Eric Fischer
08ff40e42f
Trying to follow Vlad's suggestion and work from the top down.
...
Doesn't quite work yet, but seems like the right thing to do.
From the bottom up, indexing is a mess because of the buffering,
and includes unneeded areas in the bounding boxes of big features.
2014-12-03 16:18:43 -08:00
Eric Fischer
a1e7426956
Sprinkle consts and casts to make it valid C++ as well as C
2014-12-02 14:17:49 -08:00
Eric Fischer
bb13dc03e5
Gzip -9 tiles. Don't bother writing tiles whose features all got clipped away.
2014-11-20 14:33:45 -08:00
Eric Fischer
896f05aea6
Use gzip tile compression to match current tile server
2014-11-14 14:21:12 -08:00
Eric Fischer
ffe4c95376
Calculate point dropping during encoding instead of tiling
...
so it is consistent across tile boundaries, and so every point
that exists at zoom x always also exists at zoom x+1.
2014-11-14 12:42:50 -08:00
Eric Fischer
e7aa05e7e4
Don't immediately fail when a very low tile resolution is specified manually
2014-11-12 17:06:58 -08:00
Eric Fischer
5595a087c7
Quick reject by bounding box of linear features too small to draw
2014-11-12 15:57:45 -08:00
Eric Fischer
575a24b095
Use a set instead of a map to reduce needless memory overhead
2014-11-06 11:29:11 -08:00
Eric Fischer
ea292def47
Make duplicate trimming less bad: only trim features indexed >1 time
2014-11-06 11:27:10 -08:00
Eric Fischer
c24cfada61
Move feature type to index so point skipping can avoid looking up geometry
...
Which should avoid having to page the geometry in if there is
more data than will fit in memory
2014-11-05 12:18:03 -08:00
Eric Fischer
3f9f50258c
Just use a standard C++ type for deduplicating instead of reinventing the wheel
2014-10-31 16:50:28 -07:00
Eric Fischer
3cd6fcfbe0
Add option to set buffer size. Pass it through to clipping functions.
2014-10-27 13:20:17 -07:00
Eric Fischer
4cb56a6b11
The indexing side of buffering features at the edges of tiles
2014-10-27 12:56:51 -07:00
Eric Fischer
e0826a12c6
Much faster at low zooms to check for duplicates before simplifying
2014-10-27 10:00:46 -07:00
Eric Fischer
9bb4a549ca
Silence warnings about unused variables
2014-10-24 17:24:04 -07:00
Eric Fischer
46100c34df
Move projection and tile arithmetic to its own file
2014-10-24 17:22:14 -07:00
Eric Fischer
17384d5da7
Move geometric transformations into their own file
2014-10-24 15:12:02 -07:00
Eric Fischer
ac00c84fb2
Save a little more space by sub-ordering within the same index
...
to put the lowest-numbered endpoint last
within lines that start at the same point,
since that is the one that might come next.
2014-10-21 17:59:42 -07:00
Eric Fischer
1cae0e85b3
Reindex features by their first coordinates so they coalesce better
2014-10-21 16:16:09 -07:00
Eric Fischer
9dab4b9fb6
Reverse lines that point the "wrong" direction so they coalesce better
2014-10-21 16:09:51 -07:00
Eric Fischer
043b35eede
Take another opportunity to simplify lines after coalescing features
2014-10-18 11:40:09 -07:00
Eric Fischer
151f0c116c
Fix clipping of polygons with multiple rings.
2014-10-16 15:17:18 -07:00
Eric Fischer
3bdcc3ca90
Heavy-handedly fix clipping errors at the edges of z0 and z1 tiles
2014-10-15 17:03:30 -07:00
Eric Fischer
8fec4ef9fc
Return immediately instead of failing to clip polygons to the z0 tile
2014-10-15 16:34:49 -07:00
Eric Fischer
43b27e2dfd
Don't closepath twice in multipolygons
2014-10-15 16:30:33 -07:00
Eric Fischer
ac34a292e9
Add an flag to remove all properties and use only geometry
2014-10-09 15:11:43 -07:00
Eric Fischer
dea4335162
Less accurate but much faster estimate of size savings from excluding metadata
2014-10-09 14:02:20 -07:00
Eric Fischer
389c89bdf9
Put the suggestions in order
2014-10-08 16:59:00 -07:00
Eric Fischer
67b8b984e5
Suggest properties you could drop to make the tiles smaller
2014-10-08 16:39:44 -07:00
Eric Fischer
e1c5bfe3e1
Factor out aspects of tile creation
2014-10-08 14:01:47 -07:00
Eric Fischer
376bab74a5
Larger minimum polygon size looks better than dropping to larger pixels
2014-10-07 16:19:18 -07:00
Eric Fischer
49b36e09cc
Tiny polygons (smaller than a pixel) decay into probabilities.
2014-10-07 15:38:06 -07:00
Eric Fischer
adbd9f6edb
Add a flag to select the rate at which points are dropped at low zooms.
2014-10-07 13:54:13 -07:00
Eric Fischer
8594dbe56b
If a tile is too big, try backing off to lower resolutions before failing
2014-10-07 10:27:17 -07:00
Eric Fischer
3433d17a96
Clip polygons to just outside the tile boundary
2014-10-03 16:33:22 -07:00
Eric Fischer
722e5f16eb
Dashing: better than shortening, but still too visible
2014-10-01 14:01:58 -07:00
Eric Fischer
6922a57aea
Still looks terrible, but make segment shortening work again
2014-10-01 11:09:35 -07:00
Eric Fischer
7df269219c
Fix memory leak of field names
2014-10-01 10:20:04 -07:00
Eric Fischer
6a64cf1b04
Convert metadata to use a vector too.
2014-09-30 17:24:34 -07:00
Eric Fischer
6bbe023187
Return to sorting features before coalescing
2014-09-30 17:18:23 -07:00
Eric Fischer
3ad8a919d9
Crash if the geometry is corrupted
2014-09-30 16:37:09 -07:00
Eric Fischer
c32b0207f2
Use a constructor, and fix an encoding bug for lines clipped away entirely
2014-09-30 16:27:00 -07:00
Eric Fischer
1b3620dabc
Start converting C arrays to C++ vectors
2014-09-30 15:53:45 -07:00
Eric Fischer
04c8a4441d
Remove no-ops after coalescing, which will often generate empty moves.
2014-09-29 17:32:01 -07:00
Eric Fischer
e09bec0824
Don't coalesce point features, because multipoints don't seem to draw right.
2014-09-29 16:18:20 -07:00
Eric Fischer
099623cd32
Arrays for coalescing can get too big for the stack.
2014-09-29 15:59:36 -07:00
Eric Fischer
98b0cb5e3f
Tiles are smaller if coalesced features are sorted geographically
2014-09-29 15:38:33 -07:00
Eric Fischer
a08e57357a
Coalesce geometries of features with the same attributes
2014-09-29 15:33:14 -07:00
Eric Fischer
c02c91df5a
Make an array of features in each tile to prepare to coalesce features
2014-09-29 14:46:45 -07:00
Eric Fischer
26e9feda02
Move mbtiles code to its own source file
2014-09-29 12:48:58 -07:00
