我在雪花中运行一个存储过程并且代码编译失败
Execution error in store procedure SP_POPULATE_SYS_AUDIT_DATA_BMC: SQL compilation error: error line 6 at position 2,776 Invalid argument types for function 'IFF': (VARCHAR(16777216), VARCHAR(16777216), VARCHAR(16777216)) At Statement.execute, line 150 position 61
以下是发生错误的代码片段
v_sql_3 = `insert into `+v_mds_db_name+`.`+v_source_table_name+`_temp
(sys_id, documentkey,tablename,fieldname,value, sys_created_on, action, sys_updated_on, sys_created_by ,
sys_updated_by, sourceinstance, cdctype,_numerify_batchid, cdchash, cdchash_sequence, cdctime, etl_run_number)
select concat(`+v_sys_id_field+`,'~',replace('`+v_field+`',' ','_')) as sys_id,
original_request_id as documentkey,'`+v_source_table_name+`' as tablename, '`+v_field+`' as fieldname,
trim(replace(replace(`+v_sql_2+`concat(case when split_part(substring(log,position('`+v_field_lookup+`',log)+`+v_field_length+`),': ',2)
then substring(split_part(substring(log,position('`+v_field_lookup+`',log)+`+v_field_length+`),': ',2),1,
position('(',split_part(substring(log,position('`+v_field_lookup+`',log)+`+v_field_length+`),': ',2))-1)
else split_part(substring(log,position('`+v_field_lookup+`',log)+`+v_field_length+`),': ',2) end ,': '),
`+v_sql_1+`': ',''),'\n','')) as value, audit_date as sys_created_on,
case when action = 4 then 'Create' when action = 2 then 'Set' when action = 8 then 'Delete' when action = 16 then 'Merge' end,
modified_date as sys_updated_on,submitter as sys_created_by, user as sys_updated_by,sourceinstance,cdctype as cdctype ,
_numerify_batchid, cdchash, cdchash_sequence, cdctime , `+v_etl_run_number+`
from `+v_mds_db_name+`.`+v_mds_table_name+`
where fields_changed like '%;`+v_field+`;%' and log like '%`+v_field_lookup+`%'
union
select concat(`+v_sys_id_field+`,'~',replace('`+v_field+`',' ','_')) as sys_id,
original_request_id as documentkey,'`+v_source_table_name+`' as tablename, '`+v_field+`' as fieldname,
trim(replace(replace(`+v_sql_2+`concat(case when split_part(substring(log,position('`+v_field_lookup+`',log)+`+v_field_length+`),': ',2) REGEXP '.*\\([0-9]\\).*'
then substring(split_part(substring(log,position('`+v_field_lookup+`',log)+`+v_field_length+`),': ',2),1,
position('(',split_part(substring(log,position('`+v_field_lookup+`',log)+`+v_field_length+`),': ',2))-1)
else split_part(substring(log,position('`+v_field_lookup+`',log)+`+v_field_length+`),': ',2) end ,': '),
`+v_sql_1+`': ',''),'\n','')) as value, audit_date as sys_created_on,
case when action = 4 then 'Create' when action = 2 then 'Set' when action = 8 then 'Delete' when action = 16 then 'Merge' end,
modified_date as sys_updated_on,submitter as sys_created_by, user as sys_updated_by,sourceinstance,'X' as cdctype ,
_numerify_batchid, cdchash, cdchash_sequence, cdctime , `+v_etl_run_number+`
from `+v_mds_db_name+`.`+v_audit_table_name+`_final
where original_request_id in (select distinct original_request_id
from `+v_mds_db_name+`.`+v_mds_table_name+`
where fields_changed like '%;`+v_field+`;%' and log like '%`+v_field_lookup+`%' )
and fields_changed like '%;`+v_field+`;%' and log like '%`+v_field_lookup+`%'
and `+v_sys_id_field+` not in ( select v_sys_id_field
from `+v_mds_db_name+`.`+v_mds_table_name+`
where fields_changed like '%;`+v_field+`;%' and log like '%`+v_field_lookup+`%')`;
snowflake.createStatement( { sqlText: v_sql_3} ).execute();
PS：我对雪花很陌生
答案 0 :(得分：2)
问题是这一行
case when action = 4 then 'Create' when action = 2 then 'Set' when action = 8 then 'Delete' when action = 16 then 'Merge' end
末尾应该有一个 AS action
因为前一行是 sys_created_on 而后一行是 sys_updated_on
在 etl_run_number 行之后也应该命名。
所以如果我的编辑是正确的
s1 = 'split_part(substring(log,position('`+v_field_lookup+`',log)+`+v_field_length+`),': ',2)';
v_sql_3 = `insert into `+v_mds_db_name+`.`+v_source_table_name+`_temp
(sys_id, documentkey,tablename,fieldname,value, sys_created_on,
action, sys_updated_on, sys_created_by, sys_updated_by,
sourceinstance, cdctype,_numerify_batchid, cdchash,
cdchash_sequence, cdctime, etl_run_number)
select concat(`+v_sys_id_field+`,'~',replace('`+v_field+`',' ','_')) as sys_id,
original_request_id as documentkey,
'`+v_source_table_name+`' as tablename,
'`+v_field+`' as fieldname,
trim(replace(replace(`+v_sql_2+`concat(
case when '+s1+' IS NOT NULL
then substring('+s1+',1,position('(','+s1+')-1)
else '+s1+' end ,': '),
`+v_sql_1+`': ',''),'\n','')) as value,
audit_date as sys_created_on,
case when action = 4 then 'Create' when action = 2 then 'Set' when action = 8 then 'Delete' when action = 16 then 'Merge' end as action,
modified_date as sys_updated_on,
submitter as sys_created_by,
user as sys_updated_by,
sourceinstance,
cdctype as cdctype,
_numerify_batchid,
cdchash,
cdchash_sequence,
cdctime ,
`+v_etl_run_number+` as etl_run_number
from `+v_mds_db_name+`.`+v_mds_table_name+`
where fields_changed like '%;`+v_field+`;%' and log like '%`+v_field_lookup+`%'
union
select concat(`+v_sys_id_field+`,'~',replace('`+v_field+`',' ','_')) as sys_id,
original_request_id as documentkey,
'`+v_source_table_name+`' as tablename,
'`+v_field+`' as fieldname,
trim(replace(replace(`+v_sql_2+`concat(case when '+s1+' REGEXP '.*\\([0-9]\\).*'
then substring('+s1+',1,
position('(','+s1+')-1)
else '+s1+' end ,': '),
`+v_sql_1+`': ',''),'\n','')) as value,
audit_date as sys_created_on,
case when action = 4 then 'Create' when action = 2 then 'Set' when action = 8 then 'Delete' when action = 16 then 'Merge' end as action,
modified_date as sys_updated_on,
submitter as sys_created_by,
user as sys_updated_by,
sourceinstance,
'X' as cdctype ,
_numerify_batchid,
cdchash,
cdchash_sequence,
cdctime ,
`+v_etl_run_number+` as etl_run_number
from `+v_mds_db_name+`.`+v_audit_table_name+`_final
where original_request_id in (select distinct original_request_id
from `+v_mds_db_name+`.`+v_mds_table_name+`
where fields_changed like '%;`+v_field+`;%' and log like '%`+v_field_lookup+`%' )
and fields_changed like '%;`+v_field+`;%' and log like '%`+v_field_lookup+`%'
and `+v_sys_id_field+` not in ( select v_sys_id_field
from `+v_mds_db_name+`.`+v_mds_table_name+`
where fields_changed like '%;`+v_field+`;%' and log like '%`+v_field_lookup+`%')`;
snowflake.createStatement( { sqlText: v_sql_3} ).execute();
应该是你想要的。
我提取了
split_part(substring(log,position('`+v_field_lookup+`',log)+`+v_field_length+`),': ',2)
放入一个名为 s1 的变量中以使其更具可读性，因为它被大量重复。
答案 1 :(得分：0)
这里的问题是，与 MySQL 相比：
select IF(1,TRUE,FALSE); -- this works
但在雪花中：
select IFF(1,TRUE,FALSE); -- this fails
相反，正确的方法是
select IFF(1= 1,TRUE,FALSE); -- this works
因此，无论是 case 还是 IFF，都必须根据某个真值对其进行评估，否则编译失败
同样是上面提到的代码案例中的问题，当没有对任何事物进行两次评估时，因此出现错误