[osmosis-dev] Improve relation extraction handling

[Maxim Dubinin]

Hi Frederik,

+1 for the scheme you proposed, we're creating daily subtracts for
former USSR countries (with completeways) and are concerned with missing relations as well. The
scheme you proposed was along the lines of the solution that we've
discussed to address this issue just recently.

We did not start any development.

Maxim
http://gis-lab.info

Вы писали 7 ноября 2010 г., 5:44:26:

FR> Hi,

FR>     I'm a major user of the --bp task in Osmosis, for the nightly 
FR> Geofabrik country extracts. I cannot use the complete-something flags as
FR> that would take far too much time, and on the whole the 
FR> "clip-incomplete-entities" mode seems not to be too bad for users.

FR> There's one thing that has popped up on this list however ("lost 
FR> relations while clipping") which is an increasing concern to our users
FR> and which I hope I can fix. The problem is that with cascading 
FR> relations, Osmosis will only include a relation if one of its children
FR> has been included already, so if

FR> node N1 is inside the clipping area,
FR> relation R1 contains node N1,
FR> relation R2 contains relation R1

FR> then relation R1 will alway be included in the output, but relation R2
FR> will only be included if it comes after R1 in the input file.

FR> Now while using the temporary storage for nodes and ways would be too 
FR> expensive for the nightly Geofabrik build, I think that temporarily 
FR> storing relations would be acceptable (as there aren't too many). I'm 
FR> thinking of adding a "cascadingRelations=yes" flag to the --bp/--bb 
FR> tasks that would

FR> * process relations not containing other relations as usual,
FR> * throw all other relations onto a stack and process them in a suitable
FR> order afterwards

FR> I am aware that determining a "suitable order" could be difficult if you
FR> have many levels of cascasing (or even circular) relations, and that the
FR> output file would not be ordered properly, but I think such an option 
FR> would mean a big improvement at a relatively low cost.

FR> 1. Any comments?
FR> 2. Anyone done something like that already?
FR> 3. Is any of Osmosis' internal storage mechanisms suitable for buffering
FR> relations in this way or do I have to invent something new?

FR> Bye
FR> Frederik