[mkgmap-dev] Question regarding alignment of tiles in splitter
From Steve Ratcliffe steve at parabola.me.uk on Fri Nov 23 19:49:42 GMT 2012
On 23/11/12 17:18, GerdP wrote: > Hi all, > > I am working on a better split algorithm. Now I wonder why the split > algorithm in r202 makes sure that tiles are aligned in a special way. A > comment in main says: > // The maximum resolution of the map to be produced by mkgmap. This is a > value in the range > // 0-24. Higher numbers mean higher detail. The resolution determines how > the tiles must > // be aligned. Eg a resolution of 13 means the tiles need to have their > edges aligned to > // multiples of 2 ^ (24 - 13) = 2048 map units, and their widths and > heights must be a multiple > // of 2 * 2 ^ (24 - 13) = 4096 units. The tile widths and height multiples > are double the tile > // alignment *because the center point of the tile is stored, *and that > must be aligned the > // same as the tile edges are. > private int resolution; > I think this is no longer true. Would it be okay to split tiles to single > alignment (e.g. 2048 for resolution 13) > or does something in mkgmap rely on these doubled values? There are actually no required alignment requirements at all in the img format. A tile can have arbitary boundaries, it does not even have to be rectangular; it can follow a country border for example. So it is all about how mkgmap deals with it. I think that the statement "because the center point of the tile is stored," refers to the subdivisions within a tile, so it is still true and always will be. My maths may be wrong though ;) If you don't have a boundary that is aligned then at low zooms you have points that should be on the boundary that fall between two points on the low resolution grid. In this case the questions are: 1. Should you choose a point inside the boundary? 2. The point inside one tile and the point outside on the next tile (so that it is the same point) 3. Does it matter? You might get gaps at low zooms. I can't remember if that happened. Routing might be affected (though I doubt it). The tdbfile may need fixing. The areas in the tdbfile are always rectangular. They enclose the whole tile and can overlap. You should just try it. It is quite possible that none of this matters very and it just works with boundaries at arbitary alignments already. If not it can be made to work because any limitation is in mkgmap and not the format as far as I understand. ..Steve
- Previous message: [mkgmap-dev] Question regarding alignment of tiles in splitter
- Next message: [mkgmap-dev] Question regarding alignment of tiles in splitter
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the mkgmap-dev mailing list