detect nodes with missing highway=crossing

[mkoniecz]

definitions

A - node that may be left by pedestrians in at least two directions. Naive check for eligible ways: highway=footway or highway=path or (highway=cycleway and (foot=yes or foot=designed))
B - node that may be left by cyclists in at least two directions. Naive check for eligible ways: highway=cycleway or ((highway=path or highway=footway) and (bicycle=yes or bicycle=designed))
C - node that may be left by cars in two directions. Naive check: (highway=residential or highway=unclassified or highway=primary or highway=secondary or highway=tertiary or highway=trunk (...))

Any node with ((A or B) AND C AND without highway=crossing AND without crossing=no) should be reported as "highway=crossing or crossing=no is probably missing".

[mkoniecz]

example of node that should has highway=crossing tag. JOSM validator is failing to detect anything suspicious.

[Don-vip]

In 6303/josm:

fix #9141 - new validator test: detect nodes with missing highway=crossing

[Don-vip]

Thanks for the pseudo-code !

[skyper]

• The code does not work with highway=traffic_signal and crossing=*
• what about railways ?

[Don-vip]

I'm leaving today for 3 weeks in USA, I'll look into it after, if nobody fixes it before :)

[aceman]

Interesting, I like that this is taken care of in some way. But isn't a node connecting a way for pedestrians and a way for cars (as defined in the bug description) enough of a clue that it is a crossing? What if it is not marked on the road in reality? What else can it be if it is not a crossing? If there is no crossing, the ways would not meet in a shared node and probably be in different layers. Are there any guidelines on this?

[skyper]

This test is too strict. Do we really need crossing=unmarked on every service way without any sidewalk nor kerb ?

Would prefer to downgrade it to an info in cases where only three ways connect and when paths only connect from one side like in curves.

[anonymous]

A driveway crossing over a sidewalk does not make the sidewalk point a highway=crossing, don't you think? Drivers turning onto the driveway are crossing a sidewalk, not a pedestrian crossing. Like this node ​http://www.openstreetmap.org/?node=714087855 as seen here ​http://goo.gl/maps/lO8iv

[mkoniecz]

Detecting driveways is simple, just look for service=driveway.

IMHO it would count as highway=crossing. "Crossing infrastructure for the convenience of pedestrians, cyclists etc. should first be tagged with highway=crossing" is a bit tricky as this type of construction is for convenience of both pedestrians (etc) and drivers.

For related discussion about driveways and highway=crossing look at ​http://wiki.openstreetmap.org/wiki/Talk:Key:crossing#Proposal:_crossing_ref.3Ddriveway

Also, in ​http://josm.openstreetmap.de/ticket/9245 I suggested detecting reverse situation - unneeded crossings.

[Don-vip]

In 6368/josm:

see #9141 - validator: Remove service and road support for missing crossing on highway for next release at least

[mdk]

Replying to skyper:

This test is too strict. Do we really need crossing=unmarked on every service way without any sidewalk nor kerb ?

In 6378: The problem is, that the validator still reports "Missing crossing information" if you use crossing=unmarked (At least when highway=unclassified is crossing a highway=cycleway). I think all crossing=* (except of crossing=unknown) should always satisfy this test.

[mkoniecz]

Replying to mdk:

Replying to skyper:

This test is too strict. Do we really need crossing=unmarked on every service way without any sidewalk nor kerb ?

In 6378: The problem is, that the validator still reports "Missing crossing information" if you use crossing=unmarked (At least when highway=unclassified is crossing a highway=cycleway). I think all crossing=* (except of crossing=unknown) should always satisfy this test.

Why highway=crossing tag should be omitted in this case?

[mdk]

Replying to Bulwersator:

Replying to mdk:

Replying to skyper:

This test is too strict. Do we really need crossing=unmarked on every service way without any sidewalk nor kerb ?

In 6378: The problem is, that the validator still reports "Missing crossing information" if you use crossing=unmarked (At least when highway=unclassified is crossing a highway=cycleway). I think all crossing=* (except of crossing=unknown) should always satisfy this test.

Why highway=crossing tag should be omitted in this case?

Sorry, I was not precise enough: It is a combined cycleway and footway (highway=cycleway & foot=yes) not segregated.

btw: I'm a little bit confused about how to tag such a way. I found so many opinions ...

I tried a little bit: Validator is reporting a crossing between highway=unclassified and highway=cycleway always as "Missing crossing information" (with or without foot=yes), so this would be an error

[mkoniecz]

Replying to mdk:

Replying to Bulwersator:

Replying to mdk:

Replying to skyper:

This test is too strict. Do we really need crossing=unmarked on every service way without any sidewalk nor kerb ?

In 6378: The problem is, that the validator still reports "Missing crossing information" if you use crossing=unmarked (At least when highway=unclassified is crossing a highway=cycleway). I think all crossing=* (except of crossing=unknown) should always satisfy this test.

Why highway=crossing tag should be omitted in this case?

Sorry, I was not precise enough: It is a combined cycleway and footway (highway=cycleway & foot=yes) not segregated.

I tried a little bit: Validator is reporting a crossing between highway=unclassified and highway=cycleway always as "Missing crossing information" (with or without foot=yes), so this would be an error

I tag such locations as highway=crossing (including cyclist only crossings that I tag as highway=crossing, bicycle=yes, foot=no). I added suggestion on wiki to document this usage ( see ​http://wiki.openstreetmap.org/wiki/Talk:Key:crossing#Add_an_example_with_cyclist_only_crossing ).

btw: I'm a little bit confused about how to tag such a way. I found so many opinions ...

I ended with highway=path, bicycle=designated/yes, foot=designated/yes as at least highway=path, foot=designated, bicycle=designated is rendered by JOSM. And it was declared as a proper scheme in my country (Poland).

[skyper]

Replying to Bulwersator:

Replying to mdk:

Replying to Bulwersator:

Replying to mdk:

Replying to skyper:

This test is too strict. Do we really need crossing=unmarked on every service way without any sidewalk nor kerb ?

In 6378: The problem is, that the validator still reports "Missing crossing information" if you use crossing=unmarked (At least when highway=unclassified is crossing a highway=cycleway). I think all crossing=* (except of crossing=unknown) should always satisfy this test.

Why highway=crossing tag should be omitted in this case?

Sorry, I was not precise enough: It is a combined cycleway and footway (highway=cycleway & foot=yes) not segregated.

I would always tag a combined way with highway=path. In your case probably bicycle and foot designated.

I tried a little bit: Validator is reporting a crossing between highway=unclassified and highway=cycleway always as "Missing crossing information" (with or without foot=yes), so this would be an error

I tag such locations as highway=crossing (including cyclist only crossings that I tag as highway=crossing, bicycle=yes, foot=no). I added suggestion on wiki to document this usage ( see ​osmwiki:Talk:Key:crossing#Add_an_example_with_cyclist_only_crossing ).

Thought we use a subtag of the highway value: path=crossing/sidewalk
For a bicycle only crossing I would use designated and not only "yes". I like crossing_ref=driveway.

btw: I'm a little bit confused about how to tag such a way. I found so many opinions ...

I ended with highway=path, bicycle=designated/yes, foot=designated/yes as at least highway=path, foot=designated, bicycle=designated is rendered by JOSM. And it was declared as a proper scheme in my country (Poland).

But this is all off topic:

1. We need a highway=* with crossing=*
2. So far you get a warning about highway!=crossing and crossing=* which is wrong.

[josm@…]

Replying to anonymous:

A driveway crossing over a sidewalk does not make the sidewalk point a highway=crossing, don't you think? Drivers turning onto the driveway are crossing a sidewalk, not a pedestrian crossing. Like this node ​http://www.openstreetmap.org/?node=714087855 as seen here ​http://goo.gl/maps/lO8iv

I totally agree. We need a way to distinguish between an highway crossing a footway and a footway crossing a highway. Until now I assumed that an untagged crossing of a footway and a highway was in fact a highway crossing a footway since you cannot tag tat otherwise. I find it rather odd to make missing crossings an error when there is no possibility yet to correctly mark a vast part of "crossings".

[simon04]

I'd consider the current warning "Missing crossing information" (this text is sometimes really confusing) as Severity.OTHER and re-word it to "Missing pedestrian crossing information":

src/org/openstreetmap/josm/data/validation/tests/Highways.java

@@ -144 +144 @@
                     handleCarWay(n, w);
                 }
                 if ((leftByPedestrians || leftByCyclists) && leftByCars) {
-                    errors.add(new TestError(this, Severity.WARNING, tr("Missing crossing information"), MISSING_PEDESTRIAN_CROSSING, n));
+                    errors.add(new TestError(this, Severity.OTHER, tr("Missing pedestrian crossing information"), MISSING_PEDESTRIAN_CROSSING, n));^M
                     return;
                 }
             }

[Don-vip]

In 6434/josm:

see #9141 - adapt severity and wording

[mkoniecz]

This patch will produce wrong messages for crossings with cyclists, this code should be something like this:

                     handleCarWay(n, w);
                 }
                 if ((leftByPedestrians && leftByCyclists) && leftByCars) {
                     errors.add(new TestError(this, Severity.OTHER, tr("Missing pedestrian and cyclist crossing information"), MISSING_PEDESTRIAN_CROSSING, n));
                     return;
                 } else if (leftByPedestrians && leftByCars) {
                     errors.add(new TestError(this, Severity.OTHER, tr("Missing pedestrian crossing information"), MISSING_PEDESTRIAN_CROSSING, n));
                     return;
                 } else if (leftByCyclists && leftByCars) {
                     errors.add(new TestError(this, Severity.OTHER, tr("Missing cyclist crossing information"), MISSING_PEDESTRIAN_CROSSING, n));
                     return;
                 }
             }

Sorry for not formatting this as a patch.

[Don-vip]

Ticket #9386 has been marked as a duplicate of this ticket.

[mdk]

Is there any open case left?

I hadn't found any problems since a long time!

[Klumbumbus]

Replying to mdk:

Is there any open case left?

seems not

[stoecker]

Milestone renamed