[mkgmap-dev] Splitter feature request

Wed Dec 6 11:23:09 GMT 2023

Hi Carlos,

since nobody else reacted:
I wonder in what situation an L-shaped tile could be the result of a split-process.
Do you think about some kind of post-processor which would recombine 2 or more tiles so that they still don't
reach the maxnodes boundary? Thinking about the way how a split is done, I think it can happen that the combination of two connected tiles are still below the limit,
but I would assume that an L-shape is rather rare.
Anyway, if splitter would write non-rectangular tiles we need
a) a further output file (or a new output format) to store the tile-boundary for input in mkgmap
b) some new logic in mkgmap to cut ways and polygons along that boundary

It might be easier to implement b) when boundary lines are only horizontal or vertical, but I am pretty sure this wouldn't solve most of the problems
reg.  right/left hand side driving.

ciao
Gerd

________________________________________
Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von Carlos Dávila <carlos at alternativaslibres.org>
Gesendet: Sonntag, 3. Dezember 2023 12:26
An: Development list for mkgmap
Betreff: [mkgmap-dev] Splitter feature request

Hi all

I usually let splitter decide tiles based on max-nodes value, but for
some maps I need to supply splitter an areas.list file which I adjust
manually. Reasons for that include reducing the total number of tiles,
avoiding cutting of some islands and separating right/left hand side
driving areas or reducing number of tiles with both sides driving. For
these cases, it would be helpful to be able to build "L shaped tiles",
as in the attached image. I think these tiles could be defined by three
coordinates (A,B,C), which would define the two rectangles in which the
L shaped tile can be split, and a switch (1,2,3,4) to determine de
orientation of the "L".

1: tile = A-lat, A-lon to B-lat, B-lon + A-lat,B-lon to C-lat,C-lon

2: tile = B-lat, A-lon to C-lat, C-lon + A-lat,B-lon to B-lat,C-lon

3: tile = A-lat, A-lon to B-lat, B-lon + B-lat,A-lon to C-lat,C-lon

4: tile = A-lat, A-lon to C-lat, B-lon + A-lat,B-lon to B-lat,C-lon

This raises several questions. First of all, if is feasible, then if it
is worth the effort to implement it, how would if affect splitter
performance and if it makes sense for other users. Perhaps it could be a
first step towards irregular tiles...