[OSM-dev] question about node identifiers

[John Smith]

--- On Sun, 23/8/09, Frank O'Dwyer <frank-osm at wordonthestreethq.com> wrote:

> But would still have to pair them up myself, right?  I
> would parse out a set of deleted nodes and a set of added
> nodes and would need to figure out that (delete x) and (add
> y) referrred to the same logical POI - e.g. because they had
> the same / similar name and properties and near each other?
> Or are they already paired up in the change set?

It's unlikely someone would delete and add a POI in the same spot, they are more likely to be modified.

> Also when tracking changesets, is the maximum period daily
> or is it possible to get them on a weekly / monthly cycle?
> Or to put it another way, I'm currently experimenting with a
> snapshot of planet.osm I grabbed a few days ago. How often
> do I need to grab changesets to stay current? Say if I want
> to get a single changeset to cover a period of a month, is
> that possible, or do I need to be mirroring changesets every
> day?

I tend to use the minutely ones to keep my DB up to date as much as possible, but I think daily is the longest snap shot available, you could just grab the 7 daily change files if you only wanted to update once a week.

> I also only want to get POI nodes and not ways/relations
> etc, so I don't really need the whole planet.osm, just a
> subset - is there a recommended way to ask for only those to
> save bandwidth and load on the server? It would also save me
> time on parsing as currently parsing out the amenity nodes
> takes ages.

The vast majority of data in changesets is nodes, however most of these wouldn't be amenity nodes, on a half decent machine it doesn't take much to process a daily file through an xml parser to pull out just the amenity changes. Also node changes could turn an amenity into something else if it was tagged wrong so you would need to keep tabs on this and delete them.

> Last thing I am wondering is if there is a description
> somewhere of the mandatory / optional tags per different
> type of amenity? I assume this changes over time and so if
> I'm parsing these I need to handle cases where a particular
> expected tag is not there.

There is a list on the Map Features wiki page:

http://wiki.openstreetmap.org/wiki/Map_Features

However people are free to choose their own tags as well.