postgresql json取值慢的原因分析

一、缘起

慢sql分析，总行数80w+，通过监控分析慢SQL，某个查询耗时超1s。

比较特殊的是：其中有个字段info是jsonb类型，写法：info::json->'length' as length

同样的查询条件查这个字段和不查这个字段相差3.3倍

那看来就是json取值拖垮了查询的性能。

取jsonb中的字段有多种取法(如下)，那他们有什么区别呢，对性能有啥影响呢？

info::json->'length'
info::jsonb->'length'

info::json->>'length'
info::jsonb->>'length'

info->'length'
info->'length'

info->>'length'
info->>'length'

二、对比

2.1 输出类型对比

查询不同写法的类型：

select
info::json->\’length\’ AS \”info::json->\”, pg_typeof(info::json->\’length\’ ) ,
info::jsonb->\’length\’ AS \”info::jsonb->\” , pg_typeof(info::jsonb->\’length\’ ),
info::json->>\’length\’ AS \”info::json->>\” , pg_typeof(info::json->>\’length\’ ),
info::jsonb->>\’length\’ AS \”info::jsonb->>\” , pg_typeof(info::jsonb->>\’length\’),
info->\’length\’ AS \”info->\” , pg_typeof(info->\’length\’ ),
info->\’length\’ AS \”info->\” , pg_typeof(info->\’length\’ ),
info->>\’length\’ AS \”info->>\” , pg_typeof(info->>\’length\’ ),
info->>\’length\’ AS \”info->>\” , pg_typeof(info->>\’length\’ )
from t_test_json limit 1;

结果

分析小结

->> 输出类型为text
->输出到底为何得看调用它的数据类型，比如：info类型是jsonb, 那么info->'length'为jsonb类型
::json、::jsonb起到类型转换的作用。
info本来就是jsonb类型，info::jsonb算无效转换，是否对性能有影响，待会验证

2.2 性能对比

jihite=> EXPLAIN ANALYSE
jihite-> select
jihite-> info::json->\’length\’ AS \”info::json->\”, pg_typeof(info::json->\’length\’ )
jihite-> from t_test_json limit 1;
QUERY PLAN
—————————————————————————————————————
Limit (cost=0.00..0.04 rows=1 width=36) (actual time=0.028..0.028 rows=1 loops=1)
-> Seq Scan on t_test_json (cost=0.00..30.62 rows=750 width=36) (actual time=0.027..0.027 rows=1 loops=1)
Planning time: 0.056 ms
Execution time: 0.047 ms
(4 rows)
jihite=> EXPLAIN ANALYSE
jihite-> select
jihite-> info::jsonb->\’length\’ AS \”info::jsonb->\” , pg_typeof(info::jsonb->\’length\’ )
jihite-> from t_test_json limit 1
jihite-> ;
QUERY PLAN
—————————————————————————————————————
Limit (cost=0.00..0.03 rows=1 width=36) (actual time=0.017..0.017 rows=1 loops=1)
-> Seq Scan on t_test_json (cost=0.00..23.12 rows=750 width=36) (actual time=0.015..0.015 rows=1 loops=1)
Planning time: 0.053 ms
Execution time: 0.031 ms
(4 rows)
jihite=> EXPLAIN ANALYSE
jihite-> select
jihite-> info::jsonb->\’length\’ AS \”info::jsonb->\” , pg_typeof(info::jsonb->\’length\’ )
jihite-> from t_test_json limit 1;
QUERY PLAN
—————————————————————————————————————
Limit (cost=0.00..0.03 rows=1 width=36) (actual time=0.010..0.010 rows=1 loops=1)
-> Seq Scan on t_test_json (cost=0.00..23.12 rows=750 width=36) (actual time=0.009..0.009 rows=1 loops=1)
Planning time: 0.037 ms
Execution time: 0.022 ms
(4 rows)
jihite=>
jihite=> EXPLAIN ANALYSE
jihite-> select
jihite-> info::json->>\’length\’ AS \”info::json->>\” , pg_typeof(info::json->>\’length\’ )
jihite-> from t_test_json limit 1;
QUERY PLAN
—————————————————————————————————————
Limit (cost=0.00..0.04 rows=1 width=36) (actual time=0.026..0.027 rows=1 loops=1)
-> Seq Scan on t_test_json (cost=0.00..30.62 rows=750 width=36) (actual time=0.025..0.025 rows=1 loops=1)
Planning time: 0.056 ms
Execution time: 0.046 ms
(4 rows)
jihite=>
jihite=> EXPLAIN ANALYSE
jihite-> select
jihite-> info::jsonb->>\’length\’ AS \”info::jsonb->>\” , pg_typeof(info::jsonb->>\’length\’)
jihite-> from t_test_json limit 1;
QUERY PLAN
—————————————————————————————————————
Limit (cost=0.00..0.03 rows=1 width=36) (actual time=0.012..0.012 rows=1 loops=1)
-> Seq Scan on t_test_json (cost=0.00..23.12 rows=750 width=36) (actual time=0.011..0.011 rows=1 loops=1)
Planning time: 0.053 ms
Execution time: 0.029 ms
(4 rows)
jihite=>
jihite=> EXPLAIN ANALYSE
jihite-> select
jihite-> info->\’length\’ AS \”info->\” , pg_typeof(info->\’length\’ )
jihite-> from t_test_json limit 1;
QUERY PLAN
—————————————————————————————————————
Limit (cost=0.00..0.03 rows=1 width=36) (actual time=0.014..0.014 rows=1 loops=1)
-> Seq Scan on t_test_json (cost=0.00..23.12 rows=750 width=36) (actual time=0.013..0.013 rows=1 loops=1)
Planning time: 0.052 ms
Execution time: 0.030 ms
(4 rows)
jihite=>
jihite=> EXPLAIN ANALYSE
jihite-> select
jihite-> info->\’length\’ AS \”info->\” , pg_typeof(info->\’length\’ )
jihite-> from t_test_json limit 1;
QUERY PLAN
—————————————————————————————————————
Limit (cost=0.00..0.03 rows=1 width=36) (actual time=0.013..0.013 rows=1 loops=1)
-> Seq Scan on t_test_json (cost=0.00..23.12 rows=750 width=36) (actual time=0.012..0.012 rows=1 loops=1)
Planning time: 0.051 ms
Execution time: 0.029 ms
(4 rows)
jihite=>
jihite=> EXPLAIN ANALYSE
jihite-> select
jihite-> info->>\’length\’ AS \”info->>\” , pg_typeof(info->>\’length\’ )
jihite-> from t_test_json limit 1;
QUERY PLAN
—————————————————————————————————————
Limit (cost=0.00..0.03 rows=1 width=36) (actual time=0.012..0.013 rows=1 loops=1)
-> Seq Scan on t_test_json (cost=0.00..23.12 rows=750 width=36) (actual time=0.011..0.011 rows=1 loops=1)
Planning time: 0.053 ms
Execution time: 0.030 ms
(4 rows)
jihite=>
jihite=> EXPLAIN ANALYSE
jihite-> select
jihite-> info->>\’length\’ AS \”info->>\” , pg_typeof(info->>\’length\’ )
jihite-> from t_test_json limit 1;
QUERY PLAN
—————————————————————————————————————
Limit (cost=0.00..0.03 rows=1 width=36) (actual time=0.012..0.013 rows=1 loops=1)
-> Seq Scan on t_test_json (cost=0.00..23.12 rows=750 width=36) (actual time=0.011..0.011 rows=1 loops=1)
Planning time: 0.053 ms
Execution time: 0.029 ms
(4 rows)

从执行耗时(Execution time)分析小结

执行了类型转换 jsonb->json，转换性能（0.46ms）显然低出不转换（0.3ms）

三、优化

把查询字段：info::json->'length' 改为info->>'length',减少类型转换导致性能的损耗。

四、待调查

4.1 同类型转换是否影响性能

字段本身是jsonb，进行强转::jsonb 是否对性能造成影响，还是在执行预编译时就已被优化

从大量数据的压测看，转换会对性能有影响，但是不大

4.2 如何分析函数的耗时

在explain analyze时，主要分析了索引对性能的影响，那函数的具体影响如何查看呢？

五、附

5.1 json、jsonb区别

jsonb 性能优于json
jsonb 支持索引
【最大差异：效率】jsonb 写入时会处理写入数据，写入相对较慢，json会保留原始数据（包括无用的空格）

推荐把JSON 数据存储为jsonb

5.2 postgresql查看字段类型函数

pg_typeof()

5.3 性能分析指令

如果您有一条执行很慢的 SQL 语句，您想知道发生了什么以及如何优化它。EXPLAIN ANALYSE 能够获取数据库执行 sql 语句，所经历的过程，以及耗费的时间，可以协助优化性能。

关键参数：

Execution time: *** ms 表明了实际的SQL 执行时间，其中不包括查询计划的生成时间

5.4 示例中的建表语句

# 建表语句

create table t_test_json
(
id bigserial not null PRIMARY KEY,
task character varying not null,
info jsonb not null,
create_time timestamp not null default current_timestamp
);

# 压测数据

insert into t_test_json(task, info) values(\’1\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’2\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’3\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’4\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’5\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’6\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’7\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’8\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’9\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’10\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’11\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’12\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’13\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’14\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’15\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’16\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’17\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’18\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’19\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);
insert into t_test_json(task, info) values(\’20\’, \'{\”length\”: 123.9, \”avatar\”: \”avatar_url\”, \”tags\”: [\”python\”, \”golang\”, \”db\”]}\’);

5.5 示例中的压测脚本

import time
import psycopg
dbname, user, pwd, ip, port = \’\’, \’\’, \’\’, \’\’, \’5432\’
connection = \”dbname=%s user=%s password=%s host=%s port=%s\” % (dbname, user, pwd, ip, port)
db = psycopg.connect(connection)
cur = db.cursor()
ss = 0
lens = 20
for i in range(lens):
s = time.time()
sql = \’\’\’ select
id,
info::json->\’length\’ as length
from
t_test_json
order by id
offset %s limit 1000 \’\’\’ % (i * 1000)
#print(\”sql:\”, sql)
cur.execute(sql)
rev = cur.fetchall()
e = time.time()
print(\”scan:\”, i, e – s)
ss += (e – s)
print(\’avg\’, ss / lens)