July 2013 - mkgmap-dev - The mkgmap lists

misshapen roundabouts
by Kay F. Jahnke 26 Jul '13

26 Jul '13

I think I finally found what the problem is: the garmin devices can't handle the precise data I want them to display, look at these two quotes: 1. from http://www.cgpsmapper.com/download/GM8DocV2.pdf When declaring the location of objects in the world, the highest precision available uses 24 bits of resolution, which resolves to an accuracy of about 2.5 meters. 2. in 'Garmin IMG File Format' by John Mechalas: The reference coordinates in the TRE section are given as 24-bit integers (three bytes). This implies that the maximum granularity for map elements in the image file is approximately .0000214576721 degrees of latitude or longitude. At the equator, a degree of latitude is roughly 365228.16 feet, making the highest resolution for map points about 7.8369461 feet. So even if we have precise coordinates in OSM, in the garmin image they will be rounded to the nearest(?) grid point. In fact, everything on the garmin map looks like it's been snapped-to-grid, if you look at it closely. Sorry for the noise; maybe my posting is still helpful for others facing the same problem. Kay

3 2

Splitter Crash
by News 25 Jul '13

25 Jul '13

I'm getting the following error when trying to split some large .osm files containing contour data. After generating the contour files I preprocess them and then split them so it's most likely an error in my preprocesing but I can't track it down. Additionally this only happens if the input file needs splitting i.e if I tweak the contour file size and the --max-nodes option to the point that the input file doesn't actually need splitting then splitter continues through the file and all is well but this approach isn't feasible for what I need to achieve. Command line is:- java -Xmx1024M -jar splitter.jar --mapid=12345001 --max-nodes=200000 --output=xml input.osm Any pointers as to where to look would be appreciated. Thanks Paul Exception in thread "main" java.lang.IndexOutOfBoundsException: Index: 2147483647, Size: 6 at java.util.ArrayList.rangeCheck(ArrayList.java:604) at java.util.ArrayList.get(ArrayList.java:382) at uk.me.parabola.splitter.WriterDictionaryInt.getBitSet(WriterDictionaryInt.java:85) at uk.me.parabola.splitter.SplitProcessor.writeNode(SplitProcessor.java:297) at uk.me.parabola.splitter.SplitProcessor.processNode(SplitProcessor.java:102) at uk.me.parabola.splitter.OSMParser.endElement(OSMParser.java:219) at uk.me.parabola.splitter.AbstractXppParser.parse(AbstractXppParser.java:61) at uk.me.parabola.splitter.Main.processMap(Main.java:831) at uk.me.parabola.splitter.Main.writeAreas(Main.java:769) at uk.me.parabola.splitter.Main.split(Main.java:277) at uk.me.parabola.splitter.Main.start(Main.java:162) at uk.me.parabola.splitter.Main.main(Main.java:146)

4 11

Commit: r2654: Improve matching of street names for housenumber assignment
by svn commit 24 Jul '13

24 Jul '13

Version 2654 was committed by wanmil on Sun, 30 Jun 2013 Improve matching of street names for housenumber assignment The housenumber handling uses the style sheet generated name of streets and matches that to the addr:street tag of housenumbers. The style sheet name is now stripped by the shield information (usually a ref) and text within brackets (also usually the ref name). This increases the match rate of housenumbers especially for streets with a ref.

6 8

No transliteration with CP1250
by Andrzej Popowski 22 Jul '13

22 Jul '13

Hi, I'm creating map of Poland, for Polish I need CP1250. My map covers some area outside Polish border, where object names are written in Cyrillic. In this area I get names like '???????'. I would like to get possibility to use transliterate as a function in mkgmap style. For example I could use it like that: mkgmap:country=RUS & name=* { set name='${name:en}' | '${int_name}' | transliterate_to_ascii(${name}) } Or maybe there could be full transliteration support for CP1250? -- Best regards, Andrzej

1 0

Commit: r2658: Use mkgmap:street tag to get street name for house number search
by svn commit 21 Jul '13

21 Jul '13

Version 2658 was committed by wanmil on Wed, 17 Jul 2013 Use mkgmap:street tag to get street name for house number search The house number matching algorithm now checks the mkgmap:street tag on roads. If it is set it uses it to match to the addr:street tags of house number elements. This enables style developers to put refs and other infos into the street name and at the same time have a good matching for house number search.

4 6

Re: [mkgmap-dev] Commit: r2659: Make parameter housenumbers official
by "Maxim Düster" 20 Jul '13

20 Jul '13

Hello everyone! Can somebody explain me how the option --housenumbers works? It doesn't make any difference to me whether I use this option or not - the resulting gmapsupp.img has the same size. Should this function on the real device? On my nüvi I can enter any house number in the address search - the result is always one and the same point on the street and not the point nearest to the house I sought for. I tried it with the default style. The houses on the streets have addr:housenumber and addr:street set. The latter matches the street name. What should I take into account so that it works as expected? Do I misunderstand something? Max

1 0

Re: [mkgmap-dev] Commit: r2659: Make parameter housenumbers official
by "Maxim Düster" 20 Jul '13

20 Jul '13

1 0

Commit: r2659: Make parameter housenumbers official
by svn commit 17 Jul '13

17 Jul '13

Version 2659 was committed by wanmil on Wed, 17 Jul 2013 Make parameter housenumbers official Add documentation for the housenumbers parameter so that it can be used without --x- prefix.

1 0

splitter crashes on Windows x64
by Andrzej Popowski 14 Jul '13

14 Jul '13

Hi, I'm creating maps for Europe, which is a big piece of data. Splitting it is a main problem. I'm using Windows 7 x64 and recent splitter 306. My experiences are following: Splitter on java 32bit and standard settings can split Europe. If I add contours data to Europe file, splitter on java 32 fails because of not enough memory, i can use maximum -Xmx1600M. Any use of splitter on java 64bit most probably will crash java VM. Usually this happens after 1 hour of work, often when starting processing second set of areas. I have many logs form crashes, but don't know how useful it could be. Logs contains info like: # JRE version: 7.0_21-b11 # Java VM: Java HotSpot(TM) 64-Bit Server VM (23.21-b01 mixed mode windows-amd64 compressed oops) # Problematic frame: # J java.util.HashMap.put(Ljava/lang/Object;Ljava/lang/Object;)Ljava/lang/Object; # Problematic frame: # J java.util.HashMap.containsKey(Ljava/lang/Object;)Z # Problematic frame: # V [jvm.dll+0x360038] # Problematic frame: # j java.util.AbstractCollection.toArray([Ljava/lang/Object;)[Ljava/lang/Object;+88 # Problematic frame: # J uk.me.parabola.splitter.BinaryMapWriter$PBFSerializer$Prim.addStringsToStringtable()V Splitting contours created for Europe by pphyghtmap with default settings usually fails. I have changed max-nodes-per-way form default 2000 to 500 which allowed splitter to finish this task. I understand that it is java x64 which crashes. But since 32bit version is already too limited to process big data, wouldn't be good to investigate this problem? Maybe you could make spliter compatible with 64bit version. -- Best regards popej

4 7

mkgmap multiple input files
by Andrzej Popowski 14 Jul '13

14 Jul '13

Hi, I would be interested in creating single img tile form multiple input files. Since description of "input-file" parameter is quite imprecise, I have done some testing. I have used template.args in a following form: mapname: 08900000 input-file: 08900001.osm.pbf input-file: 08900002.osm.pbf mapname: 08900001 input-file: 08900003.osm.pbf input-file: 08900004.mp What I have observed: If pbf files cover different area, then mkgmap creates separate img for each pbf. Second mapname seems to be ignored. Mkgmap outputs files with sequential numbers. Polish format mp is read, since mkgmap issues warnings, but these data aren't included in any img. And most interesting: if pbf files cover exactly the same area, they are merged into single tile. This is what I'd like to get. Unfortunately merging is not recognized when creating tdb, index and preview map. For this files, each pbf is treated as separate tile. Resulting mapset is not accepted by Mapsource, it refuses to start. I guess, that merging pbf into single tile is not supported and it is only some incidence, that it works for pbf with the same area. I would like to suggest, to make this procedure official and correct algorithms for Mapsource files accordingly. Could you for example consider new option, like "add-file", which would cause merging input file with a previous one? There shouldn't be much work for pbf, since it nearly works. And it would be a great help if it worked for polish format mp, since it would allow to add data, which is not available in OSM format. -- Best regards Andrzej

3 3