Non-existent town on map

[krankyd]

Most likely this is due to the fact you're using the map editors scheme (which is one that was created by users, and might not behave exactly like the official ones).
Regardless, this is a good chance for us to review the behavior of 'ghost' streets and cities (ones that no longer have any segments attached to them). We're considering how to handle these cases, where a user deletes or moves the last segment of a street, leaving a street / city empty. Most likely we can delete the city and street, but not allow to 'undo' the deletion (as we allow when deleting a regular segment). Once deleted, and assuming the city *should* exist, an editor will have to enter it again.

[krankyd]

hi,
I saw the correspondance also via email, and in no way was I trying to suggest that you are wrong or not trying to help - sorry if it came out that way. Without your feedback there would be many bugs did we would have never discovered - such as this one.
Ghost towns were supposed to be removed from the maps, but apparently, some of them are still there. We will work on a fix to remove them from the DB, as well as from the maps. Will keep you updated. Thanks.

[phrodod]

It seems to me that there should be a different way of specifying cities than of specifying roads. Right now, if I go to make a map edit, it automatically assigns a city for me. In areas near city boundaries, that's a pain, but is manageable IF you know the boundary. I'd venture that most people don't know the actual boundaries, so if you allow those who care to set the boundaries, then you just need to find all roads within or near those boundaries to satisfy searching.

The current method violates good database principles by repeating the same information lots of times, causing endless problems.

As long as the cities issue is coming up, let me add that the ability to include community names as well as city names would be REALLY useful. Woodland Hills and Northridge are both in Los Angeles, but are different communities. But everything I do to try to break cities into communities fails spectacularly.

Thanks for reading this rant.

[phrodod]

True that municipal boundaries are currently not available to be set by any end-users, AM or not.

It seems like they can be set by labeling roads, which is where the ghost cities come from. My point is that this is the wrong way to create cities, either official or not.

[RallyChris]

If you turn on the map editor color scheme, does it show a square area?

If so, it's a known issue. Haven't gotten clear instructions on how Waze HQ wants to handle it.

[rlpl]

There are two different problems here.

Problem 1. Waze does not handle correctly two different cities (villages, settlements, etc.) having the same name. If two such cities exist, waze thinks that they are one big city and connects them. Now imagine three cities: a big triangle appears and may cover even whole map of your country, the ghost city name appears everywhere. The best way to see the area of that ghost city is to turn off the Aerial layer in Cartouche.

The solution advised by waze HQ now is to include the name of the state/province/county/whatever into the city name so the names become different and waze software can distinguish and disconnect them. The problem disappears after the next map rebuild cycle.

My recent conversations with support seemed to think it was a problem with Cartouche guessing the wrong town name when a new segment was recorded. They were talking about turning off the auto-population of the town for newly recorded segments to deal with the problem for now.

This is the side effect of this problem: the false city name appears as the hint to all newly recorded segments. If you edit the segment (even its geometry!) and don't correct the city name it becomes permanent and remains in that place until you come back to the same place and edit it manually. This is how the false city may grow if you are not careful when editing. Fortunately, if you split the cities then the false city name is removed from the segments newly recorded before the split and still not edited. That's what makes me think that the city name of the newly recorded segment is a hint and appears as you start editing rather than a permanent property.

Problem 2. Even if you solve the problem 1. correctly and successfully, the ghost city name still appears at random places in the mobile client. This is really strange because in Cartouche the city areas are corrected. Even more strange is that the ghost areas are different at different zoom levels, some of them being correct. This makes me think that the tiles for Cartouche and for the mobile client use different databases or different rendering algorithms; Cartouche is correct and the mobile client (software or database) is buggy. Cartouche is a proof that the waze HQ has got a correct algorithm and a correct database. Why don't you use them for the client tiles?

Most likely this is due to the fact you're using the map editors scheme (which is one that was created by users, and might not behave exactly like the official ones).

Map editors color scheme is fine and we don't complain about it. My opinion is quite opposite: this color scheme is fine because it exposes the problem. It is not that other schemes are more correct. Ghost city name appear in all color schemes in exactly the same way. The only difference in map editors' behavior is that it shows the area of the ghost city so you know not only its name but also where it is (incorrectly) placed. It is good for a reference and diagnostics. It helps to catch the ghost, other schemes just try to pretend that the ghost does not exist. But they don't hide the ghost successfully and it is still visible.

Most likely we can delete the city and street, but not allow to 'undo' the deletion (as we allow when deleting a regular segment). Once deleted, and assuming the city *should* exist, an editor will have to enter it again.

I have not noticed any way to undelete the city or street name. Is there really such feature in waze? Do you mean that the city or street name appears as the hint when you start typing? The hints are fine but I don't think that suggesting the name of a street or a city which has been deleted is necessary. Please note that the city/street name could have been deleted or edited because the old one has been mistyped. For me suggesting the mistyped name is even bad. If you don't like removing the name of the street or a city immediately and permanently then maybe these names should be deleted during the next map rebuild cycle or after a month of being unused.

But I don't think this is only a problem of the deleted cities because sometimes the cities which appear in the client really exist but in different place. For me it looks like some tiles have not been refreshed and still expose the city which used to cover this place. Sometimes I see the ghost squares in the places where no road segment exists. Sometimes I can also see them inside other cities while I am sure that the segments are corrected because everything is correct in Cartouche. And also it is not a problem that the map has not been rebuilt after the recent edit because in the place which I have reported to the waze HQ the ghost city was still visible 9 months after being removed.

Please krankyd note that our aim is to make waze better. The correct solution is to fix a bug, not to keep saying it does not exist. Map editors color scheme is not buggy because all schemes show the same problem, the map editors just shows it better than others.

[rlpl]

Good to know that another party is willing to cooperate. I have investigated the problem a little deeper and played with the map editors color scheme. Despite what you say it cannot behave different than the official color schemes because the color schemes do not behave at all. They define the colors, patterns and other visual properties, they do not define data sources or algorithms. Also they do not distinguish the existing cities from the deleted cities. The main difference is that other color schemes do not display the city areas but still display the city names. Yes, if we hidden the city names the problem would seem to be solved.

I have managed to display not only the areas of the cities but also the borders. Now the cities look like the jigsaw puzzle mixed from different sets. The ghost cities have the partially missing borders or no borders at all. Like if the tile was aware that it displays the internal part of the city and not its border. Sometimes the ghost city partially covers the correct city: one tile displays the correct one with the correct borders and another one display the ghost with no border. (I really wish I could make a screenshot but for the moment I cannot). That makes me suspect that some tiles are not refreshed during the rebuild. Is that because nothing has changed in their areas? But something might have changed outside of them and may influence these tiles because now they may no longer be inside the cities they used to be.

Ghost towns were supposed to be removed from the maps, but apparently, some of them are still there. We will work on a fix to remove them from the DB, as well as from the maps.

I don't know how literally I should understand the part "remove from the DB" but I am afraid that if you just purge the deleted cities and leave the algorithm unchanged it will not work for two reasons:

1. Similar ghost cities may appear in future.
2. You will fix the problem only for the cities which have been deleted while the problem also applies to the cities which still exist but their areas have been collapsed and no longer cover half of the country

Again my hint: the tiling algorithm for Cartouche works perfectly and we all will be satisfied if you apply the same for the mobile client.

Anyway, removing the deleted streets and cities from the indexes is a good idea because they probably appear as the search hints and the search results while they should not.

[rlpl]

You are right: even if you rename the streets correctly to split the cities or to move them to their correct location, the wrong name remains forever in the client tiles where nothing has changed. This may be inside a city where the roads have not been edited for some time or in the large areas without the roads. Also note that the outdated areas may be different at different zoom levels. The "Map Editors" color scheme will display not only the wrong names but also the wrong areas.

[WeeeZer14]

Yes it is a known issue. Not sure if support knows why, but they are aware.

On the client, you can pick a color scheme. The one called Map editor shows town boundaries as shaded regions.

In my area, Belmont overlaps the real towns so you can't see the square for it.

I also have some towns (Lakeside) that are narrow long regions that I can trace through multiple states.

Anyway, if you can find a lone square with the wrong name, chances are at the center is a road segment with the wrong town information applied. Too bad we can't see town boundaries in the Cartouche.

My recent conversations with support seemed to think it was a problem with Cartouche guessing the wrong town name when a new segment was recorded. They were talking about turning off the auto-population of the town for newly recorded segments to deal with the problem for now.

[WeeeZer14]

Each week or month I see a new "phantom" town in my city. Currently Belmont, but has been places such as Cleveland, Washington, Lakeside and Columbus off the top of my head.