103
yjhatfdu2 254 天前
@249239432 计算次数和算法是很有关的,比如你要圈每一个点附近 50 米的所有点,你可以两次 for 循环每两个之间算一次,也可以排序/索引之后,每个点只需要用 LogN 次查询
|
105
yjhatfdu2 253 天前
@249239432 我单机 pg 查了 1000w 个点,每个点附近 50 个点的数量(平均 5 个左右)也就 20 分钟
explain analyse select id,(select count(*) from geo where st_dwithin(point,g.point,50)) from geo g; QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------------------------ Seq Scan on geo g (cost=0.00..165365708334.00 rows=10000000 width=12) (actual time=1.076..1049434.731 rows=10000000 loops=1) SubPlan 1 -> Aggregate (cost=16536.54..16536.55 rows=1 width=8) (actual time=0.105..0.105 rows=1 loops=10000000) -> Index Scan using geo_point_idx on geo (cost=0.54..16534.04 rows=1000 width=0) (actual time=0.027..0.104 rows=14 loops=10000000) Index Cond: (point && _st_expand(g.point, '50'::double precision)) Filter: st_dwithin(point, g.point, '50'::double precision, true) Rows Removed by Filter: 7 Planning Time: 0.756 ms Execution Time: 1049626.246 ms (9 rows) Time: 1049627.962 ms (17:29.628) |