GIN(Generalized Inverted Index, 通用倒排索引) 是一个存储对(key, posting list)集合的索引结构，其中key是一个键值，而posting list 是一组出现过key的位置。如(‘hello', '14:2 23:4')中，表示hello在14:2和23:4这两个位置出现过，在PG中这些位置实际上就是元组的tid(行号，包括数据块ID（32bit）,以及item point(16 bit) )。
在表中的每一个属性，在建立索引时，都可能会被解析为多个键值，所以同一个元组的tid可能会出现在多个key的posting list中。
通过这种索引结构可以快速的查找到包含指定关键字的元组，因此GIN索引特别适用于多值类型的元素搜索，比如支持全文搜索，数组中元素的搜索，而PG的GIN索引模块最初也是为了支持全文搜索而开发的。
说到这里，你可能会觉得GIN的结构有点像b+tree，包括KEY和对应的值(posting list)。别急，请继续往下看。
- receivers跟reads字段都是数组,使用any函数来查找是个性能差的事情并且加了gin索引CBO也不会用上,使用了2295.317 ms
explain(analyze,verbose,buffers,costs,timing)
SELECT
"msgId",
"sender",
"contentType",
"content",
"expiresIn",
"timestamp",
"sessionType",
"sessionId",
"isRescission",
"isOncePush",
CAST ( "extensions" AS VARCHAR ),
NULL AS "receivers",
NULL AS "reads"
FROM
"表"
WHERE
NOT "isRescission"
AND 7189526 = ANY ( "receivers" )
AND NOT 7189526 = ANY ("reads");
QUERY PLAN
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Seq Scan on public."表" (cost=0.00..140829.79 rows=79282 width=206) (actual time=191.069..2295.282 rows=1 loops=1)
Output: "msgId", sender, "contentType", content, "expiresIn", "timestamp", "sessionType", "sessionId", "isRescission", "isOncePush", (extensions)::character varying, NULL::unknown, NULL::unknown
Filter: ((NOT "表"."isRescission") AND (7189526 = ANY ("表".receivers)) AND (7189526 <> ALL ("表".reads)))
Rows Removed by Filter: 1776067
Buffers: shared hit=79270
Planning time: 0.092 ms
Execution time: 2295.317 ms
- 对表添加gin索引并改写语句达到性能优化目的
CREATE INDEX idxrgin1 on "OnceMessages_test" USING GIN ("receivers");
CREATE INDEX idxrgin2 on "OnceMessages_test" USING GIN ("reads");
改写成array函数 , @>包含, 并强制类型转换::bigint
explain(analyze,verbose,buffers,costs,timing) SELECT "msgId", "sender", "contentType", "content", "expiresIn", "timestamp", "sessionType", "sessionId", "isRescission", "isOncePush", CAST ( "extensions" AS VARCHAR ), NULL AS "receivers", NULL AS "reads" FROM "public"."表_test" where "receivers" @> ARRAY[7189526::bigint] and NOT "isRescission" and "reads" @>ARRAY[7189526::bigint] = 'f' Bitmap Heap Scan on public."表_test" (cost=18.98..1484.00 rows=375 width=207) (actual time=0.110..0.110 rows=1 loops=1) Output: "msgId", sender, "contentType", content, "expiresIn", "timestamp", "sessionType", "sessionId", "isRescission", "isOncePush", (extensions)::character varying, NULL::unknown, NULL::unknown Recheck Cond: ("表_test".receivers @> '{7189526}'::bigint[]) Filter: ((NOT "表_test"."isRescission") AND (NOT ("表_test".reads @> '{7189526}'::bigint[]))) Rows Removed by Filter: 2 Heap Blocks: exact=3 Buffers: shared hit=1 read=6 I/O Timings: read=0.032 -> Bitmap Index Scan on idxrgin1 (cost=0.00..18.89 rows=385 width=0) (actual time=0.057..0.057 rows=3 loops=1) Index Cond: ("表_test".receivers @> '{7189526}'::bigint[]) Buffers: shared hit=1 read=3 I/O Timings: read=0.016 Planning time: 0.299 ms Execution time: 0.149 ms
- 改写之后性能达到 0.149ms , , 函数不能乱用,随便用上性能会不会更慢也不知道,要多看手册, 中国普遍程序员都是这种想用就用的模式,跟国外完全不一样的先理解再使用
否则研发思维只会越来越落后!
CREATE INDEX idxrgin1 on "OnceMessages_test" USING GIN ("receivers");
CREATE INDEX idxrgin2 on "OnceMessages_test" USING GIN ("reads");
