how to insert random records without violating primary constraint - sql

Question

Please pardon my title if misleading, but I have a table tb_party with PRIMARY KEY on party_key
I have following query -
insert TMS..tb_party
(
[party_key]
,[party_first_name]
,[tax_id]
,[party_type_cd]
,[citizenship_country_cd]
,[domicile_country_cd]
,[party_num]
,[batch_dt]
)
select distinct
[party_key]
,[party_first_name]
,[tax_id]
,[party_type_cd]
,[cit]
,[dom]
,[party_num]
,[batch_dt]
from
(select distinct
sca.[party_key]
,sca.[party_first_name]
,sca.[tax_id]
,pt.party_type_cd
,tc_cit.COUNTRY_ID as cit
,tc_dom.COUNTRY_ID as dom
,sca.[party_num]
,getdate() as batch_dt
,dense_rank() over(partition by sca.[party_key] order by sca.party_key) as rnk
from Iteration_3.dbo.staging_cust_acct sca (nolock)
join Iteration_3..STG_PARTY_UPLOAD (nolock) stg_party
on sca.party_key = stg_party.PARTY_KEY
left join Iteration_3..STG_COUNTRY_ISO tc_dom (nolock)
on sca.[domicile_country] = tc_dom.COUNTRY_NAME
left join Iteration_3..STG_COUNTRY_ISO tc_cit (nolock)
on sca.[citizenship_country] = tc_cit.COUNTRY_NAME
left join TMS..tb_party_type pt (nolock)
on sca.[party_type] = pt.party_type_desc
WHERE
SCA.party_type IS NOT NULL
) x
where rnk = 1
The insert fails because it is trying to insert duplicate party key and since the distinct is on ALL columns, it is picking up duplicate party_keys.
What I want - I want to pick up all distinct party_keys and insert 1 row in tb_party. the other rows can be ignored. Is this possible?

Answer 1

Here is where row_number came in handy.
select distinct
[party_key]
,[party_first_name]
,[tax_id]
,[party_type_cd]
,[cit]
,[dom]
,[party_num]
,[batch_dt]
from
(
select distinct
sca.[party_key]
,sca.[party_first_name]
,sca.[tax_id]
,pt.party_type_cd
,tc_cit.COUNTRY_ID as cit
,tc_dom.COUNTRY_ID as dom
,sca.[party_num]
,getdate() as batch_dt
--,dense_rank() over(partition by sca.[party_key] order by sca.party_key,sca.[tax_id],sca.[party_num],sca.[party_first_name]*/) as rnk
,row_number() over (partition by sca.party_key order by sca.party_key) rn
from Iteration_3.dbo.staging_cust_acct sca (nolock)
join Iteration_3..STG_PARTY_UPLOAD (nolock) stg_party
on sca.party_key = stg_party.PARTY_KEY
left join Iteration_3..STG_COUNTRY_ISO tc_dom (nolock)
on sca.[domicile_country] = tc_dom.COUNTRY_NAME
left join Iteration_3..STG_COUNTRY_ISO tc_cit (nolock)
on sca.[citizenship_country] = tc_cit.COUNTRY_NAME
left join TMS..tb_party_type pt (nolock)
on sca.[party_type] = pt.party_type_desc
WHERE --SCA.update_source = 'ECM' AND SCA.update_dt = #ECM_MAX_DATE
--and
SCA.party_type IS NOT NULL
) x
where rn = 1

Answer 2

Try merging into your target table on the unique columns. When not matched, insert. When matched update like this:
...
set target.[column] = coalesce(source.[column], target.[column])
...
This way you update all your target columns when you have source values, otherwise they remain unchanged.)