Hi Felix, I can now reproduce the poor result with norway when I use both --polygon-file and --precomp-sea. Problem seems to be that splitter decides to try the slower algo first. After a while it gives up and tries the alternative faster algo and that would find a good solution soon but splitter gives up too early, so this is really not an improvement. Without the polygon the slower algo finds a good solution. I am working on a new solution which executes both algos in parallel and selects the better result. This works much better for this test case but sometimes worse when option --num-tiles is used. reg. Saarland with polygon: r614 stops with an error message (and returns 1 to signal a bad split). Your scripts seems to ignore this and just counts the number of written tiles? I'll try to find a fix so that splitter doesn't try to fit the tiles into the polygon when this happens... Gerd ________________________________________ Von: mkgmap-dev <mkgmap-dev-bounces@lists.mkgmap.org.uk> im Auftrag von Felix Hartmann <extremecarver@gmail.com> Gesendet: Mittwoch, 30. Juni 2021 08:55 An: Development list for mkgmap Betreff: Re: [mkgmap-dev] splitter r609 released okay yeah I think I misunderstood the polygon-file a bit. But why is the split for Norway so much worse now? Before it was possible to get it down to 115 tiles (using the polygon-file option) now its 168 instead. Not using polygon-file no change. (guess the 1 tile more is maybe added data somewhere). On Wed, 30 Jun 2021 at 07:59, Gerd Petermann <gpetermann_muenchen@hotmail.com<mailto:gpetermann_muenchen@hotmail.com>> wrote: Hi Felix, reg. Saarland: I'll try to reproduce. Looks like an error. reg. Liechtenstein: I think we still have different ideas what the polygon-file option is about. The documentation is a bit outdated since r433 (see https://www.mkgmap.org.uk/websvn/revision.php?repname=splitter&rev=433 ) but the basic idea was that splitter should try to produce tiles which don't overlap the given polygon too much. This sometimes means that it produces more and smaller tiles. Splitter should be able to find good splits for geofabrik downloads without a polygon-file, but there are use cases: If the user only wants a part of a download in the map they can use the corresponding polygon. You seem to suggest that splitter should just use the polygon to be able to ignore data that is in the input file but not in the polygon and never try to fit the tiles into the polygon? Gerd ________________________________________ Von: mkgmap-dev <mkgmap-dev-bounces@lists.mkgmap.org.uk<mailto:mkgmap-dev-bounces@lists.mkgmap.org.uk>> im Auftrag von Felix Hartmann <extremecarver@gmail.com<mailto:extremecarver@gmail.com>> Gesendet: Dienstag, 29. Juni 2021 23:04 An: Development list for mkgmap Betreff: Re: [mkgmap-dev] splitter r609 released I found (server) time to run my test again - using splitter v614 - search limit 1000000 Problem cases in bold. I also added the number of tiles that splitter v602 needed (well on one week older geofabrik extract - but that should not make such a difference for Norway or South America). I feel 602 with search-limit 1000000 was producing more consistent results. "for netherlands use polygon-file - cnt1 = 93 cnt0 = 88" "for alps do not use polygon-file - cnt1 = 225 cnt0 = 226" "for liechtenstein do not use polygon-file - cnt1 = 1 cnt0 = 6" I thought that the new version fixes the splitting if it is not actually needed. "for monaco do not use polygon-file - cnt1 = 1 cnt0 = 2" "for slovenia do not use polygon-file - cnt1 = 27 cnt0 = 28" "for ukraine use polygon-file - cnt1 = 62 cnt0 = 61" "for norway do not use polygon-file - cnt1 = 129 cnt0 = 168" Big degradation here - with the older splitter it was: "for norway use polygon-file - cnt1 = 128 cnt0 = 115" (also using 1000000 search limit) "for switzerland do not use polygon-file - cnt1 = 29 cnt0 = 30" "for poland do not use polygon-file - cnt1 = 127 cnt0 = 128" "for sweden use polygon-file - cnt1 = 60 cnt0 = 54" "for finland do not use polygon-file - cnt1 = 65 cnt0 = 84" "for czech-republic use polygon-file - cnt1 = 63 cnt0 = 62" "for denmark use polygon-file - cnt1 = 33 cnt0 = 32" "for austria use polygon-file - cnt1 = 56 cnt0 = 54" "for andorra do not use polygon-file - cnt1 = 1 cnt0 = 4" "for estonia use polygon-file - cnt1 = 9 cnt0 = 8" "for saarland use polygon-file - cnt1 = 4 cnt0 = 0" Seems using a polygon-file no spit at all - however it should still write the data again out into the newly defined filename. "for hamburg do not use polygon-file - cnt1 = 3 cnt0 = 12" "for hessen do not use polygon-file - cnt1 = 17 cnt0 = 18" "for bayern do not use polygon-file - cnt1 = 47 cnt0 = 48" "for berlin do not use polygon-file - cnt1 = 5 cnt0 = 13" "for australia-oceania use polygon-file - cnt1 = 112 cnt0 = 109" "for south-america do not use polygon-file - cnt1 = 335 cnt0 = 463" old splitter "for south-america do not use polygon-file - cnt1 = 337 cnt0 = 339" (using default search limit) "for africa do not use polygon-file - cnt1 = 653 cnt0 = 663" "for asia use polygon-file - cnt1 = 1697 cnt0 = 1549" "for russia use polygon-file - cnt1 = 429 cnt0 = 408" 0ld identical "for central-america use polygon-file - cnt1 = 60 cnt0 = 58" "for antarctica use polygon-file - cnt1 = 7 cnt0 = 6" "for morocco do not use polygon-file - cnt1 = 19 cnt0 = 27" old identical "for congo-democratic-republic do not use polygon-file - cnt1 = 37 cnt0 = 38" "for mozambique use polygon-file - cnt1 = 24 cnt0 = 23" "for azerbaijan do not use polygon-file - cnt1 = 3 cnt0 = 4" "for malaysia-singapore-brunei do not use polygon-file - cnt1 = 16 cnt0 = 17" "for china use polygon-file - cnt1 = 104 cnt0 = 101" "for india do not use polygon-file - cnt1 = 109 cnt0 = 112" "for indonesia do not use polygon-file - cnt1 = 175 cnt0 = 200" "for japan use polygon-file - cnt1 = 163 cnt0 = 160" "for philippines do not use polygon-file - cnt1 = 50 cnt0 = 51" "for afghanistan do not use polygon-file - cnt1 = 11 cnt0 = 20" old identical "for australia use polygon-file - cnt1 = 67 cnt0 = 66" "for argentina use polygon-file - cnt1 = 31 cnt0 = 27" "for brazil use polygon-file - cnt1 = 176 cnt0 = 172" "for chile do not use polygon-file - cnt1 = 31 cnt0 = 33" "for paraguay use polygon-file - cnt1 = 17 cnt0 = 16" "for us-midwest do not use polygon-file - cnt1 = 152 cnt0 = 154" "for us-northeast do not use polygon-file - cnt1 = 109 cnt0 = 110" "for us-pacific use polygon-file - cnt1 = 20 cnt0 = 18" "for us-south do not use polygon-file - cnt1 = 259 cnt0 = 261" "for us-west do not use polygon-file - cnt1 = 224 cnt0 = 227" "for greenland use polygon-file - cnt1 = 4 cnt0 = 2" "for mexico use polygon-file - cnt1 = 43 cnt0 = 42" "for reunion do not use polygon-file - cnt1 = 3 cnt0 = 5" On Fri, 25 Jun 2021 at 16:51, Gerd Petermann <GPetermann_muenchen@hotmail.com<mailto:GPetermann_muenchen@hotmail.com><mailto:GPetermann_muenchen@hotmail.com<mailto:GPetermann_muenchen@hotmail.com>>> wrote: Hi all, I think with r609 there should be no need to use a polygon file. See https://www.mkgmap.org.uk/websvn/revision.php?repname=splitter&rev=609 Please let me know when this version produces much worse results compared to older releases (using the same options and input) and provide the files densities-out.txt and the log. If you are interested in good splits you should check the splitter log for "Solution is not nice. Can't find a better solution" When this is printed splitter did not find a good split. Expect almost empty tiles in this case. It is likely that this happens when rather small files are split with a polygon and the default resolution. Normal users don't do this, but some map providers try to use the same options for very different downloads ... Gerd _______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk<mailto:mkgmap-dev@lists.mkgmap.org.uk><mailto:mkgmap-dev@lists.mkgmap.org.uk<mailto:mkgmap-dev@lists.mkgmap.org.uk>> https://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev -- Felix Hartman - Openmtbmap.org & VeloMap.org _______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk<mailto:mkgmap-dev@lists.mkgmap.org.uk> https://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev -- Felix Hartman - Openmtbmap.org & VeloMap.org