更多细节

这是sql查询：

SELECT COUNT(*) 
FROM "Part" 
WHERE ("part_no"='aeroflex') 
   OR ("msn"='aeroflex') 
   OR ("nsn"='aeroflex') 
   OR ("name" ILIKE '%aeroflex%') 
   OR ("manufacturer" ILIKE '%aeroflex%') 
   OR ("mfg_sku"='aeroflex')

这是表的SQL转储，包含280万行

-- ----------------------------
--  Table structure for Part
-- ----------------------------
DROP TABLE IF EXISTS "electronic_parts"."Part";
CREATE TABLE "electronic_parts"."Part" (
    "id" int4 NOT NULL DEFAULT nextval('"Part_id_seq"'::regclass),
    "part_no" varchar COLLATE "default",
    "manufacturer" varchar COLLATE "default",
    "description" text COLLATE "default",
    "slug" varchar COLLATE "default",
    "nsn" varchar COLLATE "default",
    "price" numeric,
    "name" varchar COLLATE "default",
    "mfg_sku" varchar COLLATE "default",
    "msn" varchar COLLATE "default"
)
WITH (OIDS=FALSE);
ALTER TABLE "electronic_parts"."Part" OWNER TO "root";

-- ----------------------------
--  Primary key structure for table Part
-- ----------------------------
ALTER TABLE "electronic_parts"."Part" ADD PRIMARY KEY ("id") NOT DEFERRABLE INITIALLY IMMEDIATE;

-- ----------------------------
--  Indexes structure for table Part
-- ----------------------------
CREATE INDEX  "part_i1" ON "electronic_parts"."Part" USING btree(msn COLLATE "default" "pg_catalog"."text_ops" ASC NULLS LAST);
CREATE INDEX  "part_i2" ON "electronic_parts"."Part" USING gin(to_tsvector('english'::regconfig, name::text) "pg_catalog"."tsvector_ops") WITH (FASTUPDATE = YES);
CREATE INDEX  "part_i2" ON "electronic_parts"."Part" USING gin(to_tsvector('english'::regconfig, name::text) "pg_catalog"."tsvector_ops") WITH (FASTUPDATE = YES);
CREATE INDEX  "part_i3" ON "electronic_parts"."Part" USING gin(to_tsvector('english'::regconfig, manufacturer::text) "pg_catalog"."tsvector_ops") WITH (FASTUPDATE = YES);
CREATE INDEX  "part_i3" ON "electronic_parts"."Part" USING gin(to_tsvector('english'::regconfig, manufacturer::text) "pg_catalog"."tsvector_ops") WITH (FASTUPDATE = YES);
CREATE INDEX  "part_i4" ON "electronic_parts"."Part" USING gin(to_tsvector('english'::regconfig, description) "pg_catalog"."tsvector_ops") WITH (FASTUPDATE = YES);
CREATE INDEX  "part_i4" ON "electronic_parts"."Part" USING gin(to_tsvector('english'::regconfig, description) "pg_catalog"."tsvector_ops") WITH (FASTUPDATE = YES);
CREATE UNIQUE INDEX  "part_u1" ON "electronic_parts"."Part" USING btree(part_no COLLATE "default" "pg_catalog"."text_ops" ASC NULLS LAST);
CREATE UNIQUE INDEX  "part_u2" ON "electronic_parts"."Part" USING btree(nsn COLLATE "default" "pg_catalog"."text_ops" ASC NULLS LAST);
CREATE UNIQUE INDEX  "part_u3" ON "electronic_parts"."Part" USING btree(mfg_sku COLLATE "default" "pg_catalog"."text_ops" ASC NULLS LAST);
CREATE UNIQUE INDEX  "part_u4" ON "electronic_parts"."Part" USING btree(slug COLLATE "default" "pg_catalog"."text_ops" ASC NULLS LAST);

以下是EXPLAIN

https://explain.depesz.com/s/5Lh

系统规格

DB：x86_64-pc-linux-gnu上的PostgreSQL 9.5.6，由gcc编译（Ubuntu 5.4.0-6ubuntu1~16.04.4）5.4.0 20160609,64位

操作系统：Ubuntu 16.04.2 LTS

RAM：512MB（仅用于开发目的）

服务器上根本没有负载。目前测试为1个用户。

Answer 1

可能，您的查询的以下两部分会降低速度：("name" ILIKE '%aeroflex%')和("manufacturer" ILIKE '%aeroflex%')。这样做的原因是，postgres不能使用常规索引与ilike运算符不能启动字符串（即它可以使用索引来加速{{1} }，但不是'aeroflex%'。

您的索引显示您对此进行了一些研究，通过使用'%aeroflex%'创建索引，但to_tsvector不使用这些索引，而是需要使用特殊运算符{{1与ilike

要利用已创建的to_tsvector索引，请使用where约束，如：

@@

或者，如果您更喜欢使用to_tsquery和where to_tsvector('english', "name") @@ to_tsquery('english', 'aeroflex')，则可以使用like模块。

首先：激活模块

ilike

下一步：使用pgtrm或CREATE EXTENSION pg_trgm;

在要过滤的列上添加索引

like

您可以在文档页面https://www.postgresql.org/docs/current/static/pgtrgm.html中找到有关三元组的更多信息，也可以在https://www.postgresql.org/docs/current/static/datatype-textsearch.html的帖子中找到有关三元组的更多信息

适用于大型PostgreSQL表中搜索的索引设置

更多细节

系统规格

1 个答案: