我在尝试优化包含交叉连接的查询。我有一个大的查询,我继续与派生表交叉连接。在SQL中消除交叉连接
它是否通过将派生表转换为视图来提高查询速度?甚至在永久性表格中捕获这些信息?
这里是我的查询
SELECT VIEWER_ID,
QUESTION_ID,
ANSWER_ID,
sum(ANSWER_SCORE) AS ANSWER_SCORE_SUMMED
FROM(SELECT cr.COMMUNICATIONS_ID AS ANSWER_ID,
cr.CONSUMER_ID as VIEWER_ID,
nc.PARENT_COMMUNICATIONS_ID AS QUESTION_ID,
case when cr.CONSUMER_ID= nc.SENDER_CONSUMER_ID then 3*((24/(((UNIX_TIMESTAMP(NOW())-UNIX_TIMESTAMP(cal.LAST_MOD_TIME)+3600)/3600))*(ces.EXPERT_SCORE * cirm.CONSUMER_RATING) + (12.5 * scs.SIMILARITY)* (1 - EXP(-0.5 * (cal.TIPS_AMOUNT/ATV.AVG_TIPS)) + .15)))
else ((24/(((UNIX_TIMESTAMP(NOW())-UNIX_TIMESTAMP(cal.LAST_MOD_TIME)+3600)/3600))*(ces.EXPERT_SCORE * cirm.CONSUMER_RATING) + (12.5 * scs.SIMILARITY)* (1 - EXP(-0.5 * (cal.TIPS_AMOUNT/ATV.AVG_TIPS)) + .15)))
end as ANSWER_SCORE
FROM (SELECT 234 AS CONSUMER_ID,
ACTION_LOG_ID,
COMMUNICATIONS_ID
FROM consumer_action_log
WHERE COMM_TYPE_ID=4) AS cr
JOIN network_communications AS nc
ON cr.COMMUNICATIONS_ID=nc.COMMUNICATIONS_ID
JOIN consumer_action_log AS cal
ON cr.ACTION_LOG_ID=cal.ACTION_LOG_ID
JOIN communication_interest_mapping AS cim
ON nc.PARENT_COMMUNICATIONS_ID=cim.COMMUNICATION_ID
JOIN consumer_interest_rating_mapping AS cirm
ON cr.CONSUMER_ID=cirm.CONSUMER_ID
AND cim.CONSUMER_INTEREST_EXPERT_ID=cirm.CONSUMER_INTEREST_ID
JOIN consumer_expert_score AS ces
ON nc.SENDER_CONSUMER_ID=ces.CONSUMER_ID
AND cim.CONSUMER_INTEREST_EXPERT_ID=ces.CONSUMER_EXPERT_ID
JOIN survey_customer_similarity AS scs
ON cr.CONSUMER_ID=scs.CONSUMER_ID_2
AND cal.SENDER_CONSUMER_ID=scs.CONSUMER_ID_1
OR cr.CONSUMER_ID=scs.CONSUMER_ID_1
AND cal.SENDER_CONSUMER_ID=scs.CONSUMER_ID_2
CROSS JOIN
(
SELECT AVG(cal.TIPS_AMOUNT) AS AVG_TIPS
FROM CONSUMER_ACTION_LOG AS cal
JOIN (SELECT 234 AS CONSUMER_ID,
ACTION_LOG_ID,
COMMUNICATIONS_ID
FROM consumer_action_log
WHERE COMM_TYPE_ID=4) AS cr
ON cal.SENDER_CONSUMER_ID=cr.consumer_id
) ATV) AS ASM
GROUP BY ANSWER_ID
ORDER BY ANSWER_SCORE_SUMMED DESC;
这是一个很长的查询,所以没必要把它读完了。要点只是交叉连接。我对sql很陌生,但我被告知交叉连接会减慢速度。
交叉连接放慢速度的原因是它们返回很多很多的记录。假设你的交叉连接是一个有100条记录的表格,并且它将连接到一个包含1000条记录的数据集合,那么结果数据集合将会是100000条记录,明显需要比返回原始1000条记录更多的时间。但是如果你需要这些数据,你需要这些数据。 – HLGEM 2011-01-28 19:28:19
同意,另一个原因是它停止整个查询由rdbms引擎优化。我已经看到了甲骨文的CBO与交叉连接的斗争。 (并且通常交叉连接也意味着设计有些不对); – 2011-01-29 15:58:59