[Imports] Reducing primary key damage (Branch from OpenMetaMap discussion)

Bryce Nesbitt bryce2 at obviously.com
Sat Aug 27 22:07:13 UTC 2011

On 08/25/2011 07:33 AM, Serge Wroclawski wrote:
> I know Ian (who did the GNIS imports) gets ruffled feathers when I
> mention to him that sometimes if there are duplicate GNIS features in
> an area, I've just deleted one that I thought was "less correct".
> I could have (probably should have) taken the gnis id tag and merged
> it with the existing feature, but the point is in our dataset, one
> can't assume that links between the imported object and the current
> object exist. You can't assume it through object IDs and you can't
> really assume it through tags. There will be edge cases in both sides
> which keep it from always being true.
Each piece of data is a hint for future conflation process.  So I 
encourage key preservation, especially for gnis.  Those are officially 
recognized names, and the data does change slowly over time.

Here's an example where to imports both brought in "castro park".  I 
deleted one of them, but copied over the gnis key:


So hopefully a future gnis import would see that "Castro Park" need not 
be duplicated again, even if the park gets renamed by the City.

More information about the Imports mailing list