On 27 January 2011 09:53, Santosh Bhujbal (sabhujba) <sabhujba@cisco.com> wrote:
> Hi Thom,
>
> Thank you for your response.
>
> I have a application which is periodically gathering diff stats from
> diff devices and put them into database.
> Tables are created per stat, per device and per day.
> e.g. stat1_dev1_20100125, stat1_dev1_20100126, stat1_dev1_20100127, etc.
> stat1_dev2_20100125, stat1_dev2_20100126, stat1_dev2_20100127, etc.
> stat2_dev1_20100125, stat2_dev1_20100126, stat2_dev1_20100127, etc.
> stat2_dev2_20100125, stat2_dev2_20100126, stat2_dev2_20100127, etc.
>
> Now when I am upgrading my application with new version then there are
> some tables which are having some additional columns.
> In this case I have to alter each and every old tables in database with
> new column and it's default value.
> As there are large number of tables, the upgrade process is taking too
> much time (in days).
>
> To avoid above upgrade process I want to write a SQL statements such
> that it take care of newly added columns.
You want to refer to a column which doesn't exist, but PostgreSQL
expects you to know what is available beforehand. Is there any logic
to which tables have the additional column and which ones don't? For
example, do all tables with the additional column have a name
containing a date after a certain point in time?
--
Thom Brown
Twitter: @darkixion
IRC (freenode): dark_ixion
Registered Linux user: #516935