Improve profiler performance in Snowflake by reducing table scans usi…

…ng a single CTE from which column profiles are calculated (#48)

Co-authored-by: Simo Tumelius <[email protected]>

.gitignore

@@ -1,5 +1,6 @@
 target/
 dbt_modules/
+dbt_packages/
 logs/
 
 venv/

macros/get_profile.sql

@@ -50,7 +50,13 @@
   {{ log("Column data types: " ~ data_type_map, info=False) }}
 
   {% set profile_sql %}
-    with column_profiles as (
+    with source_data as (
+      select
+        *
+      from {{ relation }}
+    ),
+
+    column_profiles as (
       {% for column_name in profile_column_names %}
         {% set data_type = data_type_map.get(column_name.lower(), "") %}
         select 
@@ -88,7 +94,7 @@
           {%- endif %}
           cast(current_timestamp as {{ dbt_profiler.type_string() }}) as profiled_at,
           {{ loop.index }} as _column_position
-        from {{ relation }}
+        from source_data
 
         {% if not loop.last %}union all{% endif %}
       {% endfor %}