[OSM-dev] Suggestion about how to setup a local osm development

[Tom Hughes]

In message <841994.59520.qm at web35113.mail.mud.yahoo.com>
        Chien Nguyen Khac <threesearch at yahoo.com> wrote:

> Tom wrote:
>
> > Ah, so you&'re talking about using the data rather than working on
> > the site itself. In which case most of your steps are irrelevant as
> > there is no need to load all the data into a clone of the live
> > database just to work with it.
> >
> > In fact it probably generally makes more sense to parse the planet
> > dump yourself and load it into some sort of database that is designed
> > for whatever task it is you want to do with it.
>
> My intention to load the whole data to local database is to develop
> a routing application that query osm database as a whole.

Well a routing application is a prime example of something that almost
certainly want to run against a simplified database and not the main
database as that is far richer than is needed for routing.

> > You know there is already an live CD image for OSM that you can
> > boot in VMWare I take it? You can find it at:
> >
> > http://artem.dev.openstreetmap.org/files/osm-linux.iso
>
> Thanks for you infomation, I will try the live CD. However live CD
> is read only, I can not update the osm database in that live CD. So
> I still prefer having dev server in VMWare image that give me total
> control of the server.

I'm sure we look forward to your development of a VMWare image.

> > I think the real question here is, are you talking about developing
> > OSM itself, or just developing applications that happen to use our
> > data...
>
> I am talking about both: developing OSM itself and using OSM
>
> For developing OSM itself:

> There is the routing library for postgis database some where on the
> internet, so I want to do the same for MySQL database.
>
> This is a my long term plan and may need your help because it
> requires a lot of work and time.
>
> If I can setup a local osm dev server, I will do the following
> things with osm database:
>
> 1. write SQL statements to drop segments table and convert segments
>    data into nodes and ways tables.

I think it's a bit more complication than a few SQL statements...

> I hope osm foundation will consider implementation of dropping
> segments tables soon.

I'm not sure it's a matter for the foundation as things stand.

> If we delay it and some US developers start to import a huge Tiger
> data into osm database then we will have more trouble with
> nodes/segments/ways later on.

I don't see why - most of the problems will be with the older data
which has tags on segments and suchlike.

> 2. design table structure for routing tables (graph,
>    nodes...). These routing tables will provide a graph network for
>    routing application

Is there any consensus that we want to have something like that in
the main database?

This is exactly what I mean about probably wanting to extract the
data into a separate, differently structured, database for doing
something like routing.

I'm not saying you shouldn't load up a copy of the data, and if you
want to then I'd suggest looking at osmosis - there are a couple of
script in subversion to do it, but I don't think either of them is
really up to date with the current database schema and osmosis is
probably faster anyway.

Tom

-- 
Tom Hughes (tom at compton.nu)
http://www.compton.nu/