osmium extract keeps some ways outside of the polygon

[andrzejkurczewski-tomtom]

What version of osmium-tool are you using?

osmium version 1.14.0 libosmium version 2.17.3

What operating system version are you using?

CentOS Linux 7

What did you do exactly?

I downloaded the Planet PBF from https://planet.openstreetmap.org/ I used osmium to extract Croatia using complete_ways strategy:

osmium extract --strategy complete_ways -c extract_config.json input.pbf

where config is as follows:

{
  "extracts": [
    {
      "output": "HRV.osm.pbf",
      "output_format": "pbf",
      "polygon": {
        "file_name": "HRV.geojson",
        "file_type": "geojson"
      }
    }
  ],
  "directory": "(...)"
}

Find geojson attached: HRV.zip It should be the same as country boundary in OSM, but it has a few months already.

What did you expect to happen?

I expected an output file to contain ways and nodes located in or crossing the Croatia border.

What did happen instead?

There are ways outside Croatia (located in Slovenia, Italy and Austria) included. Below you can find the resulting PBF filtered with osmfilter for highway=path only, for easier visualization: complete_paths.zip And the visualization:

Sorry for such large-scale example. I failed to reproduce this issue when extracting smaller areas. However, it did reproduce for multiple country polygons, including Austria, Czechia, Greece, Hungary or Italy.

In all cases the unexpected ways seem to cross the north edge of the country bounding box.

[joto]

I can reproduce the problem, so there is definitely a bug there. Looks to me like some rounding problem at the bounding box edges or so. This needs some investigation.

[andrzejkurczewski-tomtom]

Maybe one more useful remark - it seems the problem doesn't occur with simple strategy.

[joto]

@andrzejkurczewski-tomtom Are you sure? When I tested it, it happened with the simple strategy, too.

[andrzejkurczewski-tomtom]

Croatia (exactly the same geojson) was the only country I checked with simple:

osmium extract --strategy simple -c extract_config.json input.pbf

And that was my result: simple_paths.zip

But of course I cannot be sure it won't happen for simple for any other input.

[joto]

I believe this is fixed in the current master. Can you check whether it works for you?

[joto]

No feedback, so assuming this has been fixed.

[andrzejkurczewski-tomtom]

Hi @joto, sorry for the late response. I can confirm it works for me with 1.15 release, thanks a lot!