Cluster feature: API into the Parser / Parser as an independent module - Mailing list pgsql-cluster-hackers

From Greg Smith
Subject Cluster feature: API into the Parser / Parser as an independent module
Date
Msg-id 4B764504.7010709@2ndquadrant.com
Whole thread Raw
List pgsql-cluster-hackers
In the first week working on this, we got the top two features at
http://wiki.postgresql.org/wiki/ClusterFeatures described well enough
that I think anyone familiar with this area at all can understand the
basics of what's involved in them.  I'd like to keep that going by
asking who would like describe the next feature on the list, "API into
the Parser / Parser as an independent module".

My usual questions are now on the Wiki for future reference.  I don't
expect specific answers to each of those, if you all just keep throwing
details around I'll continue my work of sorting through those and
summarizing at the end.  That's been a very educational exercise for me
so far, and I look forward to learning something new on this one too.

It sounds like this Parser API feature aims to make life easier for any
statement-based replication projects such as Slony.  If they could
retrieve the statement being replicated directly from the parser, rather
than having to scrape everything out of a trigger, they could do any
tweaking needed in order to replicate that statement to other nodes more
easily than how things are done now.

I could use some examples of what the typical types of constructs that
need to be replaced that way are, as well as some idea what "just first
step, before going to the catalog" is supposed to mean.  I'm also
guessing this is a necessary requirement for wanting to replicate DDL
statements or similar statements that you can't "instrument" with
triggers, right?  No idea what work has already been done in this area
or what the main difficulties are either.

--
Greg Smith    2ndQuadrant   Baltimore, MD
PostgreSQL Training, Services and Support
greg@2ndQuadrant.com  www.2ndQuadrant.com


pgsql-cluster-hackers by date:

Previous
From: Greg Smith
Date:
Subject: Re: Global Deadlock Information
Next
From: Markus Wanner
Date:
Subject: Re: Global Deadlock Information