACCOUNTNAME AND qy.money = zj.amount
) LOOP
-- 将企业的凭证号和期间带入资金
update ISSUE202412_RECON_ZJLSB
set 企业账凭证号 = s.DOC_SEQUENCE_VALUE,
企业账期间 = s.PERIOD_NAME,
规则名称 = '日期+户名+单笔金额1对1企业直接带入资金'
where URID = s.URID;
END LOOP;
COMMIT;
END;场景3虚拟户的特殊匹配
现有交易表和资金表数据，要求交易表和资金表按照日期（天）匹配，
规则：
1.如果交易表当天的所有数据能够资金表当天的数据按照日期（天）匹配，那就按当天匹配
2.如果资金表中当天不存在数据，那就找到比当前交易表日期小的资金表中存在的最大的日期进行匹配
3.如果资金表中既不存在当天数据，也不存在比当前交易表日期小的资金表中存在的最大的日期
那就找到比当前交易表日期大的资金表中存在的最小日期进行匹配。
总结：就近匹配
将与交易表当天匹配的资金表的那天的所有FUNDSFLOWCODE进行拼接，放入交易表当天的每一条数据中。
**将日期匹配条件进行优化，
原先TRUNC(TRADE_DATE_BAK) = rec.TRADE_DATE_BAK 因为牵涉到函数计算字段，导致索引失效，
优化为范围查询，大大提升性能
AND d.postdatetime >= rec.TRADE_DATE_BAK
AND d.postdatetime < rec.TRADE_DATE_BAK + 1
优化原理，TRUNC会截取到日期的天如2025-01-03 23:33:13.111会被截取为2025-01-03 00:00:00.000
，故那天的所有日期都会被截取为2025-01-03 00:00:00.000，如果想获取当天的所有日期，仅需找到当天日期范围在
2025-01-03 00:00:00.000,2025-01-04 00:00:00.000)也就是 AND d.postdatetime \>= rec.TRADE_DATE_BAK AND d.postdatetime \< rec.TRADE_DATE_BAK + 1** ```sql DECLARE CURSOR cur_data IS SELECT DISTINCT TRUNC(TRADE_DATE_BAK) AS TRADE_DATE_BAK FROM rc_tradeflow_details_napr -- 交易流水表 WHERE OURS_ACCOUNT_NUMBER = '11111.com' AND ATTRIBUTE4 IS NULL ORDER BY TRUNC(TRADE_DATE_BAK); cx_date DATE; SUM_FUNDSFLOWCODE VARCHAR2(4000); temp VARCHAR2(4000); commit_counter NUMBER := 0; BEGIN FOR rec IN cur_data LOOP BEGIN -- 找出与当前日期相同的日期 SELECT TRUNC(d.postdatetime) INTO cx_date FROM rc_fundflow_details_napr d -- 资金流水表 WHERE 1=1 AND d.MONEY_WAY ='充值' AND d.OURS_ACCOUNT_NUMBER in ('22222','33333','44444') AND d.CUSTOMER_ACCOUNT_NAME_ORIGIN = '支付宝（中国）网络技术有限公司' AND d.CUSTOMER_ACCOUNT_NUMBER = '88888' -- AND TRUNC(TRADE_DATE_BAK) = rec.TRADE_DATE_BAK AND d.postdatetime >= rec.TRADE_DATE_BAK -- 优化为范围查询 AND d.postdatetime < rec.TRADE_DATE_BAK + 1 AND ROWNUM = 1; EXCEPTION WHEN NO_DATA_FOUND THEN cx_date := NULL; END; -- 找出比当前日期小的最大日期 IF cx_date IS NULL THEN SELECT MAX(TRUNC(d.postdatetime)) INTO cx_date FROM rc_fundflow_details_napr d -- 资金流水表 WHERE 1=1 AND d.MONEY_WAY ='充值' AND d.OURS_ACCOUNT_NUMBER in ('22222','33333','44444') AND d.CUSTOMER_ACCOUNT_NAME_ORIGIN = '支付宝（中国）网络技术有限公司' AND d.CUSTOMER_ACCOUNT_NUMBER = '88888' AND d.postdatetime < rec.TRADE_DATE_BAK + INTERVAL '1' DAY; END IF; -- 找出比当前日期大的最小日期 IF cx_date IS NULL THEN SELECT MIN(TRUNC(d.postdatetime)) INTO cx_date FROM rc_fundflow_details_napr d -- 资金流水表 WHERE 1=1 AND d.MONEY_WAY ='充值' AND d.OURS_ACCOUNT_NUMBER in ('22222','33333','44444') AND d.CUSTOMER_ACCOUNT_NAME_ORIGIN = '支付宝（中国）网络技术有限公司' AND d.CUSTOMER_ACCOUNT_NUMBER = '88888' AND d.postdatetime > rec.TRADE_DATE_BAK; END IF; -- 拼接这些日期的交易流水编号 SELECT (RTRIM(EXTRACT(XMLAGG(XMLELEMENT(e, FUNDSFLOWCODE, ';') ORDER BY FUNDSFLOWCODE),'/E/text()').GETCLOBVAL(),';')) INTO SUM_FUNDSFLOWCODE FROM( SELECT FUNDFLOW_DETAILS_ID AS FUNDSFLOWCODE FROM rc_fundflow_details_napr d -- 资金流水表 WHERE 1=1 AND d.MONEY_WAY ='充值' AND d.OURS_ACCOUNT_NUMBER in ('22222','33333','44444') AND d.CUSTOMER_ACCOUNT_NAME_ORIGIN = '支付宝（中国）网络技术有限公司' AND d.CUSTOMER_ACCOUNT_NUMBER = '88888' AND d.postdatetime >= cx_date AND d.postdatetime < cx_date +1 ); -- 更新交易流水表的本天的所有数据 UPDATE rc_tradeflow_details_napr SET ATTRIBUTE4 = SUM_FUNDSFLOWCODE WHERE 1=1 AND OURS_ACCOUNT_NUMBER = '11111.com' AND ATTRIBUTE4 IS NULL AND TRADE_DATE_BAK >= rec.TRADE_DATE_BAK AND TRADE_DATE_BAK < rec.TRADE_DATE_BAK+1; -- 增加计数器 commit_counter := commit_counter + 1; -- 每处理一定数量的记录后提交一次事务 IF commit_counter >= 50 THEN COMMIT; commit_counter := 0; -- 重置计数器 END IF; END LOOP; COMMIT; END; ``` ## 场景4多对多匹配 现有一批数据，要求交易表按照账户，日期进行分组汇总得到汇总金额 资金表按照户名，日期进行分组得到汇总金额 按照日期，账户，汇总金额进行匹配，将交易表该组的所有凭证号和期间拼接带入资金表该组的每一条数据 将资金表该组的所有URID拼接带入交易表该组的每一条数据。完成所有这样的组的匹配。 ```sql DECLARE CURSOR data_cur IS( SELECT transaction_date AS mats_1 FROM ( SELECT a.BUSI_DATE AS transaction_date, sum(CASE WHEN a.ENTERED_DR <> 0 THEN a.ENTERED_DR ELSE -a.ENTERED_CR end) AS jy_money FROM rc_bank_transactions a WHERE a.bankaccno_sid = '111111111111' GROUP BY a.BUSI_DATE ) jy INNER JOIN ( SELECT trunc(b.postdatetime) AS postdatetime, sum(b.amount) AS zj_money FROM issue202412_recon_zjlsb b WHERE b.wf_account ='111111111111' AND b.企业账凭证号 IS NULL GROUP BY trunc(b.postdatetime) ) zj ON TRUNC(jy.transaction_date) = zj.postdatetime AND jy.jy_money = zj.zj_money); mats_1 date; a_vouchers varchar2(4000); a_periods varchar2(4000); b_urids varchar2(4000); BEGIN FOR res IN data_cur LOOP mats_1 := res.mats_1; SELECT EXTRACT(XMLAGG(XMLELEMENT(e, pz, ';') ORDER BY pz),'/E/text()').GETCLOBVAL() INTO a_vouchers FROM ( SELECT DISTINCT a.DOC_SEQUENCE_VALUE as pz FROM rc_bank_transactions a WHERE TRUNC(a.BUSI_DATE) = mats_1 AND a.bankaccno_sid = '111111111111' ); SELECT EXTRACT(XMLAGG(XMLELEMENT(e, qj, ';') ORDER BY qj),'/E/text()').GETCLOBVAL() INTO a_periods FROM ( SELECT DISTINCT a.FIN_PERIOD_NAME as qj FROM rc_bank_transactions a WHERE TRUNC(a.BUSI_DATE) = mats_1 AND a.bankaccno_sid = '111111111111' ); SELECT EXTRACT(XMLAGG(XMLELEMENT(e,urid, ';') ORDER BY urid),'/E/text()').GETCLOBVAL() INTO b_urids FROM ( SELECT DISTINCT b.urid as urid FROM issue202412_recon_zjlsb b WHERE trunc(b.postdatetime) = mats_1 AND b.wf_account = '111111111111' ); UPDATE rc_bank_transactions a SET a.urid = b_urids WHERE TRUNC(a.BUSI_DATE) = mats_1 AND a.bankaccno_sid = '111111111111'; UPDATE issue202412_recon_zjlsb b SET b.企业账凭证号 = a_vouchers, b.企业账期间 = a_periods WHERE trunc(b.postdatetime) = mats_1 AND b.wf_account = '111111111111'; END LOOP; COMMIT; END; ``` ## 场景5一对一匹配 现有一批数据，企业表和交易表要求按照账户，投保单，金额，日期匹配，将企业表的凭证号和期间带入交易表。 ```sql DECLARE CURSOR c_update IS SELECT a.ROWID AS row_id, b.DOC_SEQUENCE_VALUE, b.PERIOD_NAME FROM rc_bank_transactions_202501 a, rc_fin_transactions_202501 b WHERE 1=1 AND a.bankaccno_sid = b.bankaccno_sid AND a.REF_NUMBER = b.REF_NUMBER AND (CASE WHEN a.ENTERED_DR <> 0 THEN a.ENTERED_DR ELSE -a.ENTERED_CR end) = (CASE WHEN b.ENTERED_DR <> 0 THEN b.ENTERED_DR ELSE -b.ENTERED_CR end) AND TO_CHAR(a.BUSI_DATE, 'YYYY-MM') = TO_CHAR(b.DEFAULT_EFFECTIVE_DATE, 'YYYY-MM') AND a.DOC_SEQUENCE_VALUE IS NULL; BEGIN FOR rec IN c_update LOOP UPDATE rc_bank_transactions_202501 SET DOC_SEQUENCE_VALUE = rec.DOC_SEQUENCE_VALUE, FIN_PERIOD_NAME = rec.PERIOD_NAME WHERE ROWID = rec.row_id; END LOOP; COMMIT; END; ``` ## 场景6 一对多匹配 现有一批数据要求交易数据按照账户，批次号，日期进行分组汇总得到汇总金额与资金表的账户，日期，金额进行匹配， 要求将交易表的拼接后的凭证号和期间带入与之匹配的资金表的数据中，同时将资金表的FUNDSFLOWCODE带入与之匹配的所以交易表的数据。 ```sql begin FOR s in ( SELECT ll.reqbatchno,ll.transaction_date,ss.FUNDSFLOWCODE FROM ( SELECT a.reqbatchno, a.PAYSENTDATE AS transaction_date, sum(a.PAYMONEY) AS jyamount FROM issue202501_rc_batch_jylsb a WHERE a.wf_ACCOUNT = '1111111111111111111111' GROUP BY a.reqbatchno, a.PAYSENTDATE) ll INNER JOIN ( SELECT d.postdatetime AS postdatetime, d.amount, d.FUNDSFLOWCODE FROM issue202501_recon_zjlsb d WHERE d.wf_account = '1111111111111111111111') ss ON TRUNC(ss.postdatetime) = TRUNC(ll.transaction_date) AND ss.amount = ll.jyamount ) LOOP update issue202501_rc_batch_jylsb set URID = s.FUNDSFLOWCODE,资金表规则名称 = '11111更新规则1：我方账号+交易日期+单笔金额' where 1=1 AND wf_ACCOUNT = '1111111111111111111111' AND reqbatchno = s.reqbatchno AND PAYSENTDATE = s.transaction_date ; UPDATE issue202501_recon_zjlsb e SET e.规则名称= '11111更新规则1：批次号+交易提交日期+汇总金额', e.企业账凭证号 = ( SELECT RTRIM(XMLAGG(XMLELEMENT(x, DOC_SEQUENCE_VALUE, ';').EXTRACT('//text()') ORDER BY DOC_SEQUENCE_VALUE).GetClobVal(), ';') FROM ( SELECT DISTINCT DOC_SEQUENCE_VALUE FROM issue202501_rc_batch_jylsb WHERE URID = s.FUNDSFLOWCODE ) sub ), e.企业账期间 = ( SELECT RTRIM(XMLAGG(XMLELEMENT(x,FIN_PERIOD_NAME, ';').EXTRACT('//text()') ORDER BY FIN_PERIOD_NAME).GetClobVal(), ';') FROM( SELECT DISTINCT FIN_PERIOD_NAME FROM issue202501_rc_batch_jylsb WHERE URID = s.FUNDSFLOWCODE ) sua ) WHERE e.FUNDSFLOWCODE = s.FUNDSFLOWCODE; end loop; commit; -- 提交所有数据 end; ```