Comments on: OSCON 2006: TimeTravel Tables in PostgreSQL

By: s21825

s21825 — Tue, 08 Aug 2006 01:10:15 +0000

I’ve done this type of thing in the past by maintaining a ‘revisions’ table containing a timestamp and a log message. Then each other table in the database contains a ‘revision_id’ column and a ‘deleted’ column. Then you only insert new rows, no actual deletes or updates. You can then join any table to the revisions table to go back in time (based on the timestamp in the revisions table). Interesting to see a different approach to solving the same problem!

By: malikyte

malikyte — Mon, 07 Aug 2006 02:19:38 +0000

Thanks, Elein!

By: elein

elein — Mon, 07 Aug 2006 01:45:32 +0000

The URL http://www.varlena.com/GeneralBits/Tidbits contains the current presentation *along with* the code used to generate the examples.

The primary key of the timetravel table is the id + starttime. Look at the unique indexes to see how this is formed.

elein@varlena.com

By: malikyte

malikyte — Sat, 05 Aug 2006 18:05:44 +0000

I’m not too familiar with the archival use of timestamps for an entry in a database. Is the primary key a composite key of the “id” and “timestamp” fields, or is that done differently?

This interests me as I work on a local government website (library, not too big of a deal, but librarians are really in to historical references, so…) and would like to know where to look for more on this subject.

By: Joseph Scott's Blog

Joseph Scott's Blog — Sat, 05 Aug 2006 15:54:45 +0000

PostgreSQL at OSCONN 2006

SitePoint mentions a couple of PostgreSQL sessions at OSCON 2006. Good to see more PostgreSQL stuff showing up there.
Theo Schlossnagle presented a case study on moving lots of data from Oracle to PostgreSQL. They needed more Oracle servers to accomm…