Here are the remainder of the addresses in Epping for review:

https://gitlab.com/dionmoult/osm-nsw-address-import/blob/master/review/EPPING-3.osm

I am still unsure about how to accurately map multiple addresses that end up in 
one spot. If there are just two (e.g. 18 Third Avenue, and 2/18 Third Avenue) 
over a house, I can just displace them and let it be. However, I am now 
encountering 49 nodes that are all in one spot. Here's a sample of the 
addresses they have which I have sorted ascending:

... snip ...
206/38-44 Pembroke Street
207/38-44 Pembroke Street
208/38-44 Pembroke Street
210/38-44 Pembroke Street
212/38-44 Pembroke Street
301/38-44 Pembroke Street
302A/38-44 Pembroke Street
304/38-44 Pembroke Street
305/38-44 Pembroke Street
306/38-44 Pembroke Street
306B/38-44 Pembroke Street
307/38-44 Pembroke Street
310/38-44 Pembroke Street
401A/38-44 Pembroke Street
402A/38-44 Pembroke Street
... snip ...

As you can see:

 1. It is ambiguous as to what it is, is it a flat, a tenant, a "granny flat" 
as the locals call it, another apartment block, etc?
 2. It is not obvious what the interpolation pattern is. Odd? Even? Floors? 
Alphabetical suffix?
 3. I have even encountered up to over 200 nodes in one spot (say for a large 
apartment tower, although admittedly these towers have a bit more logic to the 
numbering than the example given above)

At the moment, each one of these addresses are separate and distinct address 
nodes. I indeed can continue my conservative approach of not assuming 
interpolation, not merging, and not doing anything fancy, and instead only 
displacing them, however I'm not sure how mappers would react to 200 address 
nodes arranged in a 10m radius circle.

What should I do? Continue the conservative displacement strategy, or?

Dion Moult

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
On July 29, 2018 9:41 PM, Dion Moult <[email protected]> wrote:

> Here are another ~2000 addresses in Epping for review. Again, if there are no 
> comments in 2 weeks I will upload.
>
> https://gitlab.com/dionmoult/osm-nsw-address-import/blob/master/review/EPPING-2.osm
>
> Dion Moult
>
> ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
> On July 29, 2018 6:34 PM, Dion Moult [email protected] wrote:
>
> > Apologies, I had a few weeks going overseas for holidays (and did a lot of 
> > GPS tracking on the way!)
> > I like your suggestion of moving the nodes slightly! I have written a quick 
> > script which will take overlapping nodes and shift them all radially around 
> > their current location by less than 1 meter. That way they are close enough 
> > for navigation but far apart for humans to understand, and no data is lost. 
> > In the spirit of this import, it only operates on selected nodes, which 
> > forces the mapper to run the validate and screen the overlapping nodes 
> > before actually shifting them using the script. Script is below.
> > https://gitlab.com/dionmoult/osm-nsw-address-import/blob/master/src/displace_overlapping_nodes.js
> > Updated OSM file with the shifted nodes are here:
> > https://gitlab.com/dionmoult/osm-nsw-address-import/blob/master/review/EPPING-1.osm
> > If there are no more issues I will upload in 2 weeks.
> > Dion Moult
> > ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
> > On July 8, 2018 5:50 PM, Martin Koppenhoefer [email protected] wrote:
> >
> > > sent from a phone
> > >
> > > > On 8. Jul 2018, at 08:29, Dion Moult [email protected] wrote:
> > > >
> > > > 2.  don't leave the nodes in the same position. This action has the 
> > > > effect of intentionally creating messy data for humans. (it would be 
> > > > good for navigator software, though! But OSM is humans first!)
> > >
> > > if you move these nodes slightly (e.g. < 1m) it won’t make a difference 
> > > for navigation software, but will make the situation easier to understand 
> > > for mappers
> > > Cheers,
> > > Martin
> >
> > Imports mailing list
> > [email protected]
> > https://lists.openstreetmap.org/listinfo/imports



_______________________________________________
Imports mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/imports

Reply via email to