[Patch] multipolygon areas are rendered, even when they have not been completely downloaded

[cmuelle8]

Is there a reason why incompletely downloaded MPs clutter the mapview?

There is a performance feature to turn off area rendering all together. Can we have this selectively, just for those multipolygons that have not been downloaded completely?

This would also aid correctness of MP relations - if an MP is not complete and correct, do not display it at all.

[cmuelle8]

I'd code it myself, if nobody steps forward. I'm asking to hear your opinion about it up front..

[akks]

It can be a good feature, but it does not seem good for me to turn it on by default. The reason is very big multipolygons like forests (~ 30 ways, total 2000 nodes), that usually are not downloaded completely. If you turn off green coloriring, all wood boundaryies will look like untagged ways and confuse novices.

Here is an example of working partial rendiring:

[cmuelle8]

Replying to akks:

If you turn off green coloriring, all wood boundaryies will look like untagged ways and confuse novices.

As I understand, coloring the way is unaffected by turning on "draw boundary of areas only". I've double checked, if I use the feature, ways themselves are rendered as expected. Partial rendering might buffle novices even more, as an edge will always exist somewhere that does not have a corresponding way in the editor.

Greetings

[skyper]

+1 for making drawing of incomplete MP configurable

Maybe it would help to render them diffent.

See also #6020 and #6338.

[simon04]

An example where the partial rendering is distracting and incorrect

[simon04]

attachment:7635.patch​ tries to detect problematic cases and stop drawing such multipolygons unless explicitly enabled (preference key mappaint.multipolygon.incomplete).

[simon04]

Fixed in the course of partial rendering – thanks to bastiK, see #12104

[cmuelle8]

Adding as info: In OSM any multipolygon ring, with respect to an interior, can run around clockwise or counter-clockwise around that interior.

This means for incomplete rings, we can only "guess" the side the interior will be on. Rendering an inside, partially by a buffer band or not, to a specific side of an incomplete ring will always have a chance of filling the incorrect side.

Even if we manage to correctly draw the buffer band to either the left or right side of an open ring, chance is that this side is not the side the interior is on.

If we have a complete ring, there is still the chance that role info for that complete ring is incorrect. But at least we then draw correctly with respect to the role information provided. To determine if role info is plausible, the whole MP data is needed.