impala支持的數據庫里的double(float)類型，通過遷移inceptor后，double類型的值都變成了null。
通過查閱日志發現默認將double轉換成Decimal(38,10)然而他們的數據小數點后面不止10位。
然而那邊的平台不支持decimal這個數據類型（數據目前存放在另一個平台），用戶不用那個平台自己建表設成decimal就可以正常顯示
[2020-01-20 09:30:06.804] INFO i.t.t.o.t.i.r.JDBCReaderMinorNode - === Execute SQLs in JDBC Reader Thread for Table [ecmp.a00001_data_cleaning], Size [0 MB] : ===
[2020-01-20 09:30:06.806] INFO i.t.t.o.t.i.r.JDBCReaderMinorNode - === Job splitted into [1] mapper tasks ===
[2020-01-20 09:30:07.213] INFO i.t.t.supporter.InceptorInstance - org.apache.hive.jdbc.HiveConnection@7e1b82f4 apply settings: inceptor.smallfile.automerge=true
[2020-01-20 09:30:07.218] INFO i.t.t.supporter.InceptorInstance - Execute SQL: [
CREATE TABLE `default`.`TDT__INTERNAL__6f970de4_74b9_4bb1_ac3b_94e368315cb8__a00001_data_cleaning` (
`ci_id` Int,`ci_code` String,`cusage` Decimal(38,10),`sts` String,`day` String,`year` Int,`month` Int,`dayofmonth` Int,`timeofday` String,`hour` Int,`minute` Int,`label` Int,`__record_insert_time__` String,`_store_level` TinyInt,`time_label` Int
) ROW FORMAT SERDE 'org.apache.hadoop.hive.ql.io.tdt.JDBCSerde'
STORED AS INPUTFORMAT 'org.apache.hadoop.hive.ql.io.tdt.JDBCDBInputFormat' OUTPUTFORMAT 'org.apache.hadoop.hive.ql.io.orc.OrcOutputFormat' TBLPROPERTIES('mapreduce.jdbc.driver.class'='shade.org.apache.hive.jdbc.HiveDriver', 'mapreduce.jdbc.url'='jdbc:cdh2://10.70.248.203:21050/ecmp;auth=noSasl','mapreduce.jdbc.input.query'='SELECT ci_id,ci_code,cusage,sts,day,year,month,dayofmonth,timeofday,hour,minute,label,__record_insert_time__,_store_level,time_label FROM ecmp.a00001_data_cleaning limit 1000','mapreduce.jdbc.input.count.query'='select count(*) from (SELECT ci_id,ci_code,cusage,sts,day,year,month,dayofmonth,timeofday,hour,minute,label,__record_insert_time__,_store_level,time_label FROM ecmp.a00001_data_cleaning limit 1000) tdt_sq_alias','mapreduce.jdbc.splits.number'='1','mapreduce.jdbc.username'='admin@ziadmin01', 'mapreduce.jdbc.password'='ziadmin01','mapreduce.jdbc.driver.file'='/tmp/transporter1/6f543c31-a80f-4e14-8fac-fa71a32b8270/1579483802901.jar')
]
[2020-01-20 09:30:07.438] INFO i.t.t.supporter.InceptorInstance - Execute SQL: [
CREATE VIEW `default`.`TDT__VIEW__6f970de4_74b9_4bb1_ac3b_94e368315cb8__a00001_data_cleaning` (
`ci_id`,`ci_code`,`cusage`,`sts`,`day`,`year`,`month`,`dayofmonth`,`timeofday`,`hour`,`minute`,`label`,`__record_insert_time__`,`_store_level`,`time_label`
) AS SELECT
`ci_id`,
`ci_code`,
`cusage`,
`sts`,
`day`,
`year`,
`month`,
`dayofmonth`,
`timeofday`,
`hour`,
`minute`,
`label`,
`__record_insert_time__`,
`_store_level`,
`time_label`
FROM
`default`.`TDT__INTERNAL__6f970de4_74b9_4bb1_ac3b_94e368315cb8__a00001_data_cleaning`
]
起初懷疑是transporter底層設置的問題，問了相關人士得到：
studio-1.0版本的tdt只完整測試過mysql、oracle、db2、sql server、postgresql、hana、達夢，其它數據庫都沒測試過
，不能保證其它數據庫支持或者沒有問題
然后又測試修改精度，讀寫兩邊精度都是16，但還是null
最后通過修改讀取的字段的精度實現了，無需修改底層
SELECT ci_id,ci_code,CAST(cusage AS DECIMAL(38,16)) AS cusage,sts,day,year,month,dayofmonth,timeofday,hour,minute,label,__record_insert_time__,_store_level,time_label FROM ecmp.a00001_data_cleaning limit 1000