[pgpool-hackers: 3711] Re: Query mis cached

[Tatsuo Ishii]

>> Hi pgpool hackers,
>> 
>> 
>> 
>> I found the following SQL will be cached, when enabled memory query cache,
>> 
>> 
>> 
>>         "with cte as (insert into test2 values(1) returning *) select * from test2;"
>> 
>> 
>> 
>> I think the SQL has Data-Modifying Statements in with clause should not be cached.
>> 
>> Because Once cached, the Data-Modifying Statements will not be executed again which is not expected.
> 
> You are right.
> 
>> It seems to be better to analyze the with clause like the following:
>> 
>> 
>> 
>> --------For UPDATE/INSERT/DELETE type SQL:
>> 
>>                 extract oids from its with clause which will be invalidated.
>> 
>> 
>> 
>> --------For SELECT type SQL:
>> 
>>                 If Data-Modifying Statements in its with clause, the select will not be cached,
>> 
>>                 and the oids from with clause will be invalidated too.
>> 
>> 
>> 
>> What do you think?
>> 
>> 
>> 
>> I try to solve the problem and made a patch for the code.
> 
> Thank you for the patch! I will look into this.

The patch looks good except following warnings:

git apply ~/0002-Update-pool_memqcache.c.patch 
/home/t-ishii/0002-Update-pool_memqcache.c.patch:165: trailing whitespace.
				/* 
/home/t-ishii/0002-Update-pool_memqcache.c.patch:175: trailing whitespace.
					POOL_SETMASK(&oldmask);					
warning: 2 lines add whitespace errors.

Also I think it'd be better to add a regression test for
this. Probably add to 006.memqcache?

Best regards,
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese:http://www.sraoss.co.jp