如何使此过滤器可搜索或提高索引

Question

vzla_seg为路网段。每一段都有他的大致方位角。 COUNT(*) = 3.849.834

CREATE TABLE vzla_seg (`azimuth` int);

INSERT INTO vzla_seg  (`azimuth`)
VALUES
    (330),    (335),
    (340),    (345),
    (350),    (355),
    (359),    (3),
    (5),      (15),
    (20),     (25),
    (30),     (35)
;

我试图在与我的参数 carAzimuth 相同的大体方向（+/- 30 度）上找到线段。

Azimuth 罗盘度数的范围是 [0 ..359]。因此，如果 carAzimuth = 345 与方位角为 [315, 325, 359, 0, 5, 15]

的路段在同一方向上

carAzimuth = 355
    segmentAzimuth = 340 --> (355 - 340) -->      15  < 30
    segmentAzimuth = 359 --> (355 - 359) --> abs(- 4) < 30       
    segmentAzimuth =  20 --> (355 -  20) -->      25  < 30

carAzimuth = 5
    segmentAzimuth = 340 --> (5 - 340) --> abs(- 335) > 330       
    segmentAzimuth = 359 --> (5 - 359) --> abs(- 354) > 330       
    segmentAzimuth =  20 --> (5 -  20) --> abs(  -15) < 30

explain analyze
    SELECT *
    FROM map.vzla_seg S
    WHERE
        abs(carAzimuth - S.azimuth) < 30 
     OR abs(carAzimuth - S.azimuth) > 330

索引：

CREATE INDEX vzla_seg_azimuth_idx
  ON map.vzla_seg
  USING btree
  (azimuth);

解释计划：

"Seq Scan on vzla_seg s  (cost=0.00..151243.55 rows=2138790 width=84) 
                         (actual time=0.061..14086.038 rows=690334 loops=1)"
"  Filter: ((abs((345 - azimuth)) < 30) OR (abs((345 - azimuth)) > 330))"
"  Rows Removed by Filter: 3159500"
"Total runtime: 14283.524 ms"

加分题：

如果解释说 Rows Removed by Filter: 3.159.500 和 rows=690.334 匹配总数 COUNT(*) = 3.849.834
- 为什么费用说 rows=2.138.790？

Answer 1

您可以使用 BETWEEN 和/或大于 than/less：

    SELECT *
    FROM map.vzla_seg S
    WHERE
        (S.azimuth BETWEEN (carAzimuth - 30) AND (carAzimuth + 30))
     OR  S.azimuth  < (carAzimuth - 330) 
     OR  S.azimuth  > (carAzimuth + 330)

Answer 2

看看这是否优化了索引的使用：

SELECT *
FROM map.vzla_seg S
WHERE s.azimuth > 315 and s.azimuth < 375;

如果是这样，那么我认为逻辑是：

SELECT *
FROM map.vzla_seg S
WHERE (s.azimuth > 315 and s.azimuth < 375) or
      (s.azimuth > 15 and s.azimuth < 30) ;

如果不使用索引，请使用 union all 而不是 or:

SELECT *
FROM map.vzla_seg S
WHERE (s.azimuth > 315 and s.azimuth < 375)
UNION ALL
SELECT *
FROM map.vzla_seg S
WHERE (s.azimuth > 15 and s.azimuth < 30) ;

Answer 3

你可以试试这个：

with param(car_azimuth) as (select 345) 
    select *
    from vzla_seg s, param
    where 
        s.azimuth > car_azimuth- 30 and   
        s.azimuth < car_azimuth+ 30 or
        s.azimuth < car_azimuth- 330 or 
        s.azimuth > car_azimuth+ 330;

该计划可能取决于传递参数的方式。用常量解释（在 100 万行上）：

explain 
    select *
    from vzla_seg s
    where 
        s.azimuth > 345- 30 and   
        s.azimuth < 345+ 30 or
        s.azimuth < 345- 330 or 
        s.azimuth > 345+ 330;

                                           QUERY PLAN                                            
-------------------------------------------------------------------------------------------------
 Bitmap Heap Scan on vzla_seg s  (cost=3496.96..11233.44 rows=160405 width=8)
   Recheck Cond: (((azimuth > 315) AND (azimuth < 375)) OR (azimuth < 15) OR (azimuth > 675))
   ->  BitmapOr  (cost=3496.96..3496.96 rows=165574 width=0)
         ->  Bitmap Index Scan on vzla_seg_azimuth_idx  (cost=0.00..2598.70 rows=123828 width=0)
               Index Cond: ((azimuth > 315) AND (azimuth < 375))
         ->  Bitmap Index Scan on vzla_seg_azimuth_idx  (cost=0.00..773.52 rows=41746 width=0)
               Index Cond: (azimuth < 15)
         ->  Bitmap Index Scan on vzla_seg_azimuth_idx  (cost=0.00..4.43 rows=1 width=0)
               Index Cond: (azimuth > 675)
(9 rows)

如何使此过滤器可搜索或提高索引

How make this filter sargeable or improve index

sql

postgresql

indexing

sql-execution-plan

postgresql-performance