On Sun, Feb 28, 2010 at 10:23 PM, Terry <td3201@gmail.com> wrote:
> On Sun, Feb 28, 2010 at 7:12 PM, John R Pierce <pierce@hogranch.com> wrote:
>> Terry wrote:
>>>
>>> One more question. This is a pretty decent sized table. It is
>>> estimated to be 19,038,200 rows. That said, should I see results
>>> immediately pouring into the destination table while this is running?
>>>
>>
>> SQL transactions are atomic. you wont' see anything in the 'new' table
>> until the INSERT finishes committing, then you'll see it all at once.
>>
>> you will see a fair amount of disk write activity while its running. 20M
>> rows will take a while to run the first time, and probably a fair amount of
>> memory.
>
> This is working very well. The initial load worked great. Took a
> little while but fine after that. I am using this:
> INSERT INTO client_logs SELECT * FROM clients_event_log as t1 where
> t1.ev_id > (select max(t.ev_id) from client_logs as t);
>
> However, I got lost in this little problem and overlooked another. I
> need to convert the unix time in the ev_time column to a timestamp. I
> have the idea with this little bit but not sure how to integrate it
> nicely:
> select timestamptz 'epoch' + 1267417261 * interval '1 second'
>
I love overcomplicating things:
SELECT *,to_timestamp(ev_time) FROM clients_event_log as t1 where
t1.ev_id > (select max(t.ev_id) from client_logs as t)
