Port Osmose "Unconnected waterway or wrong way flow" check to JOSM validator

[liotier]

In large complicated hydrographic networks, waterway flow direction consistency is a common issue, as the number of errors reported by Osmose attests: ​http://osmose.openstreetmap.fr/en/errors/?item=1220 - reducing the number of such errors that reach the Openstreetmap database would be valuable.

The Osmose analyzer that reports those errors is published at ​https://github.com/osm-fr/osmose-backend/blob/master/analysers/analyser_osmosis_waterway.py

I believe that this error should be a warning because there are rare but legitimate cases for an hydrographic network not being consistent with a fully hierarchical model taking root into the ocean.

[Klumbumbus]

In 10232/josm:

fix #12646 - warn about "Unconnected waterway or wrong way flow"; ignore hazmat=designated

[liotier]

Wow, thanks - I am currently conflating a humongous heap of waterways whose vector data source did not bother setting a direction... So the directions are all random and I have to check them one by one... So of course I was thinking about this wishlist item - I didn't think it would arrive so soon... This will help immensely - I'm so impatient to test tomorrow's development version !

[Klumbumbus]

If you can't wait, you can download it as jar file from jenkins :)

jenkins/job/JOSM/jdk=JDK8/

(choose josm-custom.jar to download)

[liotier]

Didn't know about the Jenkins version... I'll try that next time. I just got the daily version 10240 but merely creating a new layer gives me errors - not to mention loading the big session I'm working on. I'm still using Java 7 on this Debian host and I'm smelling that this could be the problem... Can the development version still work with Java 7 or do I have to upgrade ? I'll try from an host where I have a Java 8 setup to check if that works... But that'll have to wait until tomorrow.

For what I mean by "humongous" see ​https://www.openstreetmap.org/changeset/39388940 which I'm in the process of uploading. Then see ​http://www.openstreetmap.org/#map=13/14.3811/-12.2633 for the finished product from an earlier similar session... And I have plenty more in the pipeline for the coming months...

Well, anyway - tests tomorrow.

[Don-vip]

Download it from the JDK7 job if you cannot update to java 8 yet: jenkins/job/JOSM/jdk=JDK7

But it's only a matter of days before we abandon Java 7 for good, so yes, you should update. Java 8 is available through Jessie backports, there's no reason not to update.

[liotier]

Aaaaand... It works on the 80k changes session - perfectly ! Maybe ignore waterway=canal, which doesn't have a flow direction nor is part of a typical hydrographic network... But the essential functionality is up to spec - big thanks for this improvement !

[Klumbumbus]

In 10251/josm:

see #12646 - ignore canal in unconnected waterway test

[Klumbumbus]

The test checks only if the last node of a waterway is connected to another way. However it does not check if it is also a waterway. So if the end of a waterway is accidentaly connected to e.g. a landuse way or a street or a riverbank, then no warning is displayed. That is currently not possible with the simple mapcss validator syntax and would require a more complex Java test.

[Klumbumbus]

In 10427/josm:

see #12646, fix #12992 - limit "Unconnected waterway or wrong way flow" warning to waterway=stream|ditch|river|drain

[Klumbumbus]

In 10530/josm:

see #12646 - don't warn about "Unconnected waterway or wrong way flow" if waterway ends with a natural=sinkhole or natural=cave_entrance node

[Klumbumbus]

In 15413/josm:

fix #16334, see #12646 - don't warn about unconnected waterway=ditch