Postgis / SQL选择元组，以便第一个元组项目是唯一的，并且项目几何形状相交

Question

This question is particularly for Postgres 9.4

Lets say I have two tables:

CREATE TABLE A(id INT);
CREATE TABLE B(id INT);

I'd like to have all tuples (A, B) with a certain condition such that among selected tuples all have different A column:

SELECT DISTINCT ON (A.id) A.id, B.id WHERE condition(A,B);

However DISTINCT ON will perform sorting in memory after all the tuples have been selected and I will like to not select tuples with duplicate A.id at all.

How can this be done in an efficient way?

EDIT:

both A and B have unique ids

EDIT2:

Here is the complete setup:

CREATE EXTENSION postgis;
DROP TABLE A;
DROP TABLE B;
CREATE TABLE A(shape Geometry, id INT);
CREATE TABLE B(shape Geometry, id INT, kind INT);
CREATE INDEX ON A USING GIST (shape);`

I would like to do the following:

SELECT A.id, B.id FROM A, B
WHERE B.id = (SELECT B.id FROM B WHERE
     ST_Intersects(A.shape, B.shape)
     AND ST_Length(ST_Intersection(A.shape, B.shape)) / ST_Length(A.shape) >= 0.5 AND B.kind != 1 LIMIT 1)`

which works (I believe), however is not necessarily the most efficient way. The table A has orders of magnitude more rows than table B. So I am not even sure if the GiST index is right.

I am also aware that the order of arguments in ST_Intersects can have a significant effect on run time. What should the correct order be?

Answer 1

If you want just one row for each "A", you can use a correlated subquery (or lateral join):

select a.id,
       (select b.id
        from b
        where condition(a, b)
        limit 1
       ) as b_id
from a;

This should stop testing for rows from b when the first one is found -- which I imagine is the best approach performance-wise.

If none are found, you will get a NULL value. You can wrap this in a subquery and filter out NULL s.

Answer 2

Try something like:

WITH distinct_a as (
SELECT DISTINCT a.id 
FROM A)
SELECT A.id, B.id 
FROM distinct_a, B
WHERE condition(A,B)

The CTE ( WITH ... ) will select all distinct values first. Then selected values will be used in the next query.