How to SELECT "closest" rows from another table?

Question

I have two tables, map1 and map2, there are multiple possible combinations between the columns map1.id1 and map2.id2.

I have tried the below query:

SELECT map1.id1, map2.id2, MIN(ST_HausdorffDistance(map1.g1, map2.g2)),map2.g2
FROM map1, map2
WHERE ST_HausdorffDistance(map1.g1, map2.g2) < 2
GROUP BY map1.id1,map2.g2,map2.id2, ST_HausdorffDistance(map1.g1, map2.g2)
ORDER BY map1.id1,map2.id2, ST_HausdorffDistance(map1.g1, map2.g2)

Current output

Multiple rows for id1:

id1   id2   min               g2
----  ----  ----------------  ----------------------------------------------------------------------------------------------------------------------------
6116   338  1.8122154049353   "0102000020E610000002000000590E3EDEF5AB23409255B6B4BF1B4A4031197DBBDBAB23404AF663EEB51B4A40"
6116   645  1.82162999509807  "0102000020E61000000300000057900B73277D2340A1675831011E4A4094C55801197D23406A204C40021E4A40B0CFF7AE9C7C234079AAE8B4131E4A40"
6116   674  1.82397666934862  "0102000020E610000002000000AC0E6F8C53B723405B80118F1F1E4A404EFF48C78BB723406C9159620A1E4A40"
  65   695  1.22999509807     "01456020E61000000300000057900B73277D2340A1675831011E4A4094C55801197D23406A204C40021E4A40B0CFF7AE9C7C234079AAE8B4131E4A40"
  65   689  1.556666934862    "0202000020E610000002000000AC0E6F8C53B723405B80118F1F1E4A404EFF48C78BB723406C9159620A1E4A40"
--  many more ...

Desired output

I wanted to SELECT only 1st 1 or 2 rows for for each id1 - as defined by the minimum Hausdorff distance:

id1   id2   min               g2
----  ----  ----------------  --------------------------------------------------------------------------------------------------------------------------
6116   338  1.8122154049353   "0102000020E610000002000000590E3EDEF5AB23409255B6B4BF1B4A4031197DBBDBAB23404AF663EEB51B4A40"
6116   645  1.82162999509807  "0102000020E61000000300000057900B73277D2340A1675831011E4A4094C55801197D23406A204C40021E4A40B0CFF7AE9C7C234079AAE8B4131E4A40"
  65   695  1.22999509807     "01456020E61000000300000057900B73277D2340A1675831011E4A4094C55801197D23406A204C40021E4A40B0CFF7AE9C7C234079AAE8B4131E4A40"
  65   689  1.556666934862    "0202000020E610000002000000AC0E6F8C53B723405B80118F1F1E4A404EFF48C78BB723406C9159620A1E4A40"

Related answer on gis.SE to illustrate the term "Hausdorff distance":

• Similarity between two or more trajectories

Answer 1

This would achieve it:

SELECT m1.id1, m2.*
FROM   map1 m1
CROSS  JOIN LATERAL (
   SELECT ST_HausdorffDistance(m1.g1, m2.g2) AS h_dist, m2.id2, m2.g2
   FROM   map2 m2
   WHERE  ST_HausdorffDistance(m1.g1, m2.g2) < 2
   ORDER  BY 1, 2
   LIMIT  2
   ) m2;

Returns 1 or 2 rows for every row in map1, extended with the top 2 corresponding row(s) in map2 (as defined by minimum Hausdorff distance) and the said Hausdorff distance between them. If there is no row with Hausdorff distance < 2 in map2, no row is returned.

Key element is the LATERAL subquery. There are variants of this query, depending on exact (missing) requirements. Related:

• OFFSET and LIMIT on complex query
• How to make DISTINCT ON faster in PostgreSQL?
• How to speed up querying last values in a time series?
• Optimise a LATERAL JOIN query on a big table

I wouldn't know of any way to use an index here. So this is going to be an expensive query.

Answer 2

In postgres you can use LIMIT to limit the number of rows returned.

SELECT map1.id1, map2.id2, MIN(ST_HausdorffDistance(map1.g1, map2.g2)),map2.g2
FROM map1, map2
WHERE ST_HausdorffDistance(map1.g1, map2.g2) < 2
GROUP BY map1.id1,map2.g2,map2.id2, ST_HausdorffDistance(map1.g1, map2.g2)
ORDER BY map1.id1,map2.id2, ST_HausdorffDistance(map1.g1, map2.g2)
LIMIT 1

Adding that to the query will limit it to how many rows you specify and the example I provided will just return the first row.

https://www.postgresql.org/docs/current/queries-limit.html

Answer 3

To get your desired output, this might work. However, your result output does not match your explanation.

with a as (
    SELECT map1.id1, map2.id2, MIN(ST_HausdorffDistance(map1.g1, map2.g2)) min,map2.g2
    FROM map1, map2
    WHERE ST_HausdorffDistance(map1.g1, map2.g2) < 2
    GROUP BY map1.id1,map2.g2,map2.id2, ST_HausdorffDistance(map1.g1, map2.g2)
    ORDER BY map1.id1,map2.id2, ST_HausdorffDistance(map1.g1, map2.g2)
), b as (
    SELECT id1, MIN(min) min 
    FROM a
    GROUP BY id1
)
SELECT a.* 
FROM a
INNER JOIN b on a.min = b.min AND a.id1 = b.id1
INNER JOIN map1 on map1.id1 = a.id1
INNER JOIN map2 on map2.id2 = a.id2
ORDER BY a.id1,a.id2, ST_HausdorffDistance(map1.g1, map2.g2)
;