这个语句好怪，请高手帮忙看看

rollingpig

try
SELECT /*+no_index(fw_khhfdb INX_FW_KHHFDB_FWDWID ) */COUNT (*)
FROM fw_khhfdb
WHERE hfrq >= TO_DATE ('2005-10-01', 'yyyy-MM-dd')
AND hfrq <= TO_DATE ('2005-10-13', 'yyyy-MM-dd')
AND (fwdwid =559 or fwdwid = 1119)


or

SELECT /*+no_index(fw_khhfdb INX_FW_KHHFDB_FWRQ ) */COUNT (*)
FROM fw_khhfdb
WHERE hfrq >= TO_DATE ('2005-10-01', 'yyyy-MM-dd')
AND hfrq <= TO_DATE ('2005-10-13', 'yyyy-MM-dd')
AND (fwdwid =559 or fwdwid = 1119)

cyr1974

SELECT COUNT (*)
FROM fw_khhfdb
WHERE hfrq >= TO_DATE ('2005-10-01', 'yyyy-MM-dd')
AND hfrq <= TO_DATE ('2005-10-13', 'yyyy-MM-dd')
AND fwdwid =559
union all
SELECT COUNT (*)
FROM fw_khhfdb
WHERE hfrq >= TO_DATE ('2005-10-01', 'yyyy-MM-dd')
AND hfrq <= TO_DATE ('2005-10-13', 'yyyy-MM-dd')
AND
fwdwid = 1119

zlxzm

最初由 zlxzm 发布
[B]这个语句其实没什么大问题，我发现，原因是前几天，对表和索引进行了一次分析，得到统计信息，这些信息让oracle改变了执行计划，不是改善，反而改差了，今天中午，删掉统计信息后，一执行马上得到结果。真想不到啊。 [/B]

谢谢大家的回复，这个语句目前运行良好，执行计划中已经使用了索引。看来，解决性能问题，实际的效果是最有说服力的，理论和实际不一定完全吻合，毕竟机器的智能还没有达到较高层次。

zenghuangmao

哦!t

davidpan

最初由 rollingpig 发布
[B]try
SELECT /*+no_index(fw_khhfdb INX_FW_KHHFDB_FWDWID ) */COUNT (*)
FROM fw_khhfdb
WHERE hfrq >= TO_DATE ('2005-10-01', 'yyyy-MM-dd')
AND hfrq <= TO_DATE ('2005-10-13', 'yyyy-MM-dd')
AND (fwdwid =559 or fwdwid = 1119)


or

SELECT /*+no_index(fw_khhfdb INX_FW_KHHFDB_FWRQ ) */COUNT (*)
FROM fw_khhfdb
WHERE hfrq >= TO_DATE ('2005-10-01', 'yyyy-MM-dd')
AND hfrq <= TO_DATE ('2005-10-13', 'yyyy-MM-dd')
AND (fwdwid =559 or fwdwid = 1119) [/B]

=========================
rollingping is right.
try rollingping method.
according the good explain plan,  you didn't need to use the index fwdwid.