[mkgmap-dev] generate-sea option and faster-mp changes

Fri May 13 15:42:39 BST 2022

Hi Gerd

Great stuff. I can understand how errors could lead to exact 0/90/180
but not 45!

Thanks for tidying and applying SeaGenerator patch. I now notice that 
CreateSeaWay(enlarge = true) is never called as a result of my
restructure and this bit of logic could be deleted.

Ticker

On Fri, 2022-05-13 at 08:09 +0000, Gerd Petermann wrote:
> Hi Ticker,
> 
> ouch, the test with precompiled-sea failed. I found artifacts in some
> generated tiles, but they are not caused or fixed by your patch or
> mine.
> E.g. sea_2490368_-7766016.osm.pbf contains some nodes like
>         <node id="11858" lat="0.0" lon="-180.0"/>
>         <node id="11859" lat="-90.0" lon="-180.0"/>
>         <node id="11860" lat="-90.0" lon="-90.0"/>
>         <node id="11861" lat="0.0" lon="-90.0"/>
>         <node id="11862" lat="45.0" lon="-180.0"/>
> with a land-polygons-split-4326.zip from 2021.
> Again they disappear when I switch off the partitioning in
> MultiPolygonRelation.
> 
> I'll download current data and try to reprpduce it...
> 
> Gerd
> 
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag
> von Gerd Petermann <gpetermann_muenchen at hotmail.com>
> Gesendet: Freitag, 13. Mai 2022 09:20
> An: Development list for mkgmap
> Betreff: Re: [mkgmap-dev] generate-sea option and faster-mp changes
> 
> Hi Ticker,
> 
> attached is my patch to fix the special case when partitioning cuts a
> multipolygon so that only two equal rectangles are left.
> 
> This also fixes the problem for me, but I guess it might fail in edge
> cases where the cutting goes through a point where
> inner and outer diverge.
> 
> Reg. your patch: Looks good, I would just rename NATURAL_TAG_KEY to
> TK_NATURAL so that it matches the naming in other sources.
> 
> I am running some tests reg. precompiled sea at the moment...
> 
> Gerd
> 
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag
> von Gerd Petermann <gpetermann_muenchen at hotmail.com>
> Gesendet: Donnerstag, 12. Mai 2022 14:55
> An: Development list for mkgmap
> Betreff: Re: [mkgmap-dev] generate-sea option and faster-mp changes
> 
> Hi Ticker,
> 
> I can reproduce the error with the old code and that error disappears
> when I disable the partitioning in class MultiPolygonRelation.
> I've not yet understood why, but I fear your patch simply works
> around this problem.
> The partitioning seems to produce parts where an inner and an outer
> rectangle both seem to cover the same area.
> I think I have to fix that as well.
> 
> Gerd
> 
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag
> von Ticker Berkin <rwb-mkgmap at jagit.co.uk>
> Gesendet: Donnerstag, 12. Mai 2022 14:29
> An: Development list for mkgmap
> Betreff: Re: [mkgmap-dev] generate-sea option and faster-mp changes
> 
> Hi Gerd
> 
> Looking at older versions (2019), createLandWay() makes a polygon
> with
> the tile bounds; no extra. This hasn't changed since then.
> 
> createSeaWay() has flag to enlarge the polygon slightly and use of
> this
> hasn't changed in the release version
> 
> Ticker
> 
> On Thu, 2022-05-12 at 08:37 +0000, Gerd Petermann wrote:
> > Hi Ticker,
> > 
> > thanks, I'll take a look at it. My understanding was that
> > SeaGenerator should produce an outer way which is slightly larger
> > than the tile.
> > Did I change that?
> > 
> > Gerd
> > 
> > ________________________________________
> > Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag
> > von Ticker Berkin <rwb-mkgmap at jagit.co.uk>
> > Gesendet: Donnerstag, 12. Mai 2022 08:20
> > An: mkgmap development
> > Betreff: [mkgmap-dev] generate-sea option and faster-mp changes
> > 
> > Hi Gerd
> > 
> > I've just rebuilt britain-and-ireland-latest.osm.pbf with the
> > current
> > trunk and find that many tiles on the coast have flooded
> > rectangles.
> > The coastline is all correct and these flooded area are normally
> > well
> > inland and unrelated to where the coastline crosses the tile
> > boundary.
> > 
> > I use option
> >   --generate-sea="multipolygon,extend-sea-sectors,close-gaps=750"
> > and this worked fine up to about a year ago - I've only just
> > noticed
> > the problem.
> > 
> > Cutting this from a suitable area (Wales, UK, Britain, Europe etc)
> > 
> > 74220030: 2408448,-331776 to 2430976,-196608
> > #       : 51.679688,-7.119141 to 52.163086,-4.218750
> > 
> > and building with minimal options plus above and default style
> > should
> > demonstrate the problem. There are also extra error messages
> > starting
> > 
> > ... 030.osm.pbf: Multipolygon generated SeaPolygonRelation [nat...
> > 
> > in the log file, which I attach
> > 
> > I think it is related to the faster-mp changes and the MP structure
> > SeaGenerator created. This has 1 outer (the tile) and all bits of
> > land
> > are inners, including land touching the edge of the tile.
> > 
> > I've attached a patch for an enhanced version of SeaGenerator that
> > distinguishes between land/sea areas that touch the edge and true
> > [anti-]islands and only use MP processing for the islands. With
> > this,
> > tiles are correct.
> > 
> > It also has other benefits: It is clearer in what is happening,
> > doesn't
> > do a lot of pointless tagging of ways that will joined to become
> > polygons, etc
> > 
> > Ticker
> > 
> > _______________________________________________
> > mkgmap-dev mailing list
> > mkgmap-dev at lists.mkgmap.org.uk
> > https://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
> 
> _______________________________________________
> mkgmap-dev mailing list
> mkgmap-dev at lists.mkgmap.org.uk
> https://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
> _______________________________________________
> mkgmap-dev mailing list
> mkgmap-dev at lists.mkgmap.org.uk
> https://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
> _______________________________________________
> mkgmap-dev mailing list
> mkgmap-dev at lists.mkgmap.org.uk
> https://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev