In the application I am maintaining there is a practice of storing the newly inserted identity in a special global temporary table for use from within the procs or scripts. The way they retrieve the stored value is this:

EXEC ('DECLARE id CURSOR FOR select id from ' + @identity_table_name + ' WHERE table_name = ' + "'" + @table_name + "'")

OPEN id
FETCH id INTO @some_id
DEALLOCATE id

Does not make sense for me. Why may they want to build up this stuff instead of a simple

EXEC ('SELECT @some_id = id from ' + @identity_table_name + ' WHERE table_name = ' + "'" + @table_name + "'")

??? In the rest of the code they strictly avoid using cursors and stick with SELECT TOP 1 instead.

2All:

The reason for still using this practice is that it is a corporate standard. You all know that it is like a concrete wall... You can not break it, you can only drive along it.

The reason for storing @@identity in a ## table is that they have i, u & d triggers in each and every table and those triggers populate other tables with their own identities and they also call procs that do the same thing.

My concern is about the way they retrieve that one value from that table - why necessarily using a cursor where a simple select would do?

PS: Sorry for writing in English - I can't get ahold of w2k CD to install ru locale... And I hate translit.