记一次pgsql的查询优化
前言
这是一个子查询的场景，对于这个查询我们不能避免子查询，下面是我一次具体的优化过程。
优化策略
1、拆分子查询，将需要的数据提前在cte中查询出来
2、连表查询，直接去查询对应cte里面的内容
WITH RECURSIVE nodes AS (
SELECT t.*
FROM document_directories t
INNER JOIN document_permissions p
ON
p.resource_id = t.id
AND p.enterprise_id = t.enterprise_id
AND p.resource ='directory'
AND p.target = 'role'
INNER JOIN role_user_rel r
ON r.role_id = p.target_id AND r.enterprise_id = p.enterprise_id
WHERE 1 = 1
AND t.enterprise_id = 11
AND 1 = ANY(p.perm)
AND r.uid = 11
UNION
SELECT pn.*
FROM document_directories pn
INNER JOIN nodes n
ON pn.id = n.parent_id
WHERE 1 = 1
),
resJoin AS(
SELECT t.*, dd.perm, e.* FROM nodes t
LEFT JOIN LATERAL (
SELECT array_agg(perm order by perm) perm
FROM (SELECT DISTINCT unnest(perm) as perm
FROM document_permissions d
INNER JOIN roles r ON r.id = d.target_id
WHERE d.resource = 'directory'
AND d.target = 'role'
AND d.enterprise_id = 11
) ss
) AS dd ON TRUE
LEFT JOIN LATERAL(
SELECT NOT EXISTS(
SELECT 1 FROM document_directories as dd
LEFT JOIN documents dm ON dm.directory_id = dd.id
WHERE (dd.parent_id = t.id OR (dm.id IS NOT NULL AND dd.id = t.id AND dm.deleted = false))
) as is_empty_directory
) AS e ON TRUE
ORDER BY t.position DESC
)
SELECT
*
FROM resJoin r
一个RECURSIVE查询出所有的节点信息，后面的resJoin，查询出返回数据需要的信息，里面用到了两个LATERAL，并且里面也用到了子查询。
分析下这个sql
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES ('CTE Scan on resjoin r (cost=116084.38..116086.40 rows=101 width=117) (actual time=2423.410..2423.656 rows=432 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' CTE nodes');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Recursive Union (cost=22.13..1042.04 rows=101 width=71) (actual time=0.309..5.270 rows=432 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Nested Loop (cost=22.13..207.26 rows=1 width=71) (actual time=0.301..1.841 rows=108 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Nested Loop (cost=21.84..200.27 rows=1 width=16) (actual time=0.080..0.741 rows=110 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Join Filter: (p.target_id = r_1.role_id)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Rows Removed by Join Filter: 110');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Index Only Scan using roles_users_rel_enterprise_id_uid_role_id_uindex on role_user_rel r_1 (cost=0.15..8.17 rows=1 width=16) (actual time=0.018..0.021 rows=2 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Index Cond: ((enterprise_id = 11) AND (uid = 11))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Heap Fetches: 2');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Bitmap Heap Scan on document_permissions p (cost=21.69..190.72 rows=110 width=24) (actual time=0.056..0.295 rows=110 loops=2)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Recheck Cond: ((enterprise_id = 11) AND (target = ''role''::text) AND (resource = ''directory''::text))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Filter: (1 = ANY (perm))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Heap Blocks: exact=162');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Bitmap Index Scan on document_permissions_etr_uindex (cost=0.00..21.66 rows=110 width=0) (actual time=0.037..0.037 rows=110 loops=2)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Index Cond: ((enterprise_id = 11) AND (target = ''role''::text) AND (resource = ''directory''::text))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Index Scan using document_directories_pk on document_directories t (cost=0.29..6.96 rows=1 width=71) (actual time=0.008..0.008 rows=1 loops=110)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Index Cond: (id = p.resource_id)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Filter: (enterprise_id = 11)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Rows Removed by Filter: 0');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Nested Loop (cost=0.29..83.28 rows=10 width=71) (actual time=0.057..0.616 rows=81 loops=4)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> WorkTable Scan on nodes n (cost=0.00..0.20 rows=10 width=8) (actual time=0.001..0.034 rows=108 loops=4)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Index Scan using document_directories_pk on document_directories pn (cost=0.29..8.31 rows=1 width=71) (actual time=0.004..0.004 rows=1 loops=432)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Index Cond: (id = n.parent_id)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' CTE resjoin');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Sort (cost=115042.09..115042.34 rows=101 width=117) (actual time=2423.408..2423.451 rows=432 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Sort Key: t_1."position" DESC');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Sort Method: quicksort Memory: 139kB');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Nested Loop Left Join (cost=1437.14..115038.72 rows=101 width=117) (actual time=11.306..2422.614 rows=432 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Nested Loop Left Join (cost=301.19..304.49 rows=101 width=116) (actual time=0.897..7.130 rows=432 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> CTE Scan on nodes t_1 (cost=0.00..2.02 rows=101 width=84) (actual time=0.310..5.849 rows=432 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Materialize (cost=301.19..301.21 rows=1 width=32) (actual time=0.002..0.002 rows=1 loops=432)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Aggregate (cost=301.19..301.20 rows=1 width=32) (actual time=0.581..0.581 rows=1 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> HashAggregate (cost=298.68..299.93 rows=100 width=4) (actual time=0.558..0.559 rows=6 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Group Key: unnest(d.perm)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> ProjectSet (cost=46.31..271.18 rows=11000 width=4) (actual time=0.081..0.401 rows=660 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Hash Join (cost=46.31..215.36 rows=110 width=45) (actual time=0.078..0.229 rows=110 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Hash Cond: (d.target_id = r_2.id)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Bitmap Heap Scan on document_permissions d (cost=21.69..189.35 rows=110 width=53) (actual time=0.053..0.163 rows=110 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Recheck Cond: ((enterprise_id = 11) AND (target = ''role''::text) AND (resource = ''directory''::text))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Heap Blocks: exact=81');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Bitmap Index Scan on document_permissions_etr_uindex (cost=0.00..21.66 rows=110 width=0) (actual time=0.039..0.039 rows=110 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Index Cond: ((enterprise_id = 11) AND (target = ''role''::text) AND (resource = ''directory''::text))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Hash (cost=16.50..16.50 rows=650 width=8) (actual time=0.011..0.011 rows=2 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Buckets: 1024 Batches: 1 Memory Usage: 9kB');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Seq Scan on roles r_2 (cost=0.00..16.50 rows=650 width=8) (actual time=0.006..0.007 rows=2 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Result (cost=1135.95..1135.96 rows=1 width=1) (actual time=5.589..5.590 rows=1 loops=432)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' InitPlan 2 (returns $5)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Nested Loop Left Join (cost=0.00..1135.95 rows=1 width=0) (actual time=5.588..5.588 rows=1 loops=432)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Join Filter: (dm.directory_id = dd.id)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Rows Removed by Join Filter: 42');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Filter: ((dd.parent_id = $4) OR ((dm.id IS NOT NULL) AND (dd.id = $4) AND (NOT dm.deleted)))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Rows Removed by Filter: 1');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Seq Scan on document_directories dd (cost=0.00..1134.06 rows=2 width=16) (actual time=4.287..5.575 rows=2 loops=432)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Filter: ((parent_id = $4) OR (id = $4))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Rows Removed by Filter: 23789');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Materialize (cost=0.00..1.25 rows=17 width=17) (actual time=0.000..0.002 rows=24 loops=756)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Seq Scan on documents dm (cost=0.00..1.17 rows=17 width=17) (actual time=0.005..0.011 rows=24 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES ('Planning time: 1.665 ms');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES ('Execution time: 2424.040 ms');
可以看到性能瓶颈就在resjoin，也就是下面的两个LATERAL查询。这两个查询是不可避免的，所以也不能拿掉，但是pgsql里面的cte支持并行的查询，
我们可以先把这些资源查询出来，然后在连表查询
WITH RECURSIVE nodes AS (
SELECT t.*
FROM document_directories t
INNER JOIN document_permissions p
ON
p.resource_id = t.id
AND p.enterprise_id = t.enterprise_id
AND p.resource ='directory'
AND p.target = 'role'
INNER JOIN role_user_rel r
ON r.role_id = p.target_id AND r.enterprise_id = p.enterprise_id
WHERE 1 = 1
AND t.enterprise_id = 11
AND 1 = ANY(p.perm)
AND r.uid = 11
UNION
SELECT pn.*
FROM document_directories pn
INNER JOIN nodes n
ON pn.id = n.parent_id
WHERE 1 = 1
),
perms AS (
SELECT d.perm, d.resource_id
FROM document_permissions d
INNER JOIN roles r ON r.id = d.target_id
WHERE d.resource = 'directory'
AND d.target = 'role'
AND d.enterprise_id = 11
GROUP BY d.resource_id, d.perm
),
directory_exists AS (
SELECT distinct dd.id
FROM document_directories as dd
LEFT JOIN document_directories t on t.parent_id = dd.id
LEFT JOIN documents dm ON dm.directory_id = dd.id
WHERE (t.id IS NOT NULL OR (dm.id IS NOT NULL AND dm.deleted = false))
),
resJoin AS(
SELECT t.*, dd.perm,de.id IS NULL as is_empty_directory FROM nodes t
LEFT JOIN LATERAL (
SELECT array_agg(perm order by perm) perm
FROM (SELECT DISTINCT unnest(perm) as perm
FROM perms d
WHERE
d.resource_id = t.id
) AS t
) AS dd ON TRUE
LEFT JOIN directory_exists de ON de.id = t.id
ORDER BY t.position
)
SELECT
*
FROM resJoin r
分析下结果
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES ('CTE Scan on resjoin r (cost=8366.56..8770.60 rows=20202 width=117) (actual time=118.496..118.699 rows=432 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' CTE nodes');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Recursive Union (cost=22.13..1042.04 rows=101 width=71) (actual time=0.113..2.833 rows=432 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Nested Loop (cost=22.13..207.26 rows=1 width=71) (actual time=0.109..0.913 rows=108 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Nested Loop (cost=21.84..200.27 rows=1 width=16) (actual time=0.082..0.458 rows=110 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Join Filter: (p.target_id = r_1.role_id)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Rows Removed by Join Filter: 110');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Index Only Scan using roles_users_rel_enterprise_id_uid_role_id_uindex on role_user_rel r_1 (cost=0.15..8.17 rows=1 width=16) (actual time=0.019..0.021 rows=2 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Index Cond: ((enterprise_id = 11) AND (uid = 11))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Heap Fetches: 2');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Bitmap Heap Scan on document_permissions p (cost=21.69..190.72 rows=110 width=24) (actual time=0.052..0.188 rows=110 loops=2)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Recheck Cond: ((enterprise_id = 11) AND (target = ''role''::text) AND (resource = ''directory''::text))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Filter: (1 = ANY (perm))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Heap Blocks: exact=162');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Bitmap Index Scan on document_permissions_etr_uindex (cost=0.00..21.66 rows=110 width=0) (actual time=0.035..0.035 rows=110 loops=2)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Index Cond: ((enterprise_id = 11) AND (target = ''role''::text) AND (resource = ''directory''::text))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Index Scan using document_directories_pk on document_directories t (cost=0.29..6.96 rows=1 width=71) (actual time=0.003..0.003 rows=1 loops=110)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Index Cond: (id = p.resource_id)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Filter: (enterprise_id = 11)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Rows Removed by Filter: 0');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Nested Loop (cost=0.29..83.28 rows=10 width=71) (actual time=0.050..0.348 rows=81 loops=4)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> WorkTable Scan on nodes n (cost=0.00..0.20 rows=10 width=8) (actual time=0.000..0.015 rows=108 loops=4)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Index Scan using document_directories_pk on document_directories pn (cost=0.29..8.31 rows=1 width=71) (actual time=0.002..0.002 rows=1 loops=432)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Index Cond: (id = n.parent_id)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' CTE perms');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> HashAggregate (cost=215.91..217.01 rows=110 width=53) (actual time=0.297..0.321 rows=110 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Group Key: d.resource_id, d.perm');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Hash Join (cost=46.31..215.36 rows=110 width=53) (actual time=0.065..0.216 rows=110 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Hash Cond: (d.target_id = r_2.id)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Bitmap Heap Scan on document_permissions d (cost=21.69..189.35 rows=110 width=61) (actual time=0.044..0.150 rows=110 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Recheck Cond: ((enterprise_id = 11) AND (target = ''role''::text) AND (resource = ''directory''::text))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Heap Blocks: exact=81');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Bitmap Index Scan on document_permissions_etr_uindex (cost=0.00..21.66 rows=110 width=0) (actual time=0.031..0.031 rows=110 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Index Cond: ((enterprise_id = 11) AND (target = ''role''::text) AND (resource = ''directory''::text))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Hash (cost=16.50..16.50 rows=650 width=8) (actual time=0.010..0.010 rows=2 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Buckets: 1024 Batches: 1 Memory Usage: 9kB');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Seq Scan on roles r_2 (cost=0.00..16.50 rows=650 width=8) (actual time=0.005..0.006 rows=2 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' CTE directory_exists');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> HashAggregate (cost=3474.64..3874.68 rows=40004 width=8) (actual time=80.055..88.016 rows=30003 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Group Key: dd.id');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Hash Right Join (cost=3373.22..3374.63 rows=40004 width=8) (actual time=58.189..68.144 rows=30003 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Hash Cond: (dm.directory_id = dd.id)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Filter: ((t_1.id IS NOT NULL) OR ((dm.id IS NOT NULL) AND (NOT dm.deleted)))');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Rows Removed by Filter: 10001');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Seq Scan on documents dm (cost=0.00..1.17 rows=17 width=17) (actual time=0.012..0.018 rows=24 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Hash (cost=2873.17..2873.17 rows=40004 width=16) (actual time=57.854..57.854 rows=40004 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Buckets: 65536 Batches: 1 Memory Usage: 2310kB');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Hash Right Join (cost=1434.09..2873.17 rows=40004 width=16) (actual time=18.692..45.374 rows=40004 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Hash Cond: (t_1.parent_id = dd.id)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Seq Scan on document_directories t_1 (cost=0.00..934.04 rows=40004 width=16) (actual time=0.010..5.319 rows=40004 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Hash (cost=934.04..934.04 rows=40004 width=8) (actual time=18.373..18.373 rows=40004 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Buckets: 65536 Batches: 1 Memory Usage: 2075kB');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Seq Scan on document_directories dd (cost=0.00..934.04 rows=40004 width=8) (actual time=0.008..7.829 rows=40004 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' CTE resjoin');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Sort (cost=3182.33..3232.83 rows=20202 width=117) (actual time=118.494..118.535 rows=432 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Sort Key: t_2."position"');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Sort Method: quicksort Memory: 99kB');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Hash Right Join (cost=585.55..1737.66 rows=20202 width=117) (actual time=95.397..118.216 rows=432 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Hash Cond: (de.id = t_2.id)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> CTE Scan on directory_exists de (cost=0.00..800.08 rows=40004 width=8) (actual time=80.058..97.650 rows=30003 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Hash (cost=584.28..584.28 rows=101 width=116) (actual time=15.277..15.277 rows=432 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Buckets: 1024 Batches: 1 Memory Usage: 63kB');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Nested Loop Left Join (cost=5.74..584.28 rows=101 width=116) (actual time=0.519..15.044 rows=432 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> CTE Scan on nodes t_2 (cost=0.00..2.02 rows=101 width=84) (actual time=0.115..3.140 rows=432 loops=1)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> Aggregate (cost=5.74..5.75 rows=1 width=32) (actual time=0.027..0.027 rows=1 loops=432)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> HashAggregate (cost=3.23..4.48 rows=100 width=4) (actual time=0.023..0.024 rows=2 loops=432)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Group Key: unnest(d_1.perm)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> ProjectSet (cost=0.00..2.98 rows=100 width=4) (actual time=0.019..0.022 rows=2 loops=432)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' -> CTE Scan on perms d_1 (cost=0.00..2.48 rows=1 width=32) (actual time=0.019..0.021 rows=0 loops=432)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Filter: (resource_id = t_2.id)');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES (' Rows Removed by Filter: 110');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES ('Planning time: 1.518 ms');
INSERT INTO "MY_TABLE"("QUERY PLAN") VALUES ('Execution time: 121.197 ms');
发现有了质的飞跃
本文参考链接：https://www.cnblogs.com/ricklz/p/12666731.html
