Hi Ticker, the problems with AllElements are not new. I see them also with much older versions of mkgmap. Gerd ________________________________________ Von: mkgmap-dev <mkgmap-dev-bounces@lists.mkgmap.org.uk> im Auftrag von Ticker Berkin <rwb-mkgmap@jagit.co.uk> Gesendet: Donnerstag, 7. November 2019 12:56 An: mkgmap development Betreff: Re: [mkgmap-dev] type/subtype of points and cities Hi Gerd MdrDisplay/cityPtrSize: I didn't change any significant behaviour. Because it seemed inconsistent, I put in a diagnostic. I also made the switch stmt in printSec11_city like the similar instances that do have to handle 1/2/3 byte objects rather than switch ... case 3: ... default: ... I should also have added handling cityPtrSize of 4. I fixed mkgmap to work correctly for this. However, when I get a tile that needs 4 byte pointers, the gmapsupp.img size jumps up, so I re-split the map with a different number of nodes until the city count reduces again. My preference here would be rename one of the cityPtrSize variables and keep/extend the switch statement. Happy to get rid of the test/system.out.printf MdrCheck/group/toType: The original code in check10() tried to re-constitute a type/subtype as per mkgmap from group/someBits data and it didn't do it fully or correctly. I agree that this tool should just display/validate the understanding of the Garmin IMG structure. However, in this case, because the fullType generated above is then validated against data from mdr11, it has to do this validation based on the grouping chosen by mkgmap. Similarly, if these mappings are changed in mkgmap, equivalent changes need to be made to Display (I put in a few comments to this effect in the relevant places) AllElements: On my Etrex device, it shows some points not in the correct place according to sequence with strange types or names. Some of these points are outside the map area! I can't quite remember the details, it has been a while since I last used it. So yes, I'm sure it is generating something it shouldn't and maybe the lower levels of mkgmap code should be checking/objecting as well. It might be related to indPoints sometimes having subtypes; I think there are assumptions that these are a fixed size items. I don't know if this will cause the problems you describe, but I'll have a look at it. Ticker On Thu, 2019-11-07 at 09:15 +0000, Gerd Petermann wrote:
Hi Ticker,
I've not yet understood the changes regarding cityPtrSize. Do you think that Garmin supports one-byte pointers when there are less than 128 cities? The transalpin demo map contains only a few cities or regions but uses two bytes. I've attached a patch with my proposed changes.
One more hint: I don't like the description "Understands the same MDR groups as generated by mkgmap so it can recreate the correct full type." My understanding of display tool is this: - The main purpose is to find out how the Garmin algos encode things in IMG format, so we should use maps produced by Garmin and Mapsource/Basecamp to verify. The code in mkgmap should be the results of those findings, not vice versa. - I also use it to verify that changes made in mkgmap don't do something unexpected.
I've just learned that Mapsource doesn't like the map produced with java -jar mkgmap.jar --index --gmapi test-map:all-elements It crashes when you search for all POI and also when you search for cities (both without giving a name)
Gerd
________________________________________ Von: mkgmap-dev <mkgmap-dev-bounces@lists.mkgmap.org.uk> im Auftrag von Ticker Berkin <rwb-mkgmap@jagit.co.uk> Gesendet: Dienstag, 5. November 2019 12:44 An: mkgmap development Betreff: Re: [mkgmap-dev] type/subtype of points and cities
Hi Gerd
Here is patch for Display.
Changes are: Couple of extra diagnostics. Handles 1 byte cities. Moves a call of getSection(11).get... out of a loop. Consistent handling/display of point type/subtype. Understands the same MDR groups as generated by mkgmap so it can recreate the correct full type.
Ticker
On Tue, 2019-11-05 at 09:43 +0000, Gerd Petermann wrote:
Hi Ticker,
okay, looks good to me. In an earlier post you mentioned that display tool needs changes, too. Please post that patch as well.
Gerd
________________________________________ Von: mkgmap-dev <mkgmap-dev-bounces@lists.mkgmap.org.uk> im Auftrag von Ticker Berkin <rwb-mkgmap@jagit.co.uk> Gesendet: Montag, 4. November 2019 18:12 An: Development list for mkgmap Betreff: Re: [mkgmap-dev] type/subtype of points and cities
Hi Gerd
To stay compatible with "Openfietsmap full" style, I've updated the patch to change the range of city to be: type >= 0x0100 && type < 0x1100; Release mkgmap had this as: type >= 0x0100 && type <= 0x1100;
In release mkgmap, a point with value 0x1100 was added to the RGN as an indPoint, but not indexed as a city by MDR. Also 0x1100 is not findable as a nearby city, at least on Garmin eTrex. I couldn't detect any useful behaviour for this combination.
Ticker
On Tue, 2019-10-22 at 15:03 +0000, Gerd Petermann wrote:
Hi Minko,
"(like you say)" should have been ... "(like Ticker wrote)" ... (a small dot with the label) ... sorry, I forgot to use the typ file, so 0x1101 is displayed with an icon showing a bus and without the label. But why do you use a type that is not indexed?
Gerd
________________________________________ Von: mkgmap-dev <mkgmap-dev-bounces@lists.mkgmap.org.uk> im Auftrag von Gerd Petermann <gpetermann_muenchen@hotmail.com> Gesendet: Dienstag, 22. Oktober 2019 16:52 An: Development list for mkgmap Betreff: Re: [mkgmap-dev] type/subtype of points and cities
Hi Ticker,
thanks.
@Minko: I've created a small map with "Openfietsmap full" style and trunk r4315 , OSM data contains an amenity=bus_station. The result is that this POI appears in the map (a small dot with the label), but it is not in the mdr (like you say). I found no older versions of mkgmap where this would have worked different, so it seems useless to me because normally I'd want to be able to find the bus station under transport services.
Why do you use 0x1101, 0x1102, or 0x1108 for POI?
Gerd
________________________________________ Von: mkgmap-dev <mkgmap-dev-bounces@lists.mkgmap.org.uk> im Auftrag von Ticker Berkin <rwb-mkgmap@jagit.co.uk> Gesendet: Dienstag, 22. Oktober 2019 16:29 An: Development list for mkgmap Betreff: Re: [mkgmap-dev] type/subtype of points and cities
Hi Gerd
The old reader/osm/Gtype:checkType() didn't diagnose cities (however they might be defined) with non-zero subtypes - this is where the new message comes from.
Old general/MapPoint:isCity[Type] defined city as:
return type >= 0x0100 && type <= 0x1100
ie 0x1100 is city, 0x1101 isn't, which I thought wasn't a good definition, so I changed it to:
return type >= 0x0100 && type < 0x1200;
I chose all of 0x11XX to be considered cities so that 0x1100 processing as a city didn't change (this and 0x1000 were put into RGN indPoints, but were not indexed as cities by MDR)
Changing it to < 0x1100 would be reasonable, because the device range seems to be < 0x0e00 and MDR5 city indexing was for points < 0x1000
Ticker
On Tue, 2019-10-22 at 12:32 +0000, Gerd Petermann wrote:
Hi Ticker,
I see some new warnings for Minkos popular Openfietsmap styles with this patch: checking style: Openfietsmap full Warning: invalid type 0x1101 for POINT in style file points, line 97 Warning: invalid type 0x1102 for POINT in style file points, line 263 Warning: invalid type 0x1102 for POINT in style file points, line 264 Warning: invalid type 0x1108 for POINT in style file points, line 336 Warning: invalid type 0x1108 for POINT in style file points, line 348 Warning: invalid type 0x1108 for POINT in style file points, line 349 Warning: invalid type 0x1108 for POINT in style file points, line 350
The corresponding lines in the points file amenity=bus_station [0x1101 resolution 24 continue] railway=halt [0x1102 resolution 22] railway=station [0x1102 resolution 20-23 continue] (barrier=bollard | barrier=bus_trap | barrier=block) [0x1108 resolution 23] access:conditional=* & access:conditional ~ '!(sunset|sunrise|destination)' & bicycle!=no & bicycle!=permissive & vehicle!=no { name 'access=${access:conditional}' } [0x1108 resolution 24] bicycle:conditional=* & bicycle:conditional ~ '!(sunset|sunrise|destination)' { name 'bicycle=${bicycle:conditional}' } [0x1108 resolution 24] vehicle:conditional=* & vehicle:conditional ~ '!(sunset|sunrise|destination)' { name 'vehicle=${vehicle:conditional}' } [0x1108 resolution 24]
Can you explain what happened with those POI without the patch?
Gerd
________________________________________ Von: mkgmap-dev <mkgmap-dev-bounces@lists.mkgmap.org.uk> im Auftrag von Ticker Berkin <rwb-mkgmap@jagit.co.uk> Gesendet: Dienstag, 22. Oktober 2019 13:07 An: Development list for mkgmap Betreff: Re: [mkgmap-dev] type/subtype of points and cities
Hi Gerd
Here is the patch based on the current trunk
Ticker On Tue, 2019-04-23 at 11:51 +0100, Ticker Berkin wrote:
Hi
I think the mkgmap internal handling of types/subtypes of points is obscure.
In the points style file, the type code is always full, ie type << 8
subtype, but when the points are read back from the RGN file for the MDR processing, the representation is the same provided the subtype is not zero, otherwise it is just type! Logic then decides that if the value is <= 0xff it is because the subtype is zero.
It is simpler and much clearer to preserve the original representation. This also allows unambiguous use of type 0 with subtypes, and, possibly, city with subtypes.
It also allows the same function to be used to test a type for being in the 'city' range, regardless of the context. mkgmap wasn't consistent the ranges for the isCity test.
City POI are written in a different manner to the RGN file. The old range for this was: type >= 0x0100 && type <= 0x1100 which I've kept, except changing to: ...& type < 0x1200. A similar range was used for MDR5.
However, city POI are held in group 1 of MDR 9/10. This is used in Find
City 'By Name'. The old range was: type <= 0xf, where, if the subtype was zero, type is right shifted by 8 (see above), non-zero subtypes messed up the testing. Now it uses the same range as above.
Actually neither of these ranges are correct for my Garmin devices. Find > City show nearby POI with types in the range 0x0100..0x0d1f, regardless of the img RGN or MDR. The eTrex Legend labels POIs in this range as {Large/Medium/Small} City/Town and points in the range 0x0e00..0x111f as "*".
Find > City 'by-name' returns POI if in group 1 of MDR 9/10. I find it a useful feature to have a POI that can be searched for but doesn't flood the list of nearby points.
Generally, non-zero subtypes for cities cause problems in the RGN structure and I've added a test for this to the style validation.
There is also a fix to --make-poi-index, but I can't detect any effect of this option.
This patch won't make any difference to the img output unless the there are points 0x1000, 0x1100.
If this patch is accepted, I'll do the equivalent to the img display system.
Regards Ticker _______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev _______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev _______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev