GROUP BY 中的列太多

Question

I'm trying to aggregate some data, but I've a problem. There's my query (using 3 tables):

SELECT
            ufc.counter_id,
            gcrvf.goal_id,
            gcrvf.date_of_visit,
            ufc.utm_campaign,
            ufc.utm_source,
            ufc.utm_medium,
            ufc.utm_content,
            ufc.utm_term,
            ufc.original_join_id,
            max(gcrvf.last_update_time) AS last_update_time,
            sum(gcrvf.conversions) AS conversions, 
            c.name AS counter_name,
            c.owner_login AS owner_login,
            c.status AS counter_status,
            concat(g.goal_source,CAST('Goal','text')) AS metric_type,
            multiIf(g.is_retargeting = 0,'non-retargeting',g.is_retargeting = 1,'retargeting',NULL) AS metric_key,
            concat(g.name,' (',CAST(gcrvf.goal_id,'String'),')') AS metric_name
        FROM connectors_yandex_metrika.goal_conversions_report_v_final AS gcrvf
        INNER JOIN connectors_yandex_metrika.utm_for_collect AS ufc ON gcrvf.counter_id = ufc.counter_id
        LEFT JOIN connectors_yandex_metrika.counter AS c ON gcrvf.counter_id = c.id
        LEFT JOIN connectors_yandex_metrika.goal AS g ON gcrvf.goal_id = g.id
        WHERE 
            ((gcrvf.utm_campaign = ufc.utm_campaign) OR (ufc.utm_campaign IS NULL)) 
            AND ((gcrvf.utm_source = ufc.utm_source) OR (ufc.utm_source IS NULL)) 
            AND ((gcrvf.utm_medium = ufc.utm_medium) OR (ufc.utm_medium IS NULL)) 
            AND ((gcrvf.utm_content = ufc.utm_content) OR (ufc.utm_content IS NULL))
            AND ((gcrvf.utm_term = ufc.utm_term ) OR (ufc.utm_term IS NULL))
        GROUP BY
            ufc.counter_id,
            gcrvf.date_of_visit,
            gcrvf.goal_id,
            ufc.utm_campaign,
            ufc.utm_source,
            ufc.utm_medium,
            ufc.utm_content,
            ufc.utm_term,
            ufc.original_join_id,
            c.name,
            c.owner_login,
            c.status,
            metric_type,
            metric_key,
            metric_name 

I have to GROUP BY by almost all columns. Is it a real problem?

Columns ufc.original_join_id , c.name,c.owner_login , c.status , metric_type , metric_key,metric_name are not necessary here. I added them to group by just because I need these columns. And I want to ask: any way to make it more abbreviated? Any ways to avoid unnecessary columns from group by? Or it's okay?

And my second question: does ClickHouse cache right table when we use JOINs? So I always should put huge table as left table?

Answer 1

All columns are required in the group by. It is not possible to leaf some columns out which where mentioned as select columns.

Depending on your indexed columns you can improve the speed of the query. You should try to make an index on the key columns.

The Database will handle the cache logic for you. Depending on how often you execute the query.