I am surveying (on foot, using "analog" notes - pen and paper that is) the area near my workplace for house numbers. Since I do not have time and resources to tag every house/building that exists, I am using the information very conveniently made available on small road signs placed on almost all street corners. Example: The numbers show the first and last existing address number in each direction of each side (there are usually four of these signs at each street intersection) of the street, and sometimes, also information about the ZIP code of the area. Because of this, I am using the Karlsruhe Schema to add this information to the map. Now, my question is: "Should the ways that generate the interpolated address numbers from the known points cross other streets or not?" Example of interpolation ways crossing street (not yet uploaded): Example of interpolation ways not crossing street (not yet uploaded): If the interpolation way cross the streets between known points, it would be possible to generate "approximate locations" for address numbers that do not exist in reality (and never will - they would be ON the street that was crossed!), overcoming the limitation of Nominatim for address numbers outside of existing ranges defined by the interpolation ways. However, it would be a completely artificial construct. For example, search (in Brazil) for: - Rua João Mafra, 25 (return is correct) - Rua João Mafra, 201 (return is "not correct" - should be between 201 and 239 points) asked 19 Sep '13, 21:36 MCPicoli |
After all this time, I came with a few simple rules for deciding when to make the address interpolation way cross or not cross other highways. These rules are:
That would mean that the address interpolation way will cross other ways if one or both address numbers of adjacent blocks corners are not known (because in this case, there is the possibility of street numbers between the known corner and the non-corner street number) and will not cross other ways only if both of them are known (because, in this case, the street numbers between them are _known to be non-existent). One image says a lot more than one thousand words, so:
(Sorry for the lame MSPaint screen capture...) answered 18 Nov '13, 19:32 MCPicoli |
I would confine interpolation to one block, that is not crossing streets. answered 20 Sep '13, 09:09 gormo 1
Gormo, could you please elaborate some more on why not crossing streets? Already 20 days since I asked the question, and I have not fully decided on how to do it...
(09 Oct '13, 19:43)
MCPicoli
|
Using housenumber interpolation is just a convenience for the mapper in a hurry. It's faster to map but less precise. In effect, what you're asking is "how much less precise can we tolerate to be ?". The answer to that question is personal, and could even differ from one editing session to the next. Personally I'd make multiple ways, as precise as possible without being too tedious. Then again, the streets I map are usually short enough and often have housenumbers assigned a bit hapazardly, so I tend to not use interpolation at all. answered 19 Sep '13, 22:55 Vincent de P... ♦ Making multiple ways should not be very tedious. However, the street signs are positioned exactly on the block corners, but the address number may or may not be very near this same corner. The sign shows the first valid address number at the time of the sign's creation, probably from post office data. In Brazil, except by the means of surveying the area, using the road signs or getting all of the actual house numbers, the information about house numbering is not public. Therefore, I believe it is much more important to have some information than no information at all!
(20 Sep '13, 13:00)
MCPicoli
Not forgetting to mention that OSM is iterative, so the information will be refined/updated by other mappers in the future.
(20 Sep '13, 13:03)
MCPicoli
Nobody is suggesting to put no information at all. If you have the housenumbers at each crossroad but surveying individual houses is too tedious, then use the crosroad signs and multiple ways. And yes, it can always be refined later.
(20 Sep '13, 15:23)
Vincent de P... ♦
|