SQL Joins Taking forever on table with 60 matches

问题

I have a query that makes joins between 4 tables.

Matching my input params on a table by table basis yeilds the following

TrialBal - 8 million records matching entity and pl_date

Join to ActDetail - Execution is about 85 secs, the row count is 8672175 (with group by, the row count is 1...for now). ActDetail would return zero rows on an inner join.

Join to CalendarEngine - there are only 60 matchig records in this table (pl_date & entity), but when this is introduced to the SQL, the query just runs and runs. Is there any particular reason for this?

CalendarEngine has a composite index on entity and one other field not used in my query, so I should still be able to use it?

Why would the join between TrailBal and ActDetail be relatively quick (presumably the Outer join is running through all 8 million records against the count of 168k entries in ActDetail), but joining on 60 records in CalendarEngine leads to a seemingly endless loop.

SQL below (I need the commented out fields in the final query). Also tried Replacing the OR statement with 2 Outer Joins

SELECT 
    ad.acct_lv2 AS IFRS_lvl2,
    ad.acct_lv3 AS IFRS_lvl3,
    ad.acct_lv4 AS IFRS_lvl4,
    ad.acct_lv6 AS IFRS_lvl6,
    --(CASE WHEN tb.pl_date = cal.curr_me_date THEN cal.date END) as Reporting_Cycle_This,
    --(CASE WHEN tb.pl_date = cal.prev_me_date THEN cal.date END) as Reporting_Cycle_Last,
    --(CASE WHEN tb.pl_date = cal.prev_me_date THEN cal.date END) as Reporting_Cycle_Previous,
    null as Total,
    null as Balance_default_dim,
    null as Balance_proxy_dim,
    0 as percentage_bal_proxy_applied,
    '1 Table Data' as rowType 
FROM TrialBal tb 
INNER JOIN Partition p ON p.partition_code = tb.partition_code
    AND p.entity = tb.entity
LEFT OUTER JOIN ActDetail ah ON tb.actNum = ad.actNum
    AND tb.cpny = ad.cpny
    AND tb.pl_date = ad.pl_date
    AND ad.source = 'Ground0'
--LEFT OUTER JOIN CalendarEngine cal on tb.pl_date = cal.curr_me_date OR tb.pl_date = cal.prev_me_date
    --AND tb.entity = cal.entity
WHERE tb.pl_date = '2014-04-30'
AND tb.entity = 'My_entity'
GROUP BY ad.acct_lv2, ad.acct_lv3, ad.acct_lv4, ad.acct_lv6, --tb.actNum, tb.cpny, 
    CASE WHEN tb.pl_date = cal.curr_me_date THEN cal.date END,
    CASE WHEN tb.pl_date = cal.prev_me_date THEN cal.date END,
    CASE WHEN tb.pl_date = cal.prev_me_date THEN cal.date END

Also tried Replacing the OR statement with 2 Outer Joins:

LEFT OUTER JOIN sCalendar cal_cur on tb.pnl_date = cal_cur.curr_me_date
    AND tb.entity_code = cal_cur.entity_code
LEFT OUTER JOIN sCalendar cal_pre on tb.pnl_date = cal_pre.prev_me_date
    AND tb.entity_code = cal_pre.entity_code

回答1:

Is there an index on cal.curr_me_date and tb.pl_date ? When I have had similar issues I went looking for indexes on my joined fields and that was the fix.

回答2:

Why are you doing this join

LEFT OUTER JOIN ActDetail ah ON tb.actNum = ad.actNum
AND tb.cpny = ad.cpny
AND tb.pl_date = ad.pl_date
AND ad.source = 'Ground0'

I dont see any use, of the column, remove unnecesary joins.

if this part is date

tb.pl_date = '2014-04-30'

i suggest use

 pl_date >= dateadd(d,0,datediff(d,0,pl_date)) and pl_date < dateadd(d,0,datediff(d,0,pl_date+1))

regards

来源：https://stackoverflow.com/questions/23936271/sql-joins-taking-forever-on-table-with-60-matches

标签

sql

join

sybase

large-data