记一次pgsql中查询优化（子查询）
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
记一次pgsql中查询优化（子查询）的更多相关文章
- 详细讲述MySQL中的子查询操作 （来自脚本之家）
继续做以下的前期准备工作: 新建一个测试数据库TestDB: ? 1 create database TestDB; 创建测试表table1和table2: ? 1 2 3 4 5 6 7 8 9 1 ...
- 在 SQL Server 数据库的 WHERE 语句中使用子查询
这是关于子查询语句的一系列文章中的第三篇.在这篇文章中我们将讨论WHERE语句中的子查询语句.其他的文章讨论了其他语句中的子查询语句. 本次课程中的所有例子都是基于Microsoft SQL Serv ...
- 在update语句中使用子查询
在update 中的 where 子句中使用子查询: UPDATE mg_page_log as a SET page_num=1 WHERE id in( SELECT id from mg_ ...
- 子查询。ANY三种用法。ALL两种用法。HAVING中使用子查询。SELECT中使用子查询。
子查询存在的意义是解决多表查询带来的性能问题. 子查询返回单行多列: ANY三种用法: ALL两种用法: HAVING中的子查询返回单行单列: SELECT中使用子查询:(了解就好,避免使用这种方法! ...
- 在MySQL中使用子查询
子查询作为数据源 子查询生成的结果集包含行.列数据,因而非常适合将它与表一起包含在from子句的子查询里.例: SELECT d.dept_id, d.name, e_cnt.how_many num ...
- mysql 在update中实现子查询的方式
当使用mysql条件更新时--最先让人想到的写法 UPDATE buyer SET is_seller=1 WHERE uid IN (SELECT uid FROM seller) 此语句是错误的, ...
- SQL Fundamentals: 子查询 || WHERE,HAVING,FROM,SELECT子句中使用子查询，WITH子句
SQL Fundamentals || Oracle SQL语言 子查询(基础) 1.认识子查询 2.WHERE子句中使用子查询 3.在HAVING子句中使用子查询 4.在FROM子句中使用子查询 5 ...
- update 中实现子查询
mysql 在update中实现子查询的方式 当使用mysql条件更新时--最先让人想到的写法 UPDATE buyer SET is_seller=1 WHERE uid IN (SELECT ...
- 子查询三(在FROM子句中使用子查询)
FROM子句中使用子查询一般都是返回多行多列,可以将其当作一张数据表 示例一.查询出每个部门的编号,名称,位置,部门人数,平均工资 SELECT d.deptno,d.dname,d.loc,temp ...
随机推荐
- Natas27 Writeup（mysql溢出截断漏洞）
Natas27: 一个登录节界面,查看源码. <html> <head> <!-- This stuff in the header has nothing to do ...
- 阿里淘宝的S1级别bug，到底是谁的锅？
3月25日,阿里的淘宝APP在IOS系统上出现BUG: 在打开淘宝APP以后,用户就会收到系统弹窗通知:“您使用的程序是测试/内测版本,将于当地时间2020-03-28到期,到期后将无法使用,请尽快下 ...
- 安装RationalRose的问题解决
列出大问题:在这一步无法进行下一步,直接就只能退出. 翻译过来的意思是:IBM安装程序被完全下载之前就终止了,大概是这个意思. 然后我就直接进了IBM的官网看了一下产品支持,上面解释说是组件clear ...
- 16. nested exception is com.fasterxml.jackson.databind.exc.UnrecognizedPropertyException: Unrecognized field "auditUnitName"
org.springframework.web.servlet.mvc.support.DefaultHandlerExceptionResolver:handleHttpMessageNotRead ...
- 常见排序算法总结分析之选择排序与归并排序-C#实现
本篇文章对选择排序中的简单选择排序与堆排序,以及常用的归并排序做一个总结分析. 常见排序算法总结分析之交换排序与插入排序-C#实现是排序算法总结系列的首篇文章,包含了一些概念的介绍以及交换排序(冒泡与 ...
- Building Applications with Force.com and VisualForce(六)：Designing Applications for Multiple users: Accommodating Multiple Users in your App
Dev 401-006 Designing Applications for Multiple users: Accommodating Multiple Users in your App. Cou ...
- POJ 1062 昂贵的聘礼 最短路+超级源点
Description 年轻的探险家来到了一个印第安部落里.在那里他和酋长的女儿相爱了,于是便向酋长去求亲.酋长要他用10000个金币作为聘礼才答应把女儿嫁给他.探险家拿不出这么多金币,便请求酋长降低 ...
- 一个完整的机器学习项目在Python中的演练（一）
大家往往会选择一本数据科学相关书籍或者完成一门在线课程来学习和掌握机器学习.但是,实际情况往往是,学完之后反而并不清楚这些技术怎样才能被用在实际的项目流程中.就像你的脑海中已经有了一块块"拼 ...
- 火焰图--记一次cpu降温过程
引子 正值周末,娃儿6:30又如闹铃般准时来叫醒了我们.年前离开美菜,又回到了杭州.原本是想有更多时间陪伴娃儿,然而新的工作节奏与工作地点,让我们每天都是早上见面:这不,为了周末可以多玩一会儿,早早就 ...
- docker-compose搭建redis哨兵集群
头脑风暴 出于学习目的,您可以很轻松地在docker环境下运行redis的单个实例,但是如果您需要在生产环境中运行它,那么必须将Redis部署为HA(High Avaliable)模式. Redis ...