has decent performance. It is implemented as a lossy sparse array with a
priority queue tracking how much of each block is used to ensure we maximize
the number of nodes we fit in the given amount of memory.
Also rearrange some header definitions.
- list of tags read from file
- slim mode works again
- relations properly supported
- more efficient DB usage
It includes some restructuring of the code, in particular, the output module
manages the mid-level now and the main program doesn't call it at all. This
moves many of the previous hacks to the output module which can manage the
mid-level as appropriate for its output (i.e. slim mode requires different
semantics from ram mode)
instances out of the one DB. The default is ofcourse still "planet_osm" so
if you don't use it you won't see a difference. I did however need to change
the names of the indexes so they don't clash also, but this is unlikely to
break anything.
primarily the true spherical mercator used by Google, TileCache and others.
Add the -m option to select this. The default is still the incorrect
projection used before. Also display the used projection during processing.
Finally some minor cleanups to fix some warnings about undefined functions.
This allows the removal of the AVL & BST code and the ID field from the node/segment/way sctrucutres. This saves some memory and simplifies the code.
Fixed a memory leak in WKT which forgot to free the segment item.
Added counters for maximum node/segment/way IDs.
Split the assert(id) checks to make it obvious which is failing.
Cleaned up some white space.
Improved linestring generation for non-contiguous ways.
Added a GIST index into the SQL output.
Corrected usage info and added a gzip example.
Removed some redundant lines.
