Postgres not using GiST index in a lateral join

[Igor Stassiy <istassiy@xxxxxxxxx>]

Hello,

I have a query plan optimization question. It is formatted nicely on 

http://stackoverflow.com/questions/28856452/postgres-not-using-gist-index-in-lateral-join

But here is a copy for the archive:

Here is the setup:

CREATE EXTENSION postgis;

DROP TABLE IF EXISTS A;

DROP TABLE IF EXISTS B; 

CREATE TABLE A(shape Geometry, id INT); 

CREATE TABLE B(shape Geometry, id INT, kind INT); 

CREATE INDEX ON A USING GIST (shape); 

CREATE INDEX ON B USING GIST (shape);

I am running the following commands:

ANALYZE A; 

ANALYZE B;

-- for each row in A, select exactly one row in B (if there is one)

-- such that B contains geometry of A

EXPLAIN SELECT * FROM (SELECT A.id as aid, (SELECT B.id FROM B WHERE ST_Contains(B.shape, A.shape) AND B.kind != 1 LIMIT 1) as bid FROM A) AS TMP;

which gives me { "Plan": { "Node Type": "Seq Scan", "Relation Name": "A", "Startup Cost": 0.00, "Total Cost": 2606592.33, "Plan Rows": 549745, "Plan Width": 1646, "Plans": [ { "Node Type": "Limit", "Parent Relationship": "SubPlan", "Subplan Name": "SubPlan 1", "Startup Cost": 0.00, "Total Cost": 4.68, "Plan Rows": 1, "Plan Width": 8, "Plans": [ { "Node Type": "Seq Scan", "Parent Relationship": "Outer", "Relation Name": "B", "Startup Cost": 0.00, "Total Cost": 4.68, "Plan Rows": 1, "Plan Width": 8, "Filter": "((shape && A.shape) AND _st_contains(shape, A.shape))" } ] } ] } }


Note that there is a sequential scan inside the lateral join, however there is clearly an index available. However after setting

set enable_seqscan=false;

the index is being used. This actually affects runtime significantly (around 3 times faster) and seems that postgres should figure things like that automatically. { "Plan": { "Node Type": "Seq Scan", "Relation Name": "A", "Startup Cost": 10000000000.00, "Total Cost": 10004716493.85, "Plan Rows": 549450, "Plan Width": 1677, "Plans": [ { "Node Type": "Limit", "Parent Relationship": "SubPlan", "Subplan Name": "SubPlan 1", "Startup Cost": 0.00, "Total Cost": 8.52, "Plan Rows": 1, "Plan Width": 8, "Plans": [ { "Node Type": "Index Scan", "Parent Relationship": "Outer", "Scan Direction": "NoMovement", "Index Name": "B_shape_idx", "Relation Name": "B", "Startup Cost": 0.00, "Total Cost": 8.52, "Plan Rows": 1, "Plan Width": 8, "Index Cond": "(shape && A.shape)", "Filter": "_st_contains(shape, A.shape)" } ] } ] } }

Is there any way to tell postgres to use index in a less hacky way? Possibly by rewriting the query? From what I understand the use of set enable_... is not recommended in production.

When you actually run the commands above it will give

{ "Plan": { "Node Type": "Seq Scan", "Relation Name": "a", "Alias": "a", "Startup Cost": 0.00, "Total Cost": 10372.75, "Plan Rows": 1230, "Plan Width": 36, "Plans": [ { "Node Type": "Limit", "Parent Relationship": "SubPlan", "Subplan Name": "SubPlan 1", "Startup Cost": 0.14, "Total Cost": 8.41, "Plan Rows": 1, "Plan Width": 4, "Plans": [ { "Node Type": "Index Scan", "Parent Relationship": "Outer", "Scan Direction": "NoMovement", "Index Name": "b_shape_idx", "Relation Name": "b", "Alias": "b", "Startup Cost": 0.14, "Total Cost": 8.41, "Plan Rows": 1, "Plan Width": 4, "Index Cond": "(shape && a.shape)", "Filter": "((kind <> 1) AND _st_contains(shape, a.shape))" } ] } ] } }

Unfortunately I cannot provide data to reproduce the query plan results.

Thanks,

Igor