Tarlika Elisabeth Schmitz wrote:
> I have a database that will be populated solely by CSV import.
> There are several CSV file formats, all denormalized.
>
> I have created interim tables which match the CSV file formats. An
> insert trigger distributes the data to their appropriate destination
> tables. The destination tables themselves have insert/update triggers
> for automated data clean-up. Any unresolvable inconsistencies are
> reported in a log table.
>
> I don't want the triggers to fire for every insert/update. There might
> be situations where I have to perform some data clean-up manually.
>
> So, my idea is to create a role for import, query current_user in the
> trigger, perform the trigger actions for importuser and just return the
> row unadulterated for adminuser.
>
> I would give privileges to the importuser for the tables being
> explicitly and implicitly populated.
>
> Is that the best way to organize this?
>
>
> =====
> setup: PostgreSQL 8.4
> dbname = schema name = admin name
>
>
>
You seem to be writing denormalized import records for the sole purpose
of writing other normalized records. Have you you looked into writing a
programme in a relatively high-level jdbc-friendly language which reads
the csv file, normalizes the data (the code already in your triggers)
and flushes on every say 1000 independent records? The "clean-up" and
logging might also be done by the import app (all depending on what's
being cleaned up and logged :) )