data-dwh-dbt-project/macros/cosmos_db_utils.sql
2024-09-19 15:17:33 +02:00

31 lines
1.2 KiB
SQL

{% macro cosmos_db_record_deduplication(source_table, primary_key) %}
/*
This macro provides a deduplication logic for Cosmos DB entities.
If two or more records have a duplicated value in a field that acts as
the primary key, the most recent record will be returned. If the record
is not duplicated, it will also be returned.
Inputs:
- source_table: table that acts as source. Should be a sync model.
- primary_key: unique identifier on which the deduplication will be applied.
Output:
- Returns the set of records from the source_table that are unique according
to the primary_key.
*/
select *
from
-- Some thoughts for the future here:
-- The query below is awful performance wise - but data
-- size is tiny today. Let's tackle the problem as it comes.
(
select
*,
row_number() over (
partition by
{{ adapter.quote("documents") }} ->> '{{ primary_key }}'
order by ({{ adapter.quote("documents") }} ->> '_ts')::integer desc
) as rank
from {{ source_table }}
)
where rank = 1
{% endmacro %}