Mapping landcover with multipolygons

If you have a road with 4 landuses glued to it and 5-6 hiking/bike routes attached to it, pressing the middle mouse button in JOSM to select the right thing can still make it hard to find the right one, and sometimes it’s actually out of the screen, depending on the number of attributes…

2 Likes

Happened and happens to me once in a while, thinking to hit the F key (Follow-Trace) with eyes ‘glued’ to the screen, but then placing the fat finger on the right next G instead, and that line, whatever that line, gets unglued from it’s neighbour in JOSM. You get a warning, but the warning stays so long on the screen, you don’t see you might have done that serially. Then you map on and discover at the end when the validator kicks in that you’ve unglued an outline with hundreds of nodes. Not found a function to easily reglue all those nodes… little doubt a function or sequence exists to do that. :o(

Back to the main, A single outline with simple polygons, 360 inside, made into a MP relation has to be many/most of the times. In that an MP is unavoidable. The fun starts when side-sharing ‘inside’ SP areas have each individually been given the inner role. The wiki is though clear in that one draws a single ring along the ‘outer’ edge of those side sharing inner areas and make that ‘inner’ to the forest or whatever.

I agree that this complicates things unnecessarily, landcover mapping can be entirely done without MPs. You would just need to split the landcover if there is another type in the middle of it. However, I can understand that if you mapping landuse that splitting that is unwanted, and an MP is required to put something in the middle of it.

MPs like this: https://www.openstreetmap.org/relation/168553 are bad for maintainability and newcomers. This is just too big, Imo this should at least be split along bigger features like railways and tracks.

2 Likes

You mean touching inner ways in an MP? This has been allowed for quite a while now:

I remember this was an issue back in the days, though.

It is prone to error. Please provide a video ad i did. At any state modifying the relation you may fuck up the geometry. Even for the simple example i provided its at least 5 modification to the relation and every single left out breaks the polygon.

So keeping history is more important than keeping a valid polygon geometry? Are you sure? And BTW: Same argument as yours: “Software needs to be fixed”. But this time it only affects history, not the validity of geometry.

I chose this way because its the least “prone to error” process in working with polygons.

but they aren’t as bad as some people are trying to picture them and people can screw up ways and MPs equally hard.

I am now with OSM for 15 years and i have roughly ~200 Days mapping in a Year - I wrote validators, and other random QA stuff for OSM, and from my experience: yes - Relations are whats fucking up OSM Data. Not only in polygons, but also in turn restrictions or route relations etc.

Relations are hard as a data model, people dont understand them, error messages are hard to get, and in the end people simply press save to get their “work done”. And then you end up with broken geometries, turn restrictions or route relations.

And for me - If i see a region with MP landuse mess, i dont even touch it - Its just a huge mess and even me, with 15 years experience, i dont want to touch it. My time is to valuable to even get into this MP mess.

Just have a look at deteriorating route relations - nobody except some very brave mappers want to touch them.

See - and i dont even need to go down that path. From my Video splitting or unglueing landuses I am not issuing a single click on some glued stuff. Thankfully josm allows to select a polygon by double clicking within the polygon.

Flo

6 Likes

Not sure which of the main QA progrs keep flagging this. Anyway, it does not bother me to flash an outline around those ‘touching’ inner areas. The F key trace does like 1000 nodes per minute i.e. it’s seconds to create one and if you have large forest/wood area with an inside zone of residential with varying farming patches strewn around it, there’s enormous fragmentation here, it’s one single inner instead of 20-30.

Edit: A fresh example, courtesy of OSMI, a bunch of touching inner areas included flagged with the primary warning if 1 way in multiple rings.

Horrible! Even building outlines are multipolygons. Simple things like making the corners of a building rectangular in iD no longer works! Newbies won’t edit something like that or there’s a big risk of breaking something.

In these cases (with really a lot of nodes) is my workaround:
Split the outline (creating a MP), delete the unneeded section, redraw the correct section and convert the MP back to a normal polygon.

Flo, I’m not going to record a video to show you how to deal with MPs. In the area where I map, there are close to none, because we all use ways. It’s nice that you have 15 years of experience in mapping, really, but I don’t think that’s the point here or makes your argument stronger than anyone else’s. All I was saying is: you can f*** up ways and you can f*** up MPs, I’ve seen both and I can repair both. I personally don’t prioritize one over the other and only use MPs when I need outer/inner geometry. But I’ve been on the other side as well and if all your landuses are MPs, then it’s actually pretty nice to work with, as shown in the initial video.
By the way: you can split MPs with alt+x, so if you’re double-clicking to select the way/MP, your workflow can be the same. No need to get into this even further, I am not an advisor for MPs.

That reply was for someone else. In JOSM, if you want to change a road’s attribute and there are ways glued to it, you sometimes have to press the middle mouse-button in order to select the road and not one of the landuses. I was just explaining, why it’s annoying to glue ways to roads. It was not related to MP vs ways, even though this problem doesn’t exist if a road is part of an MP, because at least JOSM never selects the relations of a way if it’s part of any.

JOSM and ID.

You started the argument that its a matter of knowledge. So i am making the point that i pretty shure know the issue.

And the point i make is that for the average mapper there are a bazillion ways to fuck up MP based landuses, but its basically impossible to break the geometry using the way i showed.
And you still argued in favor of MPs. So i asked you to show “how easy it is” compared to the process i outlined and showed. You refuse to - so i assume that you pretty well know how error prone MP based landuses are.

This works only if there are closed rings in the MP. This was not the point here - The question here is whether MPs with individual ways would be better. When you build your MP outer ring with individual ways (As the original request was) you cant use alt-x because there is no such thing as a closed outer ring so alt-x does not work.

And when you accidentally try so by double-click in the middle and mark the whole MP Relation and press alt-x you basically end up with a broken Multipolygon/Relation without noticing (Even the validator thinks its okay)

Flo

This is why we had the huge discussion in 2018 and the result was that glueing ways and polygons is the non-preferred way and should not be used.

Flo

When we have detailed mapping, not using a way for the adjacent polygon would mean unnecessarily redrawing hundreds of nodes. This makes everything more complicated. Does someone really believes that reclicking all the nodes of this meadow for the forest would be simpler that reusing the same line for the forest?

I guess that’s a misunderstanding. You were the first person to tell @Janjko that if he thinks it’s easier to maintain, he’s using the wrong tools. To me that sounded like “you’re doing it wrong”. Let’s stop the discussion here, it can only get personal. Probably just a misunderstanding, nothing a beer or 2 can’t solve.

Uh … no? It works perfectly fine if the outer ring of the MP is made up of several individual ways. Assuming that we’re talking about MPs that don’t have an inner ring, the behavior is identical to a way. At least in my JOSM. Might be a plugin, no idea, but it just works.

That one doesn’t do anything for me :thinking: If I don’t select 2 nodes, or a way, alt+x doesn’t do anything when the MP is selected. Same as with a way, but the way will give an error saying that you have to select 2 nodes.

You can use F Help/Action/FollowLine – JOSM to add an segment to the way. This way the nodes are reused like this: Node: 759485465 | OpenStreetMap

2 Likes

I stand corrected. This seems to got fixed. It even sorts the inners into the correct MP and splits the MP. Okay - interesting.

I ended up with broken MPs with various issues over the years which made be pretty cautions of not using alt-x on anything MP related. I couldnt quickly reproduce the issues.

Flo

Thanks for the tip. It does make it easier than reclicking but not easier than making an MP.

Yes!

Btw.
Please explain to me what the difference is between
this way,
this way, '*
this way and
this way.

'* CS comment by @SekeRob:

OSMI signalled, change forest outline sections from inner to outer. Merge a number of short outer elements to reduce member count.

Obviously, there were more such short, meaningless relation members.

I suppose they were drawn at different moments from different sources, as the source tag indicates. It is difficult / impossible to edit only a part of a polygon with a different source without using relations.

“F for follow” works in other editors too (at least iD and Potlatch).

2 Likes

I do not see any need for this approach. Indicate the source in the changeset.

And why had you mapped such simple geometries as buildings as multipolygons? One click to select all points of a building outline and one keystroke to unmap it from the neighbouring building.

This ist not KISS!