sqlserver去重

select count(*) as H from (select user_id, ROW_NUMBER() OVER (PARTITION BY tb.user_id order by tb.id) rn from tb where user_id>0) T where rn =1