On Tue, Sep 13, 2011 at 5:53 PM, Ann Harrison <aharrison@...>wrote:

> On Tue, Sep 13, 2011 at 3:45 PM, Leonardo Carneiro
> <chesterman86@...>wrote:
>
> >
> > How do i get the id of the tuple of a table? I mean the ID of the tuple
> in
> > a
> > table, not the PK of the tuple. Have i jumped any doc that describe this?
> >
> >
> The id of a tuple in a table is not a relational concept. Tuples are
> identified by column values. However, you can get the RDB$DB_KEY which is
> sort of an id. However the RDB$DB_KEY is stable only for a single
> transaction. Between transactions someone could delete one tuple with a
> specific RDB$DB_KEY and another transaction could insert a new tuple that
> could get the same RDB$DB_KEY value. Backing up and restoring a database
> completely scrambles the RDB$DB_KEY values. Use a primary key.
>
>
> Good luck,
>
> Ann
>
>
> Hi Ann, very very thank you for your answer.

I'm well aware that the good way is to use a primary key, but this case is
very special.

We have some big tables in the system that we are converting to a new format
in runtime. The old data (from the the client table, for example) is being
pumped to a temporary table with the new format (table client_tmp), and as
the new data comes in, is being inserted in the two tables (client and
client_tmp). For the sake of performance, we are not using any constraints
on the temp table, since every select is being executed on the old table.

Once all old data is pumped to the new table, the system drops the old
table, creates a new one with the same name, but with the new format AND
with constraints, and pump the data from the temp table to this new one.

It looks like the software is allowing the insertion of duplicated data.
While this will be fixed in the software side, i have to make a way to
remove this duplicated tuples in this temporary table that has no PK.

Again, tks for the tip.


[Non-text portions of this message have been removed]