public inbox for [email protected]
help / color / mirror / Atom feedLATERAL subquery predicate pushdown: filter applied after JSON construction instead of inside join (PG 18)
3+ messages / 2 participants
[nested] [flat]
* LATERAL subquery predicate pushdown: filter applied after JSON construction instead of inside join (PG 18)
@ 2026-02-11 16:12 Nyasha Chigwamba <[email protected]>
2026-02-11 17:37 ` Re: LATERAL subquery predicate pushdown: filter applied after JSON construction instead of inside join (PG 18) Tom Lane <[email protected]>
0 siblings, 1 reply; 3+ messages in thread
From: Nyasha Chigwamba @ 2026-02-11 16:12 UTC (permalink / raw)
To: [email protected] <[email protected]>; +Cc: [email protected] <[email protected]>
Good day Tom,
Note: this is my first email on this forum, please excuse any missed protocols.
We exploring an option to migrate a project from MongoDB to PostgreSQL and want to ensure that the final solution is performant. We will be keeping some JSONB columns for flexibility and to minimize the scope of application changes. We plan to use views to replace some functionality that were implementing at the application layer. Views use LATERAL joins to assemble related rows into JSON structures.
Version (running inside Docker): PostgreSQL 18.1 (Debian 18.1-1.pgdg13+2) on x86_64-pc-linux-gnu, compiled by gcc (Debian 14.2.0-19) 14.2.0, 64-bit
When filtering on a right-side JSON field in a view built from LATERAL joins, PostgreSQL applies the predicate after the lateral returns. Dependent lookups and JSON construction run for every left-side row instead of only rows that match.
Putting the filter inside the LATERAL subquery avoids this and we have observed a boost in performance: 740 ms → 477 ms, 24,837 → 155 dependency index searches (~160× fewer).
*
View: 740 ms, 467K buffers; filter on Subquery Scan (after JSON built); 24,837 dep index searches
*
Filter in join: 477 ms, 369K buffers; filter on Index Scan (inside LATERAL); 155 dep index searches
Question: Can the planner push predicates on view columns into the underlying LATERAL subqueries?
I have attached the following:
*
schema (right_side_filter_demo_abstract.sql)
*
view plan (explain-plan-filter-on-view.txt)
*
inlined plan (explain-plan-filter-in-join.txt)
Looking forward to your guidance.
Regards,
Nyasha
This communication is confidential and if not addressed to you and has been received in error, you must: (i) notify the sender immediately and delete the e-mail; and (ii) refrain from copying, printing, forwarding, publishing or disclosing the contents of the e-mail.
SET jit = off;
EXPLAIN (ANALYZE, BUFFERS, VERBOSE)
SELECT
u.pkid AS "__pkid",
u.cache AS "User",
sva.service_a_json AS "UserServiceA"
FROM demo_user u
JOIN LATERAL (
SELECT
jsonb_build_object('__pkid', sva.pkid) || sva.cache
|| jsonb_build_object('ServiceADependency', dep.dependency_json) AS service_a_json
FROM demo_user_service_a sva
LEFT JOIN LATERAL (
SELECT jsonb_build_object('__pkid', dep.pkid) || dep.cache AS dependency_json
FROM demo_user_service_a_dependency dep
WHERE dep.scope_id = u.scope_id
AND sva.cache ? 'dependency_ref'
AND LOWER(dep.cache->>'object_id') = LOWER(sva.cache->>'dependency_ref')
ORDER BY dep.pkid
LIMIT 1
) dep ON true
WHERE sva.scope_id = u.scope_id
AND LOWER(sva.cache->>'user_ref') = LOWER(u.cache->>'userid')
AND (sva.cache->>'user_ref') ILIKE '%400%'
ORDER BY sva.pkid
LIMIT 1
) sva ON true
ORDER BY (u.cache->>'userid')
LIMIT 200;
"Limit (cost=17.38..3428.12 rows=200 width=176) (actual time=2.016..476.396 rows=200.00 loops=1)"
" Output: u.pkid, u.cache, (((jsonb_build_object('__pkid', sva.pkid) || sva.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache))))), ((u.cache ->> 'userid'::text))"
" Buffers: shared hit=368941"
" -> Nested Loop (cost=17.38..17053704.65 rows=1000000 width=176) (actual time=2.015..476.372 rows=200.00 loops=1)"
" Output: u.pkid, u.cache, (((jsonb_build_object('__pkid', sva.pkid) || sva.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache))))), (u.cache ->> 'userid'::text)"
" Buffers: shared hit=368941"
" -> Index Scan using idx_demo_user_userid_raw on public.demo_user u (cost=0.42..68704.65 rows=1000000 width=120) (actual time=0.035..16.573 rows=99338.00 loops=1)"
" Output: u.pkid, u.scope_id, u.cache"
" Index Searches: 1"
" Buffers: shared hit=20636"
" -> Limit (cost=16.96..16.96 rows=1 width=57) (actual time=0.004..0.004 rows=0.00 loops=99338)"
" Output: (((jsonb_build_object('__pkid', sva.pkid) || sva.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache))))), sva.pkid"
" Buffers: shared hit=348305"
" -> Sort (cost=16.96..16.96 rows=1 width=57) (actual time=0.004..0.004 rows=0.00 loops=99338)"
" Output: (((jsonb_build_object('__pkid', sva.pkid) || sva.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache))))), sva.pkid"
" Sort Key: sva.pkid"
" Sort Method: quicksort Memory: 25kB"
" Buffers: shared hit=348305"
" -> Nested Loop Left Join (cost=8.89..16.95 rows=1 width=57) (actual time=0.004..0.004 rows=0.00 loops=99338)"
" Output: ((jsonb_build_object('__pkid', sva.pkid) || sva.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache)))), sva.pkid"
" Buffers: shared hit=348305"
" -> Index Scan using idx_demo_user_service_a_user_ref on public.demo_user_service_a sva (cost=0.43..8.45 rows=1 width=72) (actual time=0.004..0.004 rows=0.00 loops=99338)"
" Output: sva.pkid, sva.scope_id, sva.cache"
" Index Cond: (lower((sva.cache ->> 'user_ref'::text)) = lower((u.cache ->> 'userid'::text)))"
" Filter: ((sva.scope_id = u.scope_id) AND ((sva.cache ->> 'user_ref'::text) ~~* '%400%'::text))"
" Rows Removed by Filter: 0"
" Index Searches: 99338"
" Buffers: shared hit=347685"
" -> Limit (cost=8.46..8.47 rows=1 width=57) (actual time=0.012..0.012 rows=0.78 loops=200)"
" Output: ((jsonb_build_object('__pkid', dep.pkid) || dep.cache)), dep.pkid"
" Buffers: shared hit=620"
" -> Sort (cost=8.46..8.47 rows=1 width=57) (actual time=0.012..0.012 rows=0.78 loops=200)"
" Output: ((jsonb_build_object('__pkid', dep.pkid) || dep.cache)), dep.pkid"
" Sort Key: dep.pkid"
" Sort Method: quicksort Memory: 25kB"
" Buffers: shared hit=620"
" -> Result (cost=0.42..8.45 rows=1 width=57) (actual time=0.010..0.010 rows=0.78 loops=200)"
" Output: (jsonb_build_object('__pkid', dep.pkid) || dep.cache), dep.pkid"
" One-Time Filter: (sva.cache ? 'dependency_ref'::text)"
" Buffers: shared hit=620"
" -> Index Scan using idx_demo_user_service_a_dep_object_id on public.demo_user_service_a_dependency dep (cost=0.42..8.45 rows=1 width=90) (actual time=0.006..0.006 rows=1.00 loops=155)"
" Output: dep.pkid, dep.scope_id, dep.cache"
" Index Cond: (lower((dep.cache ->> 'object_id'::text)) = lower((sva.cache ->> 'dependency_ref'::text)))"
" Filter: (dep.scope_id = u.scope_id)"
" Index Searches: 155"
" Buffers: shared hit=620"
"Planning:"
" Buffers: shared hit=1"
"Planning Time: 0.410 ms"
"Execution Time: 476.512 ms"
EXPLAIN (ANALYZE, BUFFERS, VERBOSE)
SELECT *
FROM user_service_view
WHERE ("UserServiceA"->>'user_ref') ILIKE '%400%'
ORDER BY ("User"->>'userid')
LIMIT 200;
"Limit (cost=17.38..3428.11 rows=200 width=176) (actual time=9.490..740.152 rows=200.00 loops=1)"
" Output: u.pkid, u.cache, sva.service_a_json, ((u.cache ->> 'userid'::text))"
" Buffers: shared hit=467669"
" -> Nested Loop (cost=17.38..17053704.65 rows=1000000 width=176) (actual time=9.485..740.121 rows=200.00 loops=1)"
" Output: u.pkid, u.cache, sva.service_a_json, (u.cache ->> 'userid'::text)"
" Buffers: shared hit=467669"
" -> Index Scan using idx_demo_user_userid_raw on public.demo_user u (cost=0.42..68704.65 rows=1000000 width=120) (actual time=0.082..22.218 rows=99338.00 loops=1)"
" Output: u.pkid, u.scope_id, u.cache"
" Index Searches: 1"
" Buffers: shared hit=20636"
" -> Subquery Scan on sva (cost=16.95..16.97 rows=1 width=32) (actual time=0.007..0.007 rows=0.00 loops=99338)"
" Output: sva.service_a_json, sva_1.pkid"
" Filter: ((sva.service_a_json ->> 'user_ref'::text) ~~* '%400%'::text)"
" Rows Removed by Filter: 0"
" Buffers: shared hit=447033"
" -> Limit (cost=16.95..16.96 rows=1 width=57) (actual time=0.007..0.007 rows=0.50 loops=99338)"
" Output: (((jsonb_build_object('__pkid', sva_1.pkid) || sva_1.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache))))), sva_1.pkid"
" Buffers: shared hit=447033"
" -> Sort (cost=16.95..16.96 rows=1 width=57) (actual time=0.007..0.007 rows=0.50 loops=99338)"
" Output: (((jsonb_build_object('__pkid', sva_1.pkid) || sva_1.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache))))), sva_1.pkid"
" Sort Key: sva_1.pkid"
" Sort Method: quicksort Memory: 25kB"
" Buffers: shared hit=447033"
" -> Nested Loop Left Join (cost=8.89..16.94 rows=1 width=57) (actual time=0.006..0.006 rows=0.50 loops=99338)"
" Output: ((jsonb_build_object('__pkid', sva_1.pkid) || sva_1.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache)))), sva_1.pkid"
" Buffers: shared hit=447033"
" -> Index Scan using idx_demo_user_service_a_user_ref on public.demo_user_service_a sva_1 (cost=0.43..8.45 rows=1 width=72) (actual time=0.003..0.003 rows=0.50 loops=99338)"
" Output: sva_1.pkid, sva_1.scope_id, sva_1.cache"
" Index Cond: (lower((sva_1.cache ->> 'user_ref'::text)) = lower((u.cache ->> 'userid'::text)))"
" Filter: (sva_1.scope_id = u.scope_id)"
" Index Searches: 99338"
" Buffers: shared hit=347685"
" -> Limit (cost=8.46..8.47 rows=1 width=57) (actual time=0.003..0.003 rows=0.50 loops=49671)"
" Output: ((jsonb_build_object('__pkid', dep.pkid) || dep.cache)), dep.pkid"
" Buffers: shared hit=99348"
" -> Sort (cost=8.46..8.47 rows=1 width=57) (actual time=0.003..0.003 rows=0.50 loops=49671)"
" Output: ((jsonb_build_object('__pkid', dep.pkid) || dep.cache)), dep.pkid"
" Sort Key: dep.pkid"
" Sort Method: quicksort Memory: 25kB"
" Buffers: shared hit=99348"
" -> Result (cost=0.42..8.45 rows=1 width=57) (actual time=0.002..0.002 rows=0.50 loops=49671)"
" Output: (jsonb_build_object('__pkid', dep.pkid) || dep.cache), dep.pkid"
" One-Time Filter: (sva_1.cache ? 'dependency_ref'::text)"
" Buffers: shared hit=99348"
" -> Index Scan using idx_demo_user_service_a_dep_object_id on public.demo_user_service_a_dependency dep (cost=0.42..8.45 rows=1 width=90) (actual time=0.003..0.003 rows=1.00 loops=24837)"
" Output: dep.pkid, dep.scope_id, dep.cache"
" Index Cond: (lower((dep.cache ->> 'object_id'::text)) = lower((sva_1.cache ->> 'dependency_ref'::text)))"
" Filter: (dep.scope_id = u.scope_id)"
" Index Searches: 24837"
" Buffers: shared hit=99348"
"Planning Time: 1.287 ms"
"Execution Time: 740.338 ms"
Attachments:
[text/plain] explain-plan-filter-in-join.txt (5.7K, ../../CH0PR10MB4908BD90B3C5155DF98D88A2B463A@CH0PR10MB4908.namprd10.prod.outlook.com/3-explain-plan-filter-in-join.txt)
download | inline:
SET jit = off;
EXPLAIN (ANALYZE, BUFFERS, VERBOSE)
SELECT
u.pkid AS "__pkid",
u.cache AS "User",
sva.service_a_json AS "UserServiceA"
FROM demo_user u
JOIN LATERAL (
SELECT
jsonb_build_object('__pkid', sva.pkid) || sva.cache
|| jsonb_build_object('ServiceADependency', dep.dependency_json) AS service_a_json
FROM demo_user_service_a sva
LEFT JOIN LATERAL (
SELECT jsonb_build_object('__pkid', dep.pkid) || dep.cache AS dependency_json
FROM demo_user_service_a_dependency dep
WHERE dep.scope_id = u.scope_id
AND sva.cache ? 'dependency_ref'
AND LOWER(dep.cache->>'object_id') = LOWER(sva.cache->>'dependency_ref')
ORDER BY dep.pkid
LIMIT 1
) dep ON true
WHERE sva.scope_id = u.scope_id
AND LOWER(sva.cache->>'user_ref') = LOWER(u.cache->>'userid')
AND (sva.cache->>'user_ref') ILIKE '%400%'
ORDER BY sva.pkid
LIMIT 1
) sva ON true
ORDER BY (u.cache->>'userid')
LIMIT 200;
"Limit (cost=17.38..3428.12 rows=200 width=176) (actual time=2.016..476.396 rows=200.00 loops=1)"
" Output: u.pkid, u.cache, (((jsonb_build_object('__pkid', sva.pkid) || sva.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache))))), ((u.cache ->> 'userid'::text))"
" Buffers: shared hit=368941"
" -> Nested Loop (cost=17.38..17053704.65 rows=1000000 width=176) (actual time=2.015..476.372 rows=200.00 loops=1)"
" Output: u.pkid, u.cache, (((jsonb_build_object('__pkid', sva.pkid) || sva.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache))))), (u.cache ->> 'userid'::text)"
" Buffers: shared hit=368941"
" -> Index Scan using idx_demo_user_userid_raw on public.demo_user u (cost=0.42..68704.65 rows=1000000 width=120) (actual time=0.035..16.573 rows=99338.00 loops=1)"
" Output: u.pkid, u.scope_id, u.cache"
" Index Searches: 1"
" Buffers: shared hit=20636"
" -> Limit (cost=16.96..16.96 rows=1 width=57) (actual time=0.004..0.004 rows=0.00 loops=99338)"
" Output: (((jsonb_build_object('__pkid', sva.pkid) || sva.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache))))), sva.pkid"
" Buffers: shared hit=348305"
" -> Sort (cost=16.96..16.96 rows=1 width=57) (actual time=0.004..0.004 rows=0.00 loops=99338)"
" Output: (((jsonb_build_object('__pkid', sva.pkid) || sva.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache))))), sva.pkid"
" Sort Key: sva.pkid"
" Sort Method: quicksort Memory: 25kB"
" Buffers: shared hit=348305"
" -> Nested Loop Left Join (cost=8.89..16.95 rows=1 width=57) (actual time=0.004..0.004 rows=0.00 loops=99338)"
" Output: ((jsonb_build_object('__pkid', sva.pkid) || sva.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache)))), sva.pkid"
" Buffers: shared hit=348305"
" -> Index Scan using idx_demo_user_service_a_user_ref on public.demo_user_service_a sva (cost=0.43..8.45 rows=1 width=72) (actual time=0.004..0.004 rows=0.00 loops=99338)"
" Output: sva.pkid, sva.scope_id, sva.cache"
" Index Cond: (lower((sva.cache ->> 'user_ref'::text)) = lower((u.cache ->> 'userid'::text)))"
" Filter: ((sva.scope_id = u.scope_id) AND ((sva.cache ->> 'user_ref'::text) ~~* '%400%'::text))"
" Rows Removed by Filter: 0"
" Index Searches: 99338"
" Buffers: shared hit=347685"
" -> Limit (cost=8.46..8.47 rows=1 width=57) (actual time=0.012..0.012 rows=0.78 loops=200)"
" Output: ((jsonb_build_object('__pkid', dep.pkid) || dep.cache)), dep.pkid"
" Buffers: shared hit=620"
" -> Sort (cost=8.46..8.47 rows=1 width=57) (actual time=0.012..0.012 rows=0.78 loops=200)"
" Output: ((jsonb_build_object('__pkid', dep.pkid) || dep.cache)), dep.pkid"
" Sort Key: dep.pkid"
" Sort Method: quicksort Memory: 25kB"
" Buffers: shared hit=620"
" -> Result (cost=0.42..8.45 rows=1 width=57) (actual time=0.010..0.010 rows=0.78 loops=200)"
" Output: (jsonb_build_object('__pkid', dep.pkid) || dep.cache), dep.pkid"
" One-Time Filter: (sva.cache ? 'dependency_ref'::text)"
" Buffers: shared hit=620"
" -> Index Scan using idx_demo_user_service_a_dep_object_id on public.demo_user_service_a_dependency dep (cost=0.42..8.45 rows=1 width=90) (actual time=0.006..0.006 rows=1.00 loops=155)"
" Output: dep.pkid, dep.scope_id, dep.cache"
" Index Cond: (lower((dep.cache ->> 'object_id'::text)) = lower((sva.cache ->> 'dependency_ref'::text)))"
" Filter: (dep.scope_id = u.scope_id)"
" Index Searches: 155"
" Buffers: shared hit=620"
"Planning:"
" Buffers: shared hit=1"
"Planning Time: 0.410 ms"
"Execution Time: 476.512 ms"
[text/plain] explain-plan-filter-on-view.txt (5.1K, ../../CH0PR10MB4908BD90B3C5155DF98D88A2B463A@CH0PR10MB4908.namprd10.prod.outlook.com/4-explain-plan-filter-on-view.txt)
download | inline:
EXPLAIN (ANALYZE, BUFFERS, VERBOSE)
SELECT *
FROM user_service_view
WHERE ("UserServiceA"->>'user_ref') ILIKE '%400%'
ORDER BY ("User"->>'userid')
LIMIT 200;
"Limit (cost=17.38..3428.11 rows=200 width=176) (actual time=9.490..740.152 rows=200.00 loops=1)"
" Output: u.pkid, u.cache, sva.service_a_json, ((u.cache ->> 'userid'::text))"
" Buffers: shared hit=467669"
" -> Nested Loop (cost=17.38..17053704.65 rows=1000000 width=176) (actual time=9.485..740.121 rows=200.00 loops=1)"
" Output: u.pkid, u.cache, sva.service_a_json, (u.cache ->> 'userid'::text)"
" Buffers: shared hit=467669"
" -> Index Scan using idx_demo_user_userid_raw on public.demo_user u (cost=0.42..68704.65 rows=1000000 width=120) (actual time=0.082..22.218 rows=99338.00 loops=1)"
" Output: u.pkid, u.scope_id, u.cache"
" Index Searches: 1"
" Buffers: shared hit=20636"
" -> Subquery Scan on sva (cost=16.95..16.97 rows=1 width=32) (actual time=0.007..0.007 rows=0.00 loops=99338)"
" Output: sva.service_a_json, sva_1.pkid"
" Filter: ((sva.service_a_json ->> 'user_ref'::text) ~~* '%400%'::text)"
" Rows Removed by Filter: 0"
" Buffers: shared hit=447033"
" -> Limit (cost=16.95..16.96 rows=1 width=57) (actual time=0.007..0.007 rows=0.50 loops=99338)"
" Output: (((jsonb_build_object('__pkid', sva_1.pkid) || sva_1.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache))))), sva_1.pkid"
" Buffers: shared hit=447033"
" -> Sort (cost=16.95..16.96 rows=1 width=57) (actual time=0.007..0.007 rows=0.50 loops=99338)"
" Output: (((jsonb_build_object('__pkid', sva_1.pkid) || sva_1.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache))))), sva_1.pkid"
" Sort Key: sva_1.pkid"
" Sort Method: quicksort Memory: 25kB"
" Buffers: shared hit=447033"
" -> Nested Loop Left Join (cost=8.89..16.94 rows=1 width=57) (actual time=0.006..0.006 rows=0.50 loops=99338)"
" Output: ((jsonb_build_object('__pkid', sva_1.pkid) || sva_1.cache) || jsonb_build_object('ServiceADependency', ((jsonb_build_object('__pkid', dep.pkid) || dep.cache)))), sva_1.pkid"
" Buffers: shared hit=447033"
" -> Index Scan using idx_demo_user_service_a_user_ref on public.demo_user_service_a sva_1 (cost=0.43..8.45 rows=1 width=72) (actual time=0.003..0.003 rows=0.50 loops=99338)"
" Output: sva_1.pkid, sva_1.scope_id, sva_1.cache"
" Index Cond: (lower((sva_1.cache ->> 'user_ref'::text)) = lower((u.cache ->> 'userid'::text)))"
" Filter: (sva_1.scope_id = u.scope_id)"
" Index Searches: 99338"
" Buffers: shared hit=347685"
" -> Limit (cost=8.46..8.47 rows=1 width=57) (actual time=0.003..0.003 rows=0.50 loops=49671)"
" Output: ((jsonb_build_object('__pkid', dep.pkid) || dep.cache)), dep.pkid"
" Buffers: shared hit=99348"
" -> Sort (cost=8.46..8.47 rows=1 width=57) (actual time=0.003..0.003 rows=0.50 loops=49671)"
" Output: ((jsonb_build_object('__pkid', dep.pkid) || dep.cache)), dep.pkid"
" Sort Key: dep.pkid"
" Sort Method: quicksort Memory: 25kB"
" Buffers: shared hit=99348"
" -> Result (cost=0.42..8.45 rows=1 width=57) (actual time=0.002..0.002 rows=0.50 loops=49671)"
" Output: (jsonb_build_object('__pkid', dep.pkid) || dep.cache), dep.pkid"
" One-Time Filter: (sva_1.cache ? 'dependency_ref'::text)"
" Buffers: shared hit=99348"
" -> Index Scan using idx_demo_user_service_a_dep_object_id on public.demo_user_service_a_dependency dep (cost=0.42..8.45 rows=1 width=90) (actual time=0.003..0.003 rows=1.00 loops=24837)"
" Output: dep.pkid, dep.scope_id, dep.cache"
" Index Cond: (lower((dep.cache ->> 'object_id'::text)) = lower((sva_1.cache ->> 'dependency_ref'::text)))"
" Filter: (dep.scope_id = u.scope_id)"
" Index Searches: 24837"
" Buffers: shared hit=99348"
"Planning Time: 1.287 ms"
"Execution Time: 740.338 ms"
[application/octet-stream] right_side_filter_demo_abstract.sql (5.9K, ../../CH0PR10MB4908BD90B3C5155DF98D88A2B463A@CH0PR10MB4908.namprd10.prod.outlook.com/5-right_side_filter_demo_abstract.sql)
download
^ permalink raw reply [nested|flat] 3+ messages in thread
* Re: LATERAL subquery predicate pushdown: filter applied after JSON construction instead of inside join (PG 18)
2026-02-11 16:12 LATERAL subquery predicate pushdown: filter applied after JSON construction instead of inside join (PG 18) Nyasha Chigwamba <[email protected]>
@ 2026-02-11 17:37 ` Tom Lane <[email protected]>
2026-02-23 10:52 ` Re: LATERAL subquery predicate pushdown: filter applied after JSON construction instead of inside join (PG 18) Nyasha Chigwamba <[email protected]>
0 siblings, 1 reply; 3+ messages in thread
From: Tom Lane @ 2026-02-11 17:37 UTC (permalink / raw)
To: Nyasha Chigwamba <[email protected]>; +Cc: [email protected] <[email protected]>
Nyasha Chigwamba <[email protected]> writes:
> Question: Can the planner push predicates on view columns into the underlying LATERAL subqueries?
It can, but it will not push them into a subquery with LIMIT,
because that would potentially change the subquery result.
I'd try to get rid of all those ORDER BY LIMIT bits in your
view definition. That's generally considered an antipattern
in SQL. It's definitely an optimization fence.
regards, tom lane
^ permalink raw reply [nested|flat] 3+ messages in thread
* Re: LATERAL subquery predicate pushdown: filter applied after JSON construction instead of inside join (PG 18)
2026-02-11 16:12 LATERAL subquery predicate pushdown: filter applied after JSON construction instead of inside join (PG 18) Nyasha Chigwamba <[email protected]>
2026-02-11 17:37 ` Re: LATERAL subquery predicate pushdown: filter applied after JSON construction instead of inside join (PG 18) Tom Lane <[email protected]>
@ 2026-02-23 10:52 ` Nyasha Chigwamba <[email protected]>
0 siblings, 0 replies; 3+ messages in thread
From: Nyasha Chigwamba @ 2026-02-23 10:52 UTC (permalink / raw)
To: Tom Lane <[email protected]>; +Cc: [email protected] <[email protected]>
Hi Tom,
Thank you for this - much appreciated. Removing the ORDER BY LIMIT does change the query and after some further investigation, it looks like I also need to project the raw columns then apply the filters on the projected columns as a opposed to filtering on the constructed JSONB.
Regards,
Nyasha
________________________________
From: Tom Lane <[email protected]>
Sent: February 11, 2026 12:37 PM
To: Nyasha Chigwamba <[email protected]>
Cc: [email protected] <[email protected]>
Subject: Re: LATERAL subquery predicate pushdown: filter applied after JSON construction instead of inside join (PG 18)
Nyasha Chigwamba <[email protected]> writes:
> Question: Can the planner push predicates on view columns into the underlying LATERAL subqueries?
It can, but it will not push them into a subquery with LIMIT,
because that would potentially change the subquery result.
I'd try to get rid of all those ORDER BY LIMIT bits in your
view definition. That's generally considered an antipattern
in SQL. It's definitely an optimization fence.
regards, tom lane
This communication is confidential and if not addressed to you and has been received in error, you must: (i) notify the sender immediately and delete the e-mail; and (ii) refrain from copying, printing, forwarding, publishing or disclosing the contents of the e-mail.
^ permalink raw reply [nested|flat] 3+ messages in thread
end of thread, other threads:[~2026-02-23 10:52 UTC | newest]
Thread overview: 3+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2026-02-11 16:12 LATERAL subquery predicate pushdown: filter applied after JSON construction instead of inside join (PG 18) Nyasha Chigwamba <[email protected]>
2026-02-11 17:37 ` Tom Lane <[email protected]>
2026-02-23 10:52 ` Nyasha Chigwamba <[email protected]>
This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox