GeoNames Forum

Hi marc,

I would be nice with an incremental synchronization file in addition to the daily syncronization files. Would you be able create e.g. incremental file from e.g. jan 1 2009, which will always hold the latest updates, and let this file be there for the whole year?

The reason why I am asking is that we typically don't update everyday. So if we want updates then we have to download the entire database again. It would save you some bandwith costs.

Let me know what you think.

If someone wants to update to the newest data once per year only then it is better to use the full dump instead of the incremental dump. Some occasional updates are done directly on the database and don't make it into the modification files.

Marc

Hi Marc,

Thanks for replying. I did not mean update once a year, but less regularly than daily. I would be great with a rollup file for each of the months.
E.g. one for january, february, march - always 3 months back in time...

I am using sql-server and it is a pain in the neck, to import the whole package, apply the updates and delete the unwanted.

Would it be easy to just run something like the following pseudo-code;

select * from geonames where lastupdate = current month - 1

on the 1. of each month and save this in a txt file a?

Just a thought - hope you will consider it because it would really save some time

Actually, I think the feature could be done totally outside of geonames.
It would be pretty easy (but kinda boring..) to take 2 dumps, import them to different db/tables, and then create a diff between the two.

and the diff could be used to update production servers once every X days/weeks/months.

Hi samokk / marc,

This is actually what we do now - and each time this will take around a 3 hours before completion.
- Download is around 5-10 min.
- Import will take around 30 minutes using BULK INSERT
- update/diff script completes in 2.5 hours.

Hope you will consider this seriously so that we could get more sparetime and less wasted time

You could look at the modification date field and filter the records that have been updated since your last synchronization. This would be considerably faster but not as thorough as reloading the entire database from scratch

Best

Marc