Hello Everyone:
I am new to PostgreSQL and I need some advice from the experts. We are
writing a trucking application in X-Windows with a PostgreSQL back end.
The application will support about 70 users and about 60% of those users
are doing a lot of querying and updating of existing records. New data
is entered daily, but the bulk of the database work is in updating
existing records as they "move" through various stages of the system. We
are using C with libpq-fe to access the database. The application
consists of many tables and interfaces that allow users to populate
certain text widgets with criteria on which they want to search for
records. We are loading condensed versions of the data that matches the
search criteria into a XbaeMatrix widget. If the user decides to
"detail" one of the records. we find that record again and populate a
different screen with the complete set of data. If they decide to edit
the record. we start a transaction, open a cursor, and find the record
again, redisplay it, and enable various widgets for edit. If they abort
the edit, we roll back and restore the interface to the pre edit mode.
If they save the record, we commit the transaction and restore to the
pre edit mode. Given this scenario, here are my questions.
1. Does this model make sense based on your work with libpq-fe and
postgres?
2. We are only using a binary cursor for the edit operation. Does
this make sense?
3. Regarding MVCC and postgres locking model, does the above scenario
seem reasonable?
I hope the above rambling makes sense. Thanks for any advice anyone can
provide :o)
--
Sparta, NC 28675 USA
336.372.6812
http://www.esc1.com