|
I am aware that this is an open (?) bug and has been the subject of discussion for a long time. My question is, is there an end in sight? I have been rendering villages in Angola as areas, and naming them (place=village, name=*) but none of these are rendered (neither name nor area). If I use landuse=residential, then they are rendered, and the map starts to have sense, but according to guidelines, this is not appropriate. If villages are not rendered (as places), then the map loses meaning - many parts of Angola appear uninhabited when in fact there are lots of small villages. Using a place point for the village is not practical, as many of the village names are not known - the important point is to have the area shading to show the rough limits on the map (IMHO) and the extent of repopulation. Should I switch to landuse=residential and revert all villages, or will this issue be addressed?
showing 5 of 15
show 10 more comments
|
The question has been closed for the following reason "This topic is turning into a long discussion of the kind not appropriate for this site. A github pull request has been submitted for the issue: discussion relevant to that can take place there. Otherwise please use the talk page of landuse=residential on the wiki." by SK53 04 Sep '17, 10:14
|
This is a quite old issue - see #103 for discussion on the issue tracker. As usually, the problem waits for somebody to make a pull request that will solve the issue. From looking at #103 nobody tried doing it so far. Unfortunately, this issue is not easy to fix so I would not recommend it as a task for somebody who never tried improving map style unlike some issues listed here (BTW, feedback from new and potential contributors is welcomed at #2291 ). As you say this is an old issue hence my original opening comment. However, technically, I believe that it is straightforward to fix, although I think that this is more than a technical issue, and that other part, while there appears to be consensus, will take longer. Unless I am mistaken, as referenced in 2792, the renderer only looks for place names in the points database. If that exclusion is removed, and it searches the polygons database as well, the rest should be good.
(01 Sep '17, 15:17)
stevenLAD
Further to the above, I have a pull that I hope will resolve this issue, but appear to have an issue with authentication. Any pointers to resolve (I am new to Github) would be appreciated
(01 Sep '17, 17:48)
stevenLAD
Pull submitted
(01 Sep '17, 21:40)
stevenLAD
As I read that that will added "landuse=residential"-style names for place=village and place=hamlet. What about places where there's both a named place=village way and also an overlapping but not necessarily identical landuse=residential? Is there an example rendering of it anywhere (or example screenshots)?
(01 Sep '17, 22:36)
SomeoneElse ♦
If there is an overlapping place and residential area, I would think that they are intended as one and the same, but that someone has just used landuse to get shading on the map. After all, these areas, unless identified as such are not administrative boundaries and so any residences outside the village area should normally fall within
(02 Sep '17, 09:33)
stevenLAD
|
Have you tried any other maps? The "standard" map is sub-optimal at many things, of which this is one. Please don't worry about what it does or does not render.
An example map that shows named place-village on ways is the cycle map on the osm.org website: https://www.openstreetmap.org/way/519612465#map=16/-11.1297/15.4294&layers=C .
Many other maps, including those used by Osmand, Maps.me and even Mapscii as well as those produced by i.e. Mapbox renders place areas fine. OSM-Carto increasingly appears to be the last hold out..
I wonder why you ask "is there an end in sight" on this forum. Only the developers of the carto-css style know what they are working on. According to a recent presentation, they do not have a real planning for the future. You can try to persuade one of them to work on the issue, or tackle it yourself. No one one this help-site has more powers than you to dictate/demand/ask what the carto-css maintainers will be working on.
Interestingly the current CartoCSS style shows these place labels on zoom 12 but not on any other zoom level: https://www.openstreetmap.org/way/519612465#map=12/-11.1296/15.3865. Looks more like a bug to me but I'm not a CartoCSS developer. Try to open a ticket at https://github.com/gravitystorm/openstreetmap-carto/issues
Its strange - the place names at zoom level 12 do not render everywhere - see this example https://www.openstreetmap.org/#map=12/-9.9837/14.8657
As you say, appears buggy.
At lower zoom levels, place names may lead to obfuscation, which is why area shading is useful. The above example shows an area that appears void of life, but in fact contains many small communities.
To answer Escada - the wiki shows this as a bug since 2016, and the related link shows that this has been discussed since 2012.
The corresponding CartoCSS issue is here: https://github.com/gravitystorm/openstreetmap-carto/issues/2794
Actually, I did have a brief look to see how easy it'd be to render these (not in openstreetmap-carto, in another style) and the tricky bit seemed to be how to avoid duplication with named landuse and with nodes that are the centre of something.
I'm not saying it's an "impossible" problem to fix, but it's not a straightforward one.
Removing duplication will take some time, especially as this has dragged on for 4 years, but it will happen, and rather look for a solution in code, lets rely on contributors to do what they do well - make good, clear, readable maps.
I have submitted code changes for villages and hamlets where the confusion should be least. If that is accepted, then it can be extended up the hierarchy if desired.
More delay = more duplication.
"If I use landuse=residential, then they are rendered, and the map starts to have sense, but according to guidelines, this is not appropriate."
What guidelines are you referring too that would prohibit this? In my opinion, this is exactly what landuse=residential is for: marking out inhabited areas that may not necessarily have a name. The Wiki only states: "Most landuse=residential objects in OSM are nameless", which doesn't prohibit using name on residential areas. In fact, I think there are quite a number of residentials with names in cities, e.g. combined with place=neighborhood/quarter tags etc, so the statement that "most landuse=residential areas are nameless" may even be disputable
"If villages are not rendered (as places), then the map loses meaning - many parts of Angola appear uninhabited when in fact there are lots of small villages. Using a place point for the village is not practical, as many of the village names are not known - the important point is to have the area shading to show the rough limits on the map (IMHO) and the extent of repopulation."
See my first remark, I don't understand the issue as presented in the context you specify of "places-without-names".
A place (village, town, city etc) is a superset of land uses - (roads, housing, amenities.....) and is the key used to assign names. It may also have an administrative boundary which may be much larger.
The wiki states that it is preferred that roads for example, should not be included in residential land use which precludes its use for correctly mapping villages etc.
All villages have names, but the contributor may not know that name when he actually maps the village hence the "places-without-names". Defining and showing the village area without a name invites other contributors with better local knowledge to complete.
To the point raised by SomeoneElse, the village should probably be mapped below the land uses with a lighter shade so that if, generally in larger places, there are multiple land uses defined, these are not masked but included, but that is a refinement and in the first instance, lets get the villages "on the map".
Here in the Netherlands where I live, and I think in many other countries, residentials do include residential roads, and quite often other road types as well, and many other stuff. I think the Wiki is certainly not a definitive guide for real world mapping practices regarding this specific tag.
I also have some doubts about the usefullness of tiny residential areas. The OpenStreetMap wiki almost suggests only private land and cadastral plots could be residential ("If you had access to land parcel data, you'd probably draw the ways with landuse=residential along the parcel edges"), while on the other hand the Wikipedia page for "Residential area" linked from the OSM wiki page, is far less strict and talks of residential in more general ways ("A residential area is a land use in which housing predominates, as opposed to industrial and commercial areas."..."The area may be large or small." - https://en.wikipedia.org/wiki/Residential_area).
Looking at this random area in the Netherlands http://www.openstreetmap.org/#map=16/51.9738/4.9415, the landuse=residential encompasses many areas and it would look as though it actually represents the town. Given that every house is drawn (showing defacto residential landuse), landuse appears to be used as the proposed place area would be used to highlight the town at lower zooms the limits. There is also an administrative boundary with the name and the residential area is unnamed
The wiki is a standard that has been set so that there are common guidelines - not everyone may agree with it, and thats why there are places where changes can be discussed and agreed.
"Looking at this random area in the Netherlands http://www.openstreetmap.org/#map=16/51.9738/4.9415, the landuse=residential encompasses many areas and it would look as though it actually represents the town."
Yes, exactly, this is how generally the landuse=residential is used in the Netherlands (but I think also in many other European countries).
@mboeringa, I do not believe that administrative entities match with landuse=residential in many countries. A village or place is more than landuse=residential. I have mentioned this on the Dutch forum in the past and people agreed.
I don't understand what administrative entities have to do with this. I wholeheartedly agree they are not the same as landuse=residential. Administrative entities almost always cover the entire surface area of a country, and often contain large swaths of natural or agricultural land. I also agree a "place" is more than landuse=residential alone. It contains industrial,commercial and railway areas as well, all of which have long established tagging schemes in OSM, and are in use in the Netherlands. The combination of these four landuses is though, a pretty good indicator of the extent of a "place". Rather than introducing rendering for yet another tag, with all the burden of extra layers in styles (yes, these things do come at a cost!), I would urge anyone to keep using these well established taggings to define the extent of a "place".