Re: UPDATE with data at exec and CURRENT OF question

Hi.

From: <lothar(dot)behrens(at)lollisoft(dot)de>

> On 9 Aug., 10:43, z-sa(dot)(dot)(dot)(at)guitar(dot)ocn(dot)ne(dot)jp ("Hiroshi Saito") wrote:
>> Hi.
>>
>> Sorry, very late reaction.
>> > I am still using 07.03.0200 on Mac OS X. I didn't get compiled the
>> > 07.03.0260 version due to missing
>> > pg_config.
>>
>> Can't you use the series of 08.02.xx, either?
>> I think It should be cleared if there are some problems.
>>
>
> My application and my ODBC wrapper class relates to full cursor
> functionality (first,previous,next,last). So I only know about 7.3.2xx
> drivers.

Um, We have 08.02.xxx on the extension. Surely it had a time of falling
out by some confusion.... However, the current version supports it.
It should be corrected if there are some problems. and, support is
short-handed up to 07.x past versions.....I want to clear a problem
by the present version as much as possible.

>
>> > What's wrong ?
>>
>> However, I can't immediately check the following code...
>> Slight time is needed. I appreciate your perseverance.
>>
>
> Now I have a better working code. It is able to put data in 1000 byte
> pieces, but the driver
> (SQLParamData(...)) does not indicate, that it needs more data. Also
> the getBinaryData
> function returns no more data, that the size of the buffer (here
> 5000).
>
> I could increase the buffer to the buffer in my value object, but I
> probably could not go
> ahead the maximum of available memory.
>
> So I tried all to put the data in pieces of, say, 1000 or 5000 bytes,
> both functions do not
> behave as expected from Microsoft documentation from here:
>
> http://msdn2.microsoft.com/en-us/library/ms713824.aspx
>
> I have tried to ignore SQL_NEED_DATA and have calculated a remaining
> size to be handled
> until it decreases to the piece size to put the rest. But then the
> result behaves like storing rubbish.
>
> The lb_I_BinaryData class is a container where I could append pieces
> and I think it is unnesesary
> to copy that code too.
>
> I have tried all I can get from documentation. I have reached the end
> of my brain :-)
>
> Hope you will see any wrong API usage or parameters. But I assume,
> there is a bug in
> the 07.03.0200 driver.
>
> Actually I do no more use CURRENT OF <cursor name>, because PostgreSQL
> doesn't support
> that syntax (at least in my old 7.4 database).
>
> I do a trick by first bind the binary column and put a unique marker
> to that column for later update
> filtering. I hope the cursorname is unique. I probably have to make
> this function thread save !

Um, I want to separate problem. It is the right motion of a cursor action.
Then, data-column is divided and treated by chunk... Probably, we need
the test code. I will work by it... However, I want to solve it by 08.02.0402.

>
> Sorry for the big code.
>
> Thanks, Lothar

Ahh, sorry, It seems that can't reproduce for the environment problem.

P.S)
Please understand that I can't solve early like Inoue-san.

Regards,
Hiroshi Saito