Re: [Gnumed-devel] Inadvertent deletion of Procedure when Deleting episode

[Busser, Jim]

On 2013-08-02, at 2:12 AM, Jim Busser <address@hidden> wrote:

>>> so how still to account for "missing" clin.clin_root_items ????
>> 
>> Sorry but I don't yet believe in them before I have seen
>> them.

I think that I finally now (better) understand postgres inheritance.

And so while it can be technically possible to store at least *some* data in 
the rows of a parent table, as in the example at postgresql.org

        http://www.postgresql.org/docs/current/interactive/ddl-inherit.html

it is *not* necessary to do so.

In fact, in GNUmed, it was decided to prevent do so, and to utilize 
clin.clin_root_item purely for two other purposes:

1. to provide a stable data definition of columns to be inherited by child 
tables, and

2. to make easier, and more efficient, certain queries that involve specifying 
clin.clin_root_item despite that this table does not, itself directly, contain 
any rows and therefore accounting for the difference between when I query:

        SELECT COUNT(*) FROM clin.clin_root_item ;
        --> 1031 rows

versus when I query (and here note the 'ONLY'):

        SELECT COUNT(*) FROM ONLY clin.clin_root_item ;
        --> 0 rows

and so any 'loss' of values of pk_item was not because of postgres permitting 
any deletion of a row in clin.clin_root_item but, rather, that a SELECT 
performed as

        SELECT pk_item … FROM clin.clin_root_item

seemed unable to return a value of 1070 for pk_item despite that I am *pretty* 
sure (though not 100.00% positive) that this value was still in existence at 
the time, in a row in the child table clin.procedure as described here:

        http://lists.gnu.org/archive/html/gnumed-devel/2013-08/msg00002.html

-- Jim