layer dropping

[mapsam]

What changed?

• Add layers option to the tile objects that allows user to specify which layers to keep in the final vector tile
• the layers array must have at least one string, cannot be empty
• if the layers array is not included, all layers are kept
• add new validation and success tests
• add layer dropping fixture to benchmarks

Housekeeping

• Update README examples
• Remove documentation auto generated docs
• remove /docs which were copied from node-cpp-skel
• remove a lot of README copied from node-cpp-skel
• remove API.md which was copied from node-cpp-skel

cc @mapbox/tilesets-api @artemp

[mapsam]

benchmarks round 1

Benchmarks look good. Mostly want to see no loss in performance with this addition, which these show. Looks like a slight increase, which is hard to believe, so I'll consider that mostly noise for now. No significant performance changes!

test	0.4.0	215e44e
1: single tile in/out	22422 runs/s (446ms)	24038 runs/s (416ms)
2: two different tiles at the same zoom level, zero buffer	5173 runs/s (1933ms)	5206 runs/s (1921ms)
3: two different tiles from different zoom levels (separated by one zoom), zero buffer	757 runs/s (13206ms)	876 runs/s (11419ms)
4: two different tiles from different zoom levels (separated by more than one zoom), zero buffer	1870 runs/s (5349ms)	1854 runs/s (5393ms)
5: tiles completely made of points, overzooming, no properties	10549 runs/s (948ms)	9242 runs/s (1082ms)
6: tiles completely made of points, same zoom, no properties	34602 runs/s (289ms)	32468 runs/s (308ms)
7: tiles completely made of points, overzoooming, lots of properties	7758 runs/s (1289ms)	7052 runs/s (1418ms)
8: tiles completely made of points, same zoom, lots of properties	27174 runs/s (368ms)	28653 runs/s (349ms)
9: buffer_size 128 - tiles completely made of points, same zoom, lots of properties	23095 runs/s (433ms)	23256 runs/s (430ms)
10: tiles completely made of linestrings, overzooming and lots of properties	1818 runs/s (5500ms)	1796 runs/s (5567ms)
11: tiles completely made of polygons, overzooming and lots of properties	362 runs/s (27616ms)	359 runs/s (27853ms)
12: tiles completely made of points and linestrings, overzooming and lots of properties	16103 runs/s (621ms)	13966 runs/s (716ms)
13: buffer_size 128 - tiles completely made of points and linestrings, overzooming and lots of properties	15823 runs/s (632ms)	12531 runs/s (798ms)
14: tiles completely made of points and linestrings, overzooming (2x) and lots of properties	21186 runs/s (472ms)	20161 runs/s (496ms)
15: tiles completely made of polygons, overzooming and lots of properties	1373 runs/s (7281ms)	1276 runs/s (7835ms)
16: tiles completely made of polygons, overzooming (2x) and lots of properties	2966 runs/s (3372ms)	3074 runs/s (3253ms)
17: buffer_size 4096 - tiles completely made of polygons, overzooming (2x) and lots of properties	1416 runs/s (7061ms)	1391 runs/s (7190ms)
18: single tile, dropping layers ...	n/a	21413 runs/s (467ms)
peak mem (max_rss, max_heap, max_heap_total)	1.77GB 21.77MB 52.23MB	1.5GB 23.02MB 51.73MB

[e-n-f]

Looks good to me in general, aside from a few questions inline

[e-n-f]

Why does this move the layers0 argument instead of using a const reference?

[mapsam]

Good question! I'm open to ideas here - the clang tidy/format builds told me to use std::move here, so I'm admittedly just plugging that in. Can you show me how I'd use a const reference?

[e-n-f]

What I was talking about was 05cbd11bd093d, but you are right and I am wrong, and the TileObject needs to contain its own copy of the vector instead of holding a reference back to the caller's vector, which apparently goes out of scope before it is used. So please ignore this comment.

[mapsam]

👍 thanks for the commit + trying it out @ericfischer!