BUG #13849: Need a parameter added similar to "edb_stmt_level_tx" - Mailing list pgsql-bugs

From skinneda@us.ibm.com
Subject BUG #13849: Need a parameter added similar to "edb_stmt_level_tx"
Date
Msg-id 20160105222215.989.32575@wrigleys.postgresql.org
Whole thread Raw
Responses Re: BUG #13849: Need a parameter added similar to "edb_stmt_level_tx"
List pgsql-bugs
The following bug has been logged on the website:

Bug reference:      13849
Logged by:          David Skinner
Email address:      skinneda@us.ibm.com
PostgreSQL version: 9.5rc1
Operating system:   Windows
Description:

To add support for PostgreSQL, we need the equivalent of EDB's
edb_stmt_level_tx parameter, which you can see explained here:
http://www.enterprisedb.com/docs/en/8.4/oracompat/Postgres_Plus_Advanced_Server_Oracle_Compatibility_Guide-04.htm

We've read all about the new "ON CONFLICT" keyword in 9.5, but coding it
into our complex ECM product would be kludgy and an incomplete fix.  Our use
case (for example) is adding a folder, and preventing a duplicate key error
on the (enforced unique) containment name.

Rather than rollback and retry the entire txn, we only need to catch the SQL
error and retry the INSERT (which failed on dup. key error).  There are
similar situations in over 50 places in our code that relies on NOT rolling
back the txn on SQL error.  NOT rolling back is the standard behavior on
Oracle, SQL Server, Informix and DB2.  We understand the reasons for PG's
current behavior, but it is not practical in the complex real world,
especially since it should be extremely simple to support as EDB has.
Thanks!  Also, when is 9.5 GA expected?

pgsql-bugs by date:

Previous
From: NBoardman@idtus.com
Date:
Subject: BUG #13853: initdb to UNC path
Next
From: Paul
Date:
Subject: Re: BUG #13846: INSERT ON CONFLICT consumessequencersonconflicts