Question

我需要通过子字符串匹配计算trade_name表中grls_data列与data ->> 'product_info'表中fsa_raw之间的交集。

表中的值计数：

grls：38596
grls_data：47434
fsa_raw：651380

我有以下SQL架构：

CREATE TABLE public.grls
(
  id integer NOT NULL DEFAULT nextval('grls_id_seq'::regclass),
  use_version integer,
  CONSTRAINT grls_pkey PRIMARY KEY (id),
  CONSTRAINT grls_use_version_foreign FOREIGN KEY (use_version)
      REFERENCES public.grls_data (id) MATCH SIMPLE
      ON UPDATE CASCADE ON DELETE RESTRICT
)
WITH (
  OIDS=FALSE
);

CREATE INDEX grls_use_version_index
  ON public.grls
  USING btree
  (use_version);

CREATE TABLE public.grls_data
(
  id integer NOT NULL DEFAULT nextval('grls_data_id_seq'::regclass),
  grls_id integer NOT NULL,
  trade_name text NOT NULL,
  // other columns
  CONSTRAINT grls_data_pkey PRIMARY KEY (id),
  CONSTRAINT grls_data_grls_id_foreign FOREIGN KEY (grls_id)
      REFERENCES public.grls (id) MATCH SIMPLE
      ON UPDATE CASCADE ON DELETE RESTRICT
)
WITH (
  OIDS=FALSE
);

CREATE INDEX grls_data_grls_id_index
  ON public.grls_data
  USING btree
  (grls_id);


CREATE INDEX grls_data_trade_name_index
  ON public.grls_data
  USING hash
  (trade_name COLLATE pg_catalog."default");

CREATE TABLE public.fsa_raw
(
  id integer NOT NULL DEFAULT nextval('fsa_raw_id_seq'::regclass),
  data jsonb,
  CONSTRAINT fsa_raw_pkey PRIMARY KEY (id)
)
WITH (
  OIDS=FALSE
);

CREATE INDEX fsa_raw_data_idx
  ON public.fsa_raw
  USING gin
  (data jsonb_path_ops);

CREATE INDEX words_trgm_gin
  ON public.fsa_raw
  USING gin
  ((data ->> 'product_info'::text) COLLATE pg_catalog."default" gin_trgm_ops);

所以，我为测试编写了以下SQL查询：

WITH tns AS (
   SELECT DISTINCT ('%' || trade_name || '%') AS tn FROM grls JOIN grls_data ON
   grls.use_version = grls_data.id
)
SELECT COUNT(*) FROM fsa_raw
WHERE (data ->> 'product_info') ILIKE ANY (SELECT tn FROM tns)

但我的查询费用非常高：

汇总（成本= 169494199.21..169494199.22行= 1宽度= 8）

我无法等待数周，数月才能获得查询结果。如何优化我的SQL查询以更快地获得结果？

Postgresql由ILIKE ANY慢速匹配

0 个答案: