[osmosis-dev] Improving pgsnapshot

[Brett Henderson]

Hi Paul,

On 9 July 2013 17:34, Paul Norman <penorman at mac.com> wrote:

> I've been working on expanding cgimap to use pgsnapshot. In the process
> I've
> come across some changes to the schema I'd like to propose. The changes
> basically consist of removing the giant (237GB) way_nodes table.
>
> Would this be best to do within the existing pgsnapshot tasks and by
> incrementing the schema version? Should it be in new tasks based on those?
>
> Advice is welcome.
>

Good question.  I can't see any reason why it couldn't be done by
incrementing the schema version.  The main downside will be that consumers
of the schema will have to upgrade their apps.  That's not a bad thing if
there's an advantage to doing so.

For now I'd suggest doing it as an upgrade.  If it becomes a major issue
for people we can possibly create a snapshot of the existing tasks under a
different name that will let them continue to be used (but not actively
maintained) in much the same way the old pgsimple tasks continue to exist.
But I'd rather not do that without a good reason.


> Schema details:
>
> I want to add a nodes.ways bigint[] column to the nodes table and use it
> for
> node->ways lookups, then use the existing ways.nodes bigint[] column for
> way->nodes
>

What is the motivation for this?  I'm guessing that performance is likely
to increase because we get to make better use of geo-spatial indexes ...

It seems like a good idea.  I haven't had a look at the existing queries to
determine the impact of making this change.

Brett
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/osmosis-dev/attachments/20130720/fc99fa53/attachment.html>