Introduce SP-GiST operator class over box.

Patch implements quad-tree over boxes, naive approach of 2D quad tree will not work for any non-point objects because splitting space on node is not efficient. The idea of pathc is treating 2D boxes as 4D points, so, object will not overlap (in 4D space). The performance tests reveal that this technique especially beneficial with too much overlapping objects, so called "spaghetti data". Author: Alexander Lebedev with editorization by Emre Hasegeli and me
2025-12-02 23:42:46 +03:00 · 2016-03-30 18:42:36 +03:00
parent 87545f5412
commit acdf2a8b37
12 changed files with 1066 additions and 3 deletions
--- a/src/test/regress/expected/box.out
+++ b/src/test/regress/expected/box.out
@@ -215,3 +215,243 @@ SELECT '' AS four, height(f1), width(f1) FROM BOX_TBL;
      |      0 |     0
 (4 rows)

+--
+-- Test the SP-GiST index
+--
+CREATE TEMPORARY TABLE box_temp (f1 box);
+INSERT INTO box_temp
+	SELECT box(point(i, i), point(i * 2, i * 2))
+	FROM generate_series(1, 50) AS i;
+CREATE INDEX box_spgist ON box_temp USING spgist (f1);
+INSERT INTO box_temp
+	VALUES (NULL),
+		   ('(-0,0)(0,100)'),
+		   ('(-3,4.3333333333)(40,1)'),
+		   ('(0,100)(0,infinity)'),
+		   ('(-infinity,0)(0,infinity)'),
+		   ('(-infinity,-infinity)(infinity,infinity)');
+SET enable_seqscan = false;
+SELECT * FROM box_temp WHERE f1 << '(10,20),(30,40)';
+        f1        
+------------------
+ (2,2),(1,1)
+ (4,4),(2,2)
+ (6,6),(3,3)
+ (8,8),(4,4)
+ (-0,100),(0,0)
+ (0,inf),(0,100)
+ (0,inf),(-inf,0)
+(7 rows)
+
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 << '(10,20),(30,40)';
+                  QUERY PLAN                  
+----------------------------------------------
+ Index Only Scan using box_spgist on box_temp
+   Index Cond: (f1 << '(30,40),(10,20)'::box)
+(2 rows)
+
+SELECT * FROM box_temp WHERE f1 &< '(10,4.333334),(5,100)';
+        f1        
+------------------
+ (2,2),(1,1)
+ (4,4),(2,2)
+ (6,6),(3,3)
+ (8,8),(4,4)
+ (10,10),(5,5)
+ (-0,100),(0,0)
+ (0,inf),(0,100)
+ (0,inf),(-inf,0)
+(8 rows)
+
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 &< '(10,4.333334),(5,100)';
+                     QUERY PLAN                     
+----------------------------------------------------
+ Index Only Scan using box_spgist on box_temp
+   Index Cond: (f1 &< '(10,100),(5,4.333334)'::box)
+(2 rows)
+
+SELECT * FROM box_temp WHERE f1 && '(15,20),(25,30)';
+          f1           
+-----------------------
+ (20,20),(10,10)
+ (22,22),(11,11)
+ (24,24),(12,12)
+ (26,26),(13,13)
+ (28,28),(14,14)
+ (30,30),(15,15)
+ (32,32),(16,16)
+ (34,34),(17,17)
+ (36,36),(18,18)
+ (38,38),(19,19)
+ (40,40),(20,20)
+ (42,42),(21,21)
+ (44,44),(22,22)
+ (46,46),(23,23)
+ (48,48),(24,24)
+ (50,50),(25,25)
+ (inf,inf),(-inf,-inf)
+(17 rows)
+
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 && '(15,20),(25,30)';
+                  QUERY PLAN                  
+----------------------------------------------
+ Index Only Scan using box_spgist on box_temp
+   Index Cond: (f1 && '(25,30),(15,20)'::box)
+(2 rows)
+
+SELECT * FROM box_temp WHERE f1 &> '(40,30),(45,50)';
+        f1         
+-------------------
+ (80,80),(40,40)
+ (82,82),(41,41)
+ (84,84),(42,42)
+ (86,86),(43,43)
+ (88,88),(44,44)
+ (90,90),(45,45)
+ (92,92),(46,46)
+ (94,94),(47,47)
+ (96,96),(48,48)
+ (98,98),(49,49)
+ (100,100),(50,50)
+(11 rows)
+
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 &> '(40,30),(45,50)';
+                  QUERY PLAN                  
+----------------------------------------------
+ Index Only Scan using box_spgist on box_temp
+   Index Cond: (f1 &> '(45,50),(40,30)'::box)
+(2 rows)
+
+SELECT * FROM box_temp WHERE f1 >> '(30,40),(40,30)';
+        f1         
+-------------------
+ (82,82),(41,41)
+ (84,84),(42,42)
+ (86,86),(43,43)
+ (88,88),(44,44)
+ (90,90),(45,45)
+ (92,92),(46,46)
+ (94,94),(47,47)
+ (96,96),(48,48)
+ (98,98),(49,49)
+ (100,100),(50,50)
+(10 rows)
+
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 >> '(30,40),(40,30)';
+                  QUERY PLAN                  
+----------------------------------------------
+ Index Only Scan using box_spgist on box_temp
+   Index Cond: (f1 >> '(40,40),(30,30)'::box)
+(2 rows)
+
+SELECT * FROM box_temp WHERE f1 <<| '(10,4.33334),(5,100)';
+            f1            
+--------------------------
+ (2,2),(1,1)
+ (4,4),(2,2)
+ (40,4.3333333333),(-3,1)
+(3 rows)
+
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 <<| '(10,4.33334),(5,100)';
+                     QUERY PLAN                     
+----------------------------------------------------
+ Index Only Scan using box_spgist on box_temp
+   Index Cond: (f1 <<| '(10,100),(5,4.33334)'::box)
+(2 rows)
+
+SELECT * FROM box_temp WHERE f1 &<| '(10,4.3333334),(5,1)';
+            f1            
+--------------------------
+ (2,2),(1,1)
+ (4,4),(2,2)
+ (40,4.3333333333),(-3,1)
+(3 rows)
+
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 &<| '(10,4.3333334),(5,1)';
+                     QUERY PLAN                     
+----------------------------------------------------
+ Index Only Scan using box_spgist on box_temp
+   Index Cond: (f1 &<| '(10,4.3333334),(5,1)'::box)
+(2 rows)
+
+SELECT * FROM box_temp WHERE f1 |&> '(49.99,49.99),(49.99,49.99)';
+        f1         
+-------------------
+ (100,100),(50,50)
+ (0,inf),(0,100)
+(2 rows)
+
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 |&> '(49.99,49.99),(49.99,49.99)';
+                        QUERY PLAN                         
+-----------------------------------------------------------
+ Index Only Scan using box_spgist on box_temp
+   Index Cond: (f1 |&> '(49.99,49.99),(49.99,49.99)'::box)
+(2 rows)
+
+SELECT * FROM box_temp WHERE f1 |>> '(37,38),(39,40)';
+        f1         
+-------------------
+ (82,82),(41,41)
+ (84,84),(42,42)
+ (86,86),(43,43)
+ (88,88),(44,44)
+ (90,90),(45,45)
+ (92,92),(46,46)
+ (94,94),(47,47)
+ (96,96),(48,48)
+ (98,98),(49,49)
+ (100,100),(50,50)
+ (0,inf),(0,100)
+(11 rows)
+
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 |>> '(37,38),(39,40)';
+                  QUERY PLAN                   
+-----------------------------------------------
+ Index Only Scan using box_spgist on box_temp
+   Index Cond: (f1 |>> '(39,40),(37,38)'::box)
+(2 rows)
+
+SELECT * FROM box_temp WHERE f1 @> '(10,11),(15,16)';
+          f1           
+-----------------------
+ (16,16),(8,8)
+ (18,18),(9,9)
+ (20,20),(10,10)
+ (inf,inf),(-inf,-inf)
+(4 rows)
+
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 @> '(10,11),(15,15)';
+                  QUERY PLAN                  
+----------------------------------------------
+ Index Only Scan using box_spgist on box_temp
+   Index Cond: (f1 @> '(15,15),(10,11)'::box)
+(2 rows)
+
+SELECT * FROM box_temp WHERE f1 <@ '(10,15),(30,35)';
+       f1        
+-----------------
+ (30,30),(15,15)
+(1 row)
+
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 <@ '(10,15),(30,35)';
+                  QUERY PLAN                  
+----------------------------------------------
+ Index Only Scan using box_spgist on box_temp
+   Index Cond: (f1 <@ '(30,35),(10,15)'::box)
+(2 rows)
+
+SELECT * FROM box_temp WHERE f1 ~= '(20,20),(40,40)';
+       f1        
+-----------------
+ (40,40),(20,20)
+(1 row)
+
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 ~= '(20,20),(40,40)';
+                  QUERY PLAN                  
+----------------------------------------------
+ Index Only Scan using box_spgist on box_temp
+   Index Cond: (f1 ~= '(40,40),(20,20)'::box)
+(2 rows)
+
+RESET enable_seqscan;
+DROP INDEX box_spgist;
--- a/src/test/regress/expected/opr_sanity.out
+++ b/src/test/regress/expected/opr_sanity.out
@@ -1728,15 +1728,19 @@ ORDER BY 1, 2, 3;
       4000 |            6 | ~=
       4000 |            7 | @>
       4000 |            8 | <@
+       4000 |            9 | &<|
+       4000 |           10 | <<|
       4000 |           10 | <^
       4000 |           11 | <
       4000 |           11 | >^
+       4000 |           11 | |>>
       4000 |           12 | <=
+       4000 |           12 | |&>
       4000 |           14 | >=
       4000 |           15 | >
       4000 |           16 | @>
       4000 |           18 | =
-(108 rows)
+(112 rows)

 -- Check that all opclass search operators have selectivity estimators.
 -- This is not absolutely required, but it seems a reasonable thing
--- a/src/test/regress/sql/box.sql
+++ b/src/test/regress/sql/box.sql
@@ -117,3 +117,65 @@ SELECT '' AS one, b1.*, b2.*
   WHERE b1.f1 @> b2.f1 and not b1.f1 ~= b2.f1;

 SELECT '' AS four, height(f1), width(f1) FROM BOX_TBL;
+
+--
+-- Test the SP-GiST index
+--
+
+CREATE TEMPORARY TABLE box_temp (f1 box);
+
+INSERT INTO box_temp
+	SELECT box(point(i, i), point(i * 2, i * 2))
+	FROM generate_series(1, 50) AS i;
+
+CREATE INDEX box_spgist ON box_temp USING spgist (f1);
+
+INSERT INTO box_temp
+	VALUES (NULL),
+		   ('(-0,0)(0,100)'),
+		   ('(-3,4.3333333333)(40,1)'),
+		   ('(0,100)(0,infinity)'),
+		   ('(-infinity,0)(0,infinity)'),
+		   ('(-infinity,-infinity)(infinity,infinity)');
+
+SET enable_seqscan = false;
+
+SELECT * FROM box_temp WHERE f1 << '(10,20),(30,40)';
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 << '(10,20),(30,40)';
+
+SELECT * FROM box_temp WHERE f1 &< '(10,4.333334),(5,100)';
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 &< '(10,4.333334),(5,100)';
+
+SELECT * FROM box_temp WHERE f1 && '(15,20),(25,30)';
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 && '(15,20),(25,30)';
+
+SELECT * FROM box_temp WHERE f1 &> '(40,30),(45,50)';
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 &> '(40,30),(45,50)';
+
+SELECT * FROM box_temp WHERE f1 >> '(30,40),(40,30)';
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 >> '(30,40),(40,30)';
+
+SELECT * FROM box_temp WHERE f1 <<| '(10,4.33334),(5,100)';
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 <<| '(10,4.33334),(5,100)';
+
+SELECT * FROM box_temp WHERE f1 &<| '(10,4.3333334),(5,1)';
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 &<| '(10,4.3333334),(5,1)';
+
+SELECT * FROM box_temp WHERE f1 |&> '(49.99,49.99),(49.99,49.99)';
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 |&> '(49.99,49.99),(49.99,49.99)';
+
+SELECT * FROM box_temp WHERE f1 |>> '(37,38),(39,40)';
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 |>> '(37,38),(39,40)';
+
+SELECT * FROM box_temp WHERE f1 @> '(10,11),(15,16)';
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 @> '(10,11),(15,15)';
+
+SELECT * FROM box_temp WHERE f1 <@ '(10,15),(30,35)';
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 <@ '(10,15),(30,35)';
+
+SELECT * FROM box_temp WHERE f1 ~= '(20,20),(40,40)';
+EXPLAIN (COSTS OFF) SELECT * FROM box_temp WHERE f1 ~= '(20,20),(40,40)';
+
+RESET enable_seqscan;
+
+DROP INDEX box_spgist;