Validator - false positives in "way end node near other way"

[bilbo]

Check for "way end node near other way" is useful, however in some areas it finds many false positives.

Minimal distance to trigger this warning seems to be 10 (by default).

But what unit it is? when the preference is read, the number is divided by some arbitrary number 6378135.0
What is meaning of that number? No documentation in the source

I've collected some examples of false positives found by validator, along with some ideas how to avoid them:

validator-deadend.png - T-junction near end of way

There is branching (T-junction) close to end of way. This shows false alarm, even though the way in warning is connected to the way with the endnode.

validator-platform.png - Parallel railway tracks or platforms

Railway platforms or tracks running in parallel - these can be as close as 5 meters.
Have some smaller distance to trigger the warning, if both ways (the one with end node and the one close to the node) are parallel or very close to parallel

validator-powerline.png - highway end node close to powerline
validator-stream-road.png - highway end node close to stream

Sometimes features are running quite close to each other "by design".
Power lines often are placed with little regard of what is below them and endpoint of highway ending up near powerline is surely not connected to it.
Sometimes stream run right behind the streets (the streets are ended just before the stream) and they are also not connected

Therefore for "incompatible" features (highway vs waterway, powerline vs waterway/railway/highway) that should not be probably connected
(highway connected to the powerline does not make sense), the distance should be lower too.

I think for these cases, some lower distance threshold (one quarter by default, perhaps ....) should be used

[bilbo]

Parallel railway tracks or platforms

[bilbo]

highway end node close to powerline

[bilbo]

highway end node close to stream

[skyper]

I think in all these example there should be a warning as long as the highway does not end with noexit=yes, but it should not be refered to a way with a different layer=* .

[bastiK]

Ticket #4363 has been marked as a duplicate of this ticket.

[skyper]

Validator is in Core now

[skyper]

Seems to be fixed

r3992

[skyper]

Sorry. Validator does not find any "way end node near other way" right now

[skyper]

I have another example with barrier=fence

[skyper]

example file

[skyper]

Ticket #5837 has been marked as a duplicate of this ticket.

[skyper]

Validator also warns about "way end node near other way" with untagged ways which are only part of landuse-relations as "other way" . Think this is a false positiv, too.

r4074

[skyper]

Ticket #4257 has been marked as a duplicate of this ticket.

[skyper]

Replying to skyper:

Validator also warns about "way end node near other way" with untagged ways which are only part of landuse-relations as "other way" . Think this is a false positiv, too.

r4074

Everything except the unntagged way is still an issue. Please, do check the main tag of both ways and filter to only check for ways with the same main tag. Thanks

r4137

[skyper]

Ticket #5435 has been marked as a duplicate of this ticket.

[simon04]

Ticket #6313 has been marked as a duplicate of this ticket.

[mdk]

It looks like this issue is not so clear, but it huts me every day. So I add a more clear example with a highway near a poweline, a waterway and a barrier. I hope this will help to solve the problem.

[mdk]

Ping :-)

(BTW why dose attaching a file did not effect 'last change date' of the issue in trac?)

[skyper]

Found another issue:

a bridge (node) which ends near a landuse=farmland (way)

This should not be a problem (no warning).

EDIT: Already reported in comment:9

[skyper]

Ticket #7742 has been marked as a duplicate of this ticket.

[skyper]

From #7742 AM909:

In r5181:

The validator reports "way node end near other way" when it shouldn't - specifically for objects of different types. Examples:

• A road (highway=*) that ends near an admin boundary (boundary=administrative + place=city + ...)

• A road (highway=*) that ends near a river (waterway=river).

• A road (highway=*) that ends near a power line (power=line). This probably doesn't happen if the power line has layer > that of the road, but few people seem to do this, and the validator doesn't report power lines crossing roads of the same layer, so it is apparently assumed that power lines are layer=6.

• A fence (barrier=fence) that ends near a road (highway=*).

Basically, it seems this check is designed to find ways that should intersect, but don't. It should, therefore, limit itself to types of ways that should actually intersect (like those with the same main key, though there may be some others).

[anonymous]

Ticket #7893 has been marked as a duplicate of this ticket.

anonymous:

I often see how the validator reports e.g. a footpath joining a street near a tram line, which leads to said warning. The warning is not correct as connecting a footpath to a railway is wrong. It should not report the warning for such heavily dissimilar ways.

[anonymous]

Another edge case that should be avoided...

I mapped the entrance to an underground parking garage from imagery as a tunnel just going off the end of the road that doesn't connect to anything. The tunnel segment ends near another road on the surface and throws this validator error. It should ignore all ways that are on a different 'layer' as indicated by the layer tag. A similar thing should probably also be done for the 'level' tag which is used for indicating which "floor" something is on for indoor mapping.

[skyper]

Replying to skyper:

From #7742 AM909:

In r5181:

The validator reports "way node end near other way" when it shouldn't - specifically for objects of different types. Examples:

• A road (highway=*) that ends near an admin boundary (boundary=administrative + place=city + ...)
• A road (highway=*) that ends near a river (waterway=river).

• A way (highway=*,waterway=* or railway=*) that ends near landuse=* or natural=*.

• A road (highway=*) that ends near a power line (power=line). This probably doesn't happen if the power line has layer > that of the road, but few people seem to do this, and the validator doesn't report power lines crossing roads of the same layer, so it is apparently assumed that power lines are layer=6.
• A fence (barrier=fence) that ends near a road (highway=*).

Basically, it seems this check is designed to find ways that should intersect, but don't. It should, therefore, limit itself to types of ways that should actually intersect (like those with the same main key, though there may be some others).

+1

All cases which lead to incorrect mapping and to misses of the valid cases as you get too many false warnings.

[Don-vip]

Ticket #9395 has been marked as a duplicate of this ticket.

[simon04]

In 6515/josm:

fix #3531 - validator: unconnected ways: handle highway/railway/waterway/natural+landuse/power separately

[simon04]

In 6518/josm:

see #3531 - validator: unconnected ways: ignore highway/railway=platform

[skyper]

another example file

[skyper]

Thanks for the improvements but I am sorry, I did find two more problem concerning waterways and water:

1. If a waterway runs into a lake and starts on the other side again I get a warning, if the lake is small and/or both ways are close.
2. If you have a short part as tunnel in between (bridges are quite rare) we still have the same problem as we had with highways.

Please use the attached example

[simon04]

In 6667/josm:

fix #3531 - validator: unconnected ways - skip "Way end node near other way" if connected

[Don-vip]

Ticket #5277 has been marked as a duplicate of this ticket.