[OSM-talk] Nominatim - Does it use the is_in tag?

Hakuch hakuch at posteo.de
Tue Aug 16 19:50:05 UTC 2016


Hey Sarah, do you have documentations that explain how nominatim
processes the queries? That could be an answer to questions like that one

On 16.08.2016 21:27, Sarah Hoffmann wrote:
> On Tue, Aug 16, 2016 at 02:29:26PM +0100, Dave F wrote:
>> Hi
>>
>> I've heard a claim from a user who still wants to use the is_in:*
>> tag as well as boundary tags that Nominatim uses is_in as preference
>> because "geospacial mathematics is resource intensive".
>>
>> Is this true?
> 
> Not at all. Nominatim happily processes boundaries and always prefers
> it over any other hierarchy information.
> 
> It is true that it still understands is_in:* tags but prbably not in
> the way you would think. First of all, they are completely ignored
> on anything at building level (e.g address points and POIs). For
> everything else Nominatim always uses a geospatial match when
> computing the address. is_in:* is just good to help make a decision
> when there are two equally well suited candidates, generally when, say,
> a road is right between two city place nodes. As soon as there are
> boundaries, multiple candidates don't happen anymore, so that is_in:*
> is ignored for all practical purposes.
> 
>> I thought geospacial calculations were fairly light on processing power.
>>
>> I also thought is_in:* was to be discouraged. Being hard coding, if
>> a boundary was to change all affected entities would become
>> inaccurate.
> 
> Yes, if possible always draw boundaries. They are more precise and easier
> to maintain. is_in is unnecessary.
> 
> Kind regards
> 
> Sarah
> 
> _______________________________________________
> talk mailing list
> talk at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0x3CBE432B.asc
Type: application/pgp-keys
Size: 3096 bytes
Desc: not available
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20160816/432668b6/attachment.key>


More information about the talk mailing list