PostgreSQL 中是否有相当于 SQLite R*Tree 的工具？

Question

我正在将 SQLite 数据库结构迁移到 PostgreSQL。该 SQLite 数据库使用带有空间索引的 R*Tree 模块来进行快速 3D 边界框搜索。

我没有空间索引和 PostgreSQL 的经验。

我已阅读有关 PostgreSQL GiST 索引的文档，这似乎表明这可以实现类似的功能。但我一直无法找到如何将此功能应用于 3D 边界框检查。

我想不知何故 <@ operator with box and point should do the trick. Can someone elaborate if this intuition is right and provide a small example of how to define a table with spatial 3D data and a query that does a bounding box check?

这是示例数据和我正在寻找的查询思路：

id	x	y	z
1	0	1	2
2	4	5	6

SELECT id FROM table WHERE x >= 0 AND x <= 2 AND y >= 0 AND y <= 2 AND z >= 0 AND z <= 2

Answer 1

使用 PostGIS，您可以将索引设置为任何其他要点索引 - 默认运算符类根据类型应用。 db<>fiddle 的演示：

设置带有样品的桌子：

create extension postgis;

create table business(
  id int generated by default as identity primary key,
  business_name text,
  x numeric,
  y numeric,
  z numeric,
  t numeric,
  geom geometry(PointZM,4326) generated always as 
      (st_setsrid(st_makepoint(x,y,z,t),4326)) stored);

select setseed(.4);
insert into business (business_name,x,y,z,t)
select 'business_'||n,
        x,
        y,
        z,
        t
from generate_series(1,1e5,1)a(n),
  lateral(select n,   random()*180 x,
                   90-random()*180 y, 
                   50-random()*100 z,
                   50-random()*100 t)b(n1,x,y,z,t);

insert into business(business_name,x,y,z,t) values
  ('business_d0',0,0,0,0),
  ('business_d1',1,0,0,0),
  ('business_d2',0,1,0,0),
  ('business_d3',0,0,1,0),
  ('business_d4',0,0,0,1);

设置索引。默认运算符类将仅使用每个几何图形的两个维度来构建 r 树，因此您需要指定

gist_geometry_ops_nd

运算符类来处理 n 维几何图形的所有维度：

drop index if exists geom_idx;
create index geom_idx on business using gist(geom gist_geometry_ops_nd);

测试操作，显示索引使用：选择距离参考点最近的 5 个点：

prepare postgis_based as
select id,business_name,x,y,z,t
      ,(select geom from business where id=42)<<->>geom distance 
from business
order by distance
limit 7;

explain analyze execute postgis_based;
execute postgis_based;

QUERY PLAN
Limit  (cost=8.59..9.53 rows=7 width=73) (actual time=0.376..1.411 rows=7 loops=1)
  InitPlan 1 (returns $0)
    ->  Index Scan using business_pkey on business business_1  (cost=0.29..8.31 rows=1 width=48) (actual time=0.007..0.008 rows=1 loops=1)
          Index Cond: (id = 42)
  ->  Index Scan using geom_idx on business  (cost=0.28..13420.38 rows=100005 width=73) (actual time=0.375..1.408 rows=7 loops=1)
        Order By: (geom <<->> $0)
Planning Time: 0.357 ms
Execution Time: 1.477 ms

id	企业名称	x	y	z	t	距离
42	业务_42	135.913060823458	7.7997970642875	2.93153014852416	13.8754985934873	0
10212	业务_10212	138.015417820462	6.16780328827525	6.99214194619289	17.9070448007494	6.310724334056296
39781	业务_39781	135.825537405277	10.1567444370512	1.08794511732091	19.9058199716022	6.732491586131694
78354	业务_78354	137.145666026389	12.8802423990292	5.95771554931872	17.565112372697	7.078226353070906
74004	业务_74004	131.450844772993	7.42647416532242	-1.66828745462648	17.208767597593	7.233238991206186
43296	业务_43296	139.082823062846	8.30873695673203	3.44753032284635	20.5087372769281	7.387321847582334
95119	业务_95119	130.772860227918	4.31568006546163	2.77471451324465	18.0819978648418	7.5019971132540055