insert overwrite table ods.customer partition (part_date="20211020") select `(part_date)?+.+` from ods.customer where part_date="20211021";

This query will overwrite the data in the partition "part_date=20211020" of the table "ods.customer" with the data from the partition "part_date=20211021" of the same table. The query uses dynamic partitioning to select the partition based on the value of the "part_date" column.

相关问题

insert overwrite table discountdw.dwd_sd_adds_order_bill_inc partition(dt = '2023-06-06') select t1.order_bill_id, t1.counterfoil_no, t1.acceptor, date_format(to_utc_timestamp(cast(t1.expiry_date as bigint) ,'GMT-8'),'YYYY-MM-dd'), t2.company_id, t1.cert_no, t1.company_name, t1.third_order_id, t1.counterfoil_amt/10000, t1.transaction_amt/10000, t1.rate, '3bp' as service_tariffing, ((DATEDIFF(to_utc_timestamp(t1.expiry_date ,'GMT-8'),to_utc_timestamp(t1.transaction_date ,'GMT-8') ) + adjust_days)* 0.0003 *(counterfoil_amt))/ 360 as service_fee, 360 as total_days, DATEDIFF(to_utc_timestamp(t1.expiry_date ,'GMT-8'),to_utc_timestamp(t1.transaction_date ,'GMT-8') ) + adjust_days as modulation_date, t3.channel_type, t3.bank_name, date_format(to_utc_timestamp(cast(t1.transaction_date as bigint) ,'GMT-8'),'YYYY-MM-dd'), t1.order_status_code, t1.order_status_msg, t4.fee_amt, t4.status, t1.tenant_id, t5.revenue, to_utc_timestamp(cast(t1.create_date as bigint) ,'GMT-8'), to_utc_timestamp(cast(t1.update_date as bigint) ,'GMT-8') from (select * from discountdw.ods_adds_order_bill_inc where dt ='2023-06-06' and channel_id=101110004 )t1 left join (select * from mecdw.ods_company_full where platform_id='sdpjw')t2 on t1.cert_no=t2.cert_no and t1.tenant_id=t2.tenant_id left join discountdw.dim_adds_product_full t3 on t1.partner_id=t3.partner_id and t1.product_id=t3.product_id left join (select * from mecdw.dwd_sc_fee_record_full where dt='2023-06-06' and biz_type=2 ) t4 on t1.order_bill_id=t4.third_id left join (select * from discountdw.ods_sd_order_ext_inc where dt='2023-06-06') t5 on t1.order_bill_id=t5.order_bill_id left join sdpjwdw.dim_holiday_info_full t6 on date_format(to_utc_timestamp(t1.expiry_date ,'GMT-8'),'YYYY-MM-dd') = t6.civic_holiday ;

这是一条 SQL 语句，看起来比较长。以下是一些可以考虑优化的建议：

1. 将表名、字段名等缩短，以减少字符数和提高可读性。
2. 将复杂的表连接操作拆分成多个子查询，以降低查询复杂度。
3. 对于一些计算操作，可以考虑使用内置函数或自定义函数，以简化代码并提高可维护性。
4. 对于一些重复的操作，可以将其提取成变量或常量，以减少重复代码。

综上所述，优化后的 SQL 可能类似于以下代码（仅供参考）：

-- 定义常量
SET adjust_days = 1;

-- 查询订单表
WITH order_data AS (
  SELECT
    o.order_bill_id,
    o.counterfoil_no,
    o.acceptor,
    DATE_FORMAT(TO_UTC_TIMESTAMP(CAST(o.expiry_date AS BIGINT), 'GMT-8'), 'YYYY-MM-dd') AS expiry_date,
    o.cert_no,
    o.company_name,
    o.third_order_id,
    o.counterfoil_amt / 10000 AS counterfoil_amt,
    o.transaction_amt / 10000 AS transaction_amt,
    o.rate,
    '3bp' AS service_tariffing,
    ((DATEDIFF(TO_UTC_TIMESTAMP(o.expiry_date, 'GMT-8'), TO_UTC_TIMESTAMP(o.transaction_date, 'GMT-8')) + adjust_days) * 0.0003 * (o.counterfoil_amt)) / 360 AS service_fee,
    360 AS total_days,
    DATEDIFF(TO_UTC_TIMESTAMP(o.expiry_date, 'GMT-8'), TO_UTC_TIMESTAMP(o.transaction_date, 'GMT-8')) + adjust_days AS modulation_date,
    o.order_status_code,
    o.order_status_msg,
    o.tenant_id,
    TO_UTC_TIMESTAMP(CAST(o.create_date AS BIGINT), 'GMT-8') AS create_date,
    TO_UTC_TIMESTAMP(CAST(o.update_date AS BIGINT), 'GMT-8') AS update_date
  FROM discountdw.ods_adds_order_bill_inc o
  WHERE o.dt = '2023-06-06' AND o.channel_id = 101110004
),

-- 查询公司表
company_data AS (
  SELECT
    c.company_id,
    c.cert_no,
    c.tenant_id
  FROM mecdw.ods_company_full c
  WHERE c.platform_id = 'sdpjw'
),

-- 查询产品表
product_data AS (
  SELECT
    p.partner_id,
    p.product_id,
    p.channel_type,
    p.bank_name
  FROM discountdw.dim_adds_product_full p
),

-- 查询费用记录表
fee_data AS (
  SELECT
    f.third_id,
    f.fee_amt,
    f.status
  FROM mecdw.dwd_sc_fee_record_full f
  WHERE f.dt = '2023-06-06' AND f.biz_type = 2
),

-- 查询订单扩展信息表
order_ext_data AS (
  SELECT
    e.order_bill_id,
    e.revenue
  FROM discountdw.ods_sd_order_ext_inc e
  WHERE e.dt = '2023-06-06'
),

-- 查询节假日表
holiday_data AS (
  SELECT
    h.civic_holiday
  FROM sdpjwdw.dim_holiday_info_full h
)

-- 合并数据
INSERT OVERWRITE TABLE discountdw.dwd_sd_adds_order_bill_inc PARTITION(dt = '2023-06-06')
SELECT
  o.order_bill_id,
  o.counterfoil_no,
  o.acceptor,
  o.expiry_date,
  c.company_id,
  o.cert_no,
  o.company_name,
  o.third_order_id,
  o.counterfoil_amt,
  o.transaction_amt,
  o.rate,
  o.service_tariffing,
  o.service_fee,
  o.total_days,
  o.modulation_date,
  p.channel_type,
  p.bank_name,
  DATE_FORMAT(TO_UTC_TIMESTAMP(CAST(o.transaction_date AS BIGINT), 'GMT-8'), 'YYYY-MM-dd'),
  o.order_status_code,
  o.order_status_msg,
  f.fee_amt,
  f.status,
  o.tenant_id,
  e.revenue,
  o.create_date,
  o.update_date
FROM order_data o
LEFT JOIN company_data c ON o.cert_no = c.cert_no AND o.tenant_id = c.tenant_id
LEFT JOIN product_data p ON o.partner_id = p.partner_id AND o.product_id = p.product_id
LEFT JOIN fee_data f ON o.order_bill_id = f.third_id
LEFT JOIN order_ext_data e ON o.order_bill_id = e.order_bill_id
LEFT JOIN holiday_data h ON o.expiry_date = DATE_FORMAT(TO_UTC_TIMESTAMP(h.civic_holiday, 'GMT-8'), 'YYYY-MM-dd');

set hive.exec.dynamic.partition.mode=nonstrict; set hive.exec.parallel=true; set hive.exec.parallel.thread.number=30; --drop table if exists dm_finrisk.dm_event_walletSettleSuccessEvent_di; create table if not exists dm_finrisk.dm_event_cashLoanEvent_di( event_uuid string ,event_code string ,dt string ,event_time string ,rawSnapshot string ) COMMENT 'cashLoanEvent' PARTITIONED BY ( day bigint)stored as parquet; drop table if exists tmp.tmp_dm_event_cashLoanEvent_di_01_${zdt.addDay(-1).format("yyyyMMdd")}; create table if not exists tmp.tmp_dm_event_cashLoanEvent_di_01_${zdt.addDay(-1).format("yyyyMMdd")} as select uuid as event_uuid ,event as event_code ,from_unixtime(unix_timestamp(cast(day as string),'yyyymmdd'),'yyyy-mm-dd') as dt ,from_unixtime(cast(cast(`time` AS BIGINT)/1000 AS BIGINT),'yyyy-MM-dd HH:mm:ss') as event_time ,get_json_object(data,'$.hotPublish.rawSnapshot') as rawSnapshot from dm_finrisk.ods_event_hdfs_snapshot_di where code = 'shield@cashLoanEvent' and day=${zdt.addDay(-1).format("yyyyMMdd")} drop table if exists tmp.tmp_dm_event_cashLoanEvent_di_02_${zdt.addDay(-1).format("yyyyMMdd")}; create table if not exists tmp.tmp_dm_event_cashLoanEvent_di_02_${zdt.addDay(-1).format("yyyyMMdd")} as select event_uuid ,event_code ,dt ,event_time ,rawSnapshot from tmp.tmp_dm_event_cashLoanEvent_di_01_${zdt.addDay(-1).format("yyyyMMdd")}; alter table dm_finrisk.dm_event_cashLoanEvent_di drop if exists partition (day = ${zdt.addDay(-1).format("yyyyMMdd")}); insert overwrite table dm_finrisk.dm_event_cashLoanEvent_di partition (day = ${zdt.addDay(-1).format("yyyyMMdd")}) select event_uuid ,event_code ,dt ,event_time ,rawSnapshot from tmp.tmp_dm_event_cashLoanEvent_di_02_${zdt.addDay(-1).format("yyyyMMdd")}分析下上面代码

上面的代码主要是一个HiveQL脚本，用于从`dm_finrisk.ods_event_hdfs_snapshot_di`表中提取`code`为`shield@cashLoanEvent`且`day`为昨天日期的数据，并将其存储到`dm_finrisk.dm_event_cashLoanEvent_di`表中的对应分区中。

具体来说，该脚本的执行步骤如下：

1. 第1行：设置Hive执行参数，包括动态分区模式、并行执行和并行执行线程数等。

2. 第3-8行：创建`dm_finrisk.dm_event_cashLoanEvent_di`表，该表包含5个列，分别为`event_uuid`、`event_code`、`dt`、`event_time`和`rawSnapshot`，其中`dt`为分区列。

3. 第10-15行：创建临时表`tmp.tmp_dm_event_cashLoanEvent_di_01_${zdt.addDay(-1).format("yyyyMMdd")}`，该表从`dm_finrisk.ods_event_hdfs_snapshot_di`表中选择`code`为`shield@cashLoanEvent`且`day`为昨天日期的记录，并将其中的`uuid`、`event`、`day`、`time`和`data`字段分别映射到`event_uuid`、`event_code`、`dt`、`event_time`和`rawSnapshot`列中。

4. 第17-22行：创建临时表`tmp.tmp_dm_event_cashLoanEvent_di_02_${zdt.addDay(-1).format("yyyyMMdd")}`，该表从临时表`tmp.tmp_dm_event_cashLoanEvent_di_01_${zdt.addDay(-1).format("yyyyMMdd")}`中选择所有列，用于后续的数据处理。

5. 第24行：如果`dm_finrisk.dm_event_cashLoanEvent_di`表中昨天的分区已经存在，则删除该分区。

6. 第26-27行：将临时表`tmp.tmp_dm_event_cashLoanEvent_di_02_${zdt.addDay(-1).format("yyyyMMdd")}`中的数据插入到`dm_finrisk.dm_event_cashLoanEvent_di`表的昨天分区中。

整个脚本的作用是将来自`dm_finrisk.ods_event_hdfs_snapshot_di`表的特定数据提取出来，并存储到`dm_finrisk.dm_event_cashLoanEvent_di`表的对应分区中，以便后续的分析和查询。