[Pgpool-hackers] Heading to pgpool-II 3.1

[Tatsuo Ishii]

Oops. I found a bug with it.
Revised patches included.
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp

>> On 24-06-11 10:15, Tatsuo Ishii wrote:
>>>> Any chance that this idea will be implemented anytime soon?
>>>> http://lists.pgfoundry.org/pipermail/pgpool-hackers/2011-May/000716.html
>>> 
>>> Revisiting it now. Since nobody proposed patches for this, I've been
>>> starting coding now:-) The rewriting code is quite complex and I am
>>> not sure whether I could finish the work by the end of this
>>> weekend. Let me try out...
>> 
>> Wow, I am impressed. I hope you will succeed.
>> 
>> As a matter of fact, a colleague of mine created a patch that handles one
>> specific case: timezone('utc'::text, now()), which is the most important
>> case for us.
>> 
>> He told me that the code is too complex for him to quickly handle the
>> generic case, so he added this specific case instead. I can't imagine the
>> patch being useful to you, but if you want to look at it, let me know.
> 
> I have finished the work so far. However I could not make it work for
> extended protocol. This means still Java programs or PHP PDO programs
> are still remain same as they were. So I am not sure includes patches
> is usefull for you or not.
> 
> What I did was, extract default expressions (such as
> timezone('utc'::text, now()) from the system catalog and get the
> result to replace a query. So you could use any expressions including
> now() for default values.
> 
> To adopt the cases for extended protocol, probably I need to modify
> bind_rewrite_timestamp.
> 
> Here are test cases:
> 
> create table t1(i int, t timestamp default timezone('utc'::text, now()),
> 	   t2 timestamp default current_timestamp);
> CREATE TABLE
> begin;
> BEGIN
> insert into t1 values(1, now());
> INSERT 0 1
> insert into t1 default values;
> INSERT 0 1
> insert into t1 values(2);
> INSERT 0 1
> insert into t1 values(3, default);
> INSERT 0 1
> insert into t1(i) values(4);
> INSERT 0 1
> update t1 set i = 100 where i is null;
> UPDATE 1
> update t1 set i = 101,t = current_timestamp where i = 100;
> UPDATE 1
> select * from t1 where i = 101;
>   i  |             t              |             t2             
> -----+----------------------------+----------------------------
>  101 | 2011-06-25 17:18:30.676202 | 2011-06-25 17:18:30.676202
> (1 row)
> 
> update t1 set i = 102,t = default where i = 101;
> UPDATE 1
> select * from t1;
>   i  |             t              |             t2             
> -----+----------------------------+----------------------------
>    1 | 2011-06-25 17:18:30.676202 | 2011-06-25 17:18:30.676202
>    2 | 2011-06-25 08:18:30.676202 | 2011-06-25 17:18:30.676202
>    3 | 2011-06-25 08:18:30.676202 | 2011-06-25 17:18:30.676202
>    4 | 2011-06-25 08:18:30.676202 | 2011-06-25 17:18:30.676202
>  102 | 2011-06-25 08:18:30.676202 | 2011-06-25 17:18:30.676202
> (5 rows)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pool_timestamp.patch
Type: text/x-patch
Size: 9395 bytes
Desc: not available
URL: <http://pgfoundry.org/pipermail/pgpool-hackers/attachments/20110627/5ed28625/attachment.bin>