[Talk-ca] Imported frustrations

[Adam Dunn]

I think this comes down to a combination of things:
1. User education. People need to be told to double-check what they
are importing, and to also slow down on their imports. Using
bulk_upload.py might be okay in the middle of Nunavut where *nothing*
exists, but it's a poor choice for any place that has even a couple
ways. People also need to be more educated in general editing
techniques. It's true that many people just map for the renderer and
don't consider routing and other functions. When I was doing GeoBase
NRN imports for Vancouver, I would select a small section of roads to
import (maybe 10 blocks by 10 blocks), then I would spend an hour or
two cleaning up messes that people had made in Potlatch before I could
actually do any importing. Bad users abound, it's just that bad users
importing are easier to spot because they cover a larger area.

2. Tools. I think OSM could really do with a "diff"-style program. An
ability to view differences between two layers within JOSM would be a
great thing. This kind of tool would make imports far easier to
perform. See http://en.wikipedia.org/wiki/Diff if you are unfamiliar
with diff. So far, the best thing I have seen is OpenJump Roadmatcher
Conflation, which was used quite heavily in the GeoBase process.
Unfortunately, OpenJump Roadmatcher only works on ways that are not
closed (not polygons). Importers should also be forced to run things
through JOSM's Validator before they submit.

Would it be beneficial to bring back OpenJump? I could take another
look at the scripts and maybe come up with some new process for Canvec
importing.

Adam

On Thu, Oct 6, 2011 at 3:00 PM, James Ewen <ve6srv at gmail.com> wrote:
> On Thu, Oct 6, 2011 at 12:45 PM, Harald Kliems
> <harald.kliems at mail.mcgill.ca> wrote:
>
>> Finding and fixing these errors has been a huge timesuck and not much fun.
>
> Wow, using the tool you just showed me made the job of finding and
> correcting the errors in my local community a very quick and easy
> task... trying to find duplicate ways is pretty tough just going by
> what you can see. Finding unconnected ways is a little easier as they
> show up visibly.
>
> I just cleaned up about 20 square miles of area in about 10 minutes.
>
> There's an even bigger issue where the canvec imports stop well shy of
> the existing roads, or issues like this:
>
> http://tools.geofabrik.de/osmi/debug.html?view=routing_non_eu&lon=-112.55995&lat=53.71582&zoom=13&opacity=0.98
>
> I was thinking that it might be an idea where instead of dropping the
> close match CanVec Imports, that if they were imported, but marked
> such that they would be easily found and deleted if not desired, or
> the existing road could be deleted, and the CanVec import modified to
> make it the new way.
>
> Such as in the case above where the CanVec way was left out of the
> import in deference to the existing road, it would be very easy to
> just modify the CanVec way to make it part of the database, and delete
> the less accurate hand drawn way. It's more of a pain to have to go
> and find the CanVec ways again, and import them so you can delete the
> hand drawn way.
>
> If there were a way to have the CanVec ways that were determined to be
> duplicates shown on a map (kind of like showing roads under
> construction), one could easily compare the two ways, and with a
> simple edit and delete combination, make the CanVec way the one to
> keep, or a simple delete to remove the CanVec way.
>
> Speaking of CanVec imports... anyone going to tackle importing roads
> into Saskatchewan some day? It gets pretty bleak on that side of the
> border in a hurry!
>
> http://www.openstreetmap.org/?lat=52.091&lon=-109.473&zoom=9&layers=O
>
> --
> James
> VE6SRV
>
> _______________________________________________
> Talk-ca mailing list
> Talk-ca at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-ca
>