我只对oracle db 11拥有阅读权限。我的问题是我现在开发了以下查询，但是它非常慢。执行此查询最多可能需要5 min。此外，我想在将来为我的查询添加更多标签，我想这次会以指数方式增加。
SELECT distinct *
FROM (
SELECT sqOne.value_number AS Interest_Rate, sqTwo.value_number AS Equity, u.update_as_of AS REFERENZ_DATUM, u.update_dat_ins AS INSERTION_DATE, p.cust_id, p.project_name
FROM update u
JOIN project p ON p.project_id= u.project_id
JOIN customer cust ON p.CUST_ID = cust.CUST_ID
JOIN value v ON u.update_id = v.update_id
JOIN tag t ON t.tag_id = v.tag_id
-- ###############################################
-- Interest Rate
-- ###############################################
LEFT OUTER JOIN
(
SELECT va.value_number, pr.project_id
FROM update up
JOIN project pr ON pr.project_id= up.project_id
JOIN value va ON up.update_id = va.update_id
JOIN tag ta ON ta.tag_id = va.tag_id
WHERE ta.tag_desc LIKE 'IR%' AND va.value_number > 0 AND va.value_number IS NOT NULL
) sqOne
ON sqOne.project_id= p.project_id
-- ###############################################
-- Equity
-- ###############################################
LEFT OUTER JOIN
(
SELECT va.value_number, pr.project_id--return the tic which is the latest!!! TODO
FROM update up
JOIN project pr ON pr.project_id= up.project_id
JOIN value va ON up.update_id = va.update_id
JOIN tag ta ON ta.tag_id = va.tag_id
WHERE ta.tag_desc LIKE 'EQUITY%' AND va.value_number > 0 AND va.value_number IS NOT NULL
) sqTwo
ON sqTwo.project_id= p.project_id
WHERE p.project_stat = 'A'
ORDER BY p.project_id
);
在这里，我尝试使用select *查询重建表定义：
select * from update; +--------------------------------------------------+ | UPDATE_ID UPDATE_AS_OF UPDATE_DAT_INS PROJECT_ID | +--------------------------------------------------+ | 1 17-DEC-12 17-DEC-12 34 | | 2 17-DEC-12 17-DEC-12 31 | | 14 17-DEC-12 17-DEC-12 29 | | 21 17-DEC-12 18-DEC-12 2 | | 22 17-DEC-12 18-DEC-12 2 | | 25 18-DEC-12 18-DEC-12 66 | +--------------------------------------------------+
select * from project; +----------------------------------------------------------------------+ | PROJECT_NAME PROJECT_DESC CUST_ID PROJECT_ID | +----------------------------------------------------------------------+ | Test null 654321 21 | | test1 null 654321 39 | | test2 null 123456 76 | | test null 123456 90 | | Usability null 323423 158 | | 5. Test null 654321 181 | | fsf null 123456 361 | | Test 100 null 123456 425 | +----------------------------------------------------------------------+
select * from customer; +-----------------------------------------------------+ | CUST_ID CUST_DAT_INS CUST_DAT_UPD CUST_NAME | +-----------------------------------------------------+ | 1 13-DEC-12 01-MAR-14 abc1 | | 2 13-DEC-12 30-OCT-13 abc2 | | 3 13-DEC-12 30-OCT-13 abds | | 4 13-DEC-12 30-OCT-13 test1 | | 5 13-DEC-12 30-OCT-13 test23 | +-----------------------------------------------------+
select * from value; +---------------------------------------------------------------------------------+ | VALUE_ID VALUE_NUMBER VALUE_DAT_INS TAG_ID PROJECT_ID UPDATE_ID | +---------------------------------------------------------------------------------+ | 1 null 17-DEC-12 118 34 1 | | 2 5000 17-DEC-12 48 34 1 | | 3 8 17-DEC-12 73 34 1 | | 4 null 17-DEC-12 187 34 1 | | 5 null 17-DEC-12 123 34 1 | +---------------------------------------------------------------------------------+
select * from tag; +------------------------------------------------------------+ | TAG_ID TAG_CODE TAG_DESC TAG_DAT_INS | +------------------------------------------------------------+ | 1 EQU EQUITY 13-DEC-12 | | 2 IR Interest Rate 13-DEC-12 | +------------------------------------------------------------+
感谢您就如何优化我的查询以加快加载速度提供答案！
更新
这是explain for查询的更新：
---------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost |
---------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 2600 | 162K| 2747 |
| 1 | HASH UNIQUE | | 2600 | 162K| 2747 |
| 2 | HASH JOIN | | 2600 | 162K| 2746 |
| 3 | VIEW | VW_DTP_104D9B21 | 4191 | 20955 | 623 |
| 4 | HASH UNIQUE | | 4191 | 20955 | 623 |
| 5 | INDEX FAST FULL SCAN | VALUE_IDX4 | 531K| 2594K| 590 |
| 6 | HASH JOIN | | 2600 | 149K| 2123 |
| 7 | HASH JOIN RIGHT OUTER | | 495 | 16830 | 2109 |
| 8 | VIEW | | 483 | 2898 | 1050 |
| 9 | NESTED LOOPS | | 483 | 19320 | 1050 |
| 10 | HASH JOIN | | 765 | 27540 | 1050 |
| 11 | NESTED LOOPS | | 4191 | 102K| 14 |
| 12 | TABLE ACCESS BY INDEX ROWID| TAG | 1 | 16 | 1 |
| 13 | INDEX UNIQUE SCAN | TAG_UDX1 | 1 | | 0 |
| 14 | TABLE ACCESS FULL | UPDATE | 4191 | 37719 | 13 |
| 15 | TABLE ACCESS FULL | VALUE | 123K| 1331K| 1035 |
| 16 | INDEX UNIQUE SCAN | PROJECT_PK | 1 | 4 | 0 |
| 17 | HASH JOIN RIGHT OUTER | | 495 | 13860 | 1058 |
| 18 | VIEW | | 242 | 1452 | 1052 |
| 19 | NESTED LOOPS | | 242 | 12826 | 1052 |
| 20 | HASH JOIN | | 383 | 18767 | 1052 |
| 21 | MERGE JOIN CARTESIAN | | 2096 | 79648 | 16 |
| 22 | TABLE ACCESS FULL | TAG | 1 | 29 | 3 |
| 23 | BUFFER SORT | | 4191 | 37719 | 13 |
| 24 | TABLE ACCESS FULL | UPDATE | 4191 | 37719 | 13 |
| 25 | TABLE ACCESS FULL | VALUE | 123K| 1331K| 1035 |
| 26 | INDEX UNIQUE SCAN | PROJECT_PK | 1 | 4 | 0 |
| 27 | TABLE ACCESS FULL | PROJECT | 495 | 10890 | 5 |
| 28 | TABLE ACCESS FULL | UPDATE | 4191 | 102K| 13 |
---------------------------------------------------------------------------------------
答案 0 :(得分：2)
从我坐的地方看来，这个问题看起来很简单。以下事情似乎是无用的代码：
customer query ...在该查询级别的任何地方都没有使用customer表中的数据。value query ...在该查询级别的任何地方都没有使用value表中的数据。tag query ...在该查询级别的任何地方都没有使用tag表中的数据。project和sqOne中的sqTwo ... pr.project_id都可以选为up.project_id。va.value_number IS NOT NULL和sqOne中的sqTwo谓词...谓词va.value_number > 0自动隐含两者。order by子句＆＃34; ......由于distinct在＆＃34;等级1＆＃34;。value和tag联接之后，select distinct ...不需要因为我怀疑那两个联接（沿着我怀疑sqOne / sqTwo仅在project_id而不是update_id上加入SELECT distinct *
FROM (
SELECT sqOne.value_number AS Interest_Rate, sqTwo.value_number AS Equity, u.update_as_of AS REFERENZ_DATUM, u.update_dat_ins AS INSERTION_DATE, p.cust_id, p.project_name
FROM update u
JOIN project p ON p.project_id = u.project_id
-- Interest Rate
LEFT OUTER JOIN (
SELECT va.value_number, up.project_id
FROM update up
JOIN value va ON up.update_id = va.update_id
JOIN tag ta ON ta.tag_id = va.tag_id
WHERE ta.tag_desc LIKE 'IR%' AND va.value_number > 0
) sqOne
ON sqOne.project_id = p.project_id
-- Equity
LEFT OUTER JOIN (
SELECT va.value_number, up.project_id --return the tic which is the latest!!! TODO
FROM update up
JOIN value va ON up.update_id = va.update_id
JOIN tag ta ON ta.tag_id = va.tag_id
WHERE ta.tag_desc LIKE 'EQUITY%' AND va.value_number > 0
) sqTwo
ON sqTwo.project_id = p.project_id
WHERE p.project_stat = 'A'
-- the following supplements the original inner joins with `tag` and `value`; those two originally left the result with only those `product` and `update` rows which had at least some tags and values
AND (sqOne.project_id is not null or sqTwo.project_id is not null)
);
/ {{1}}的错误加入了＆＃34;等级2＆ ＃34;查询。但我可能也错了，你必须自己测试一下。因此，在此基本代码清理之后，查询可能如下所示：
{{1}}
请尝试，如果它甚至运行（我可能已经删除了一些重要的东西），如果它做了你需要它做的事情，如果它运行得更快或者没有，如果没有那么发布经过的时间它的执行和新的执行计划。
答案 1 :(得分：0)
您可以尝试重新设计查询。
1）您不需要2个子查询。由于连接条件在主查询和子查询中是相同的，我们可以在那里计算兴趣和权益，例如感兴趣的是在基本查询本身中使用以下语句
(CASE WHEN t.tag_desc LIKE 'IR%' AND v.value_number > 0 AND v.value_number IS NOT NULL THEN v.value_number ELSE NULL END) AS Interest_Rate
同样写入权益并删除2个子查询。
2）只需检查是否需要使用DISTINCT
希望它有所帮助。