Optimize query performance for millions of rows - sql

Question

I have this PostgreSQL table for storing words:
CREATE TABLE IF NOT EXISTS words
(
id bigint NOT NULL DEFAULT nextval('processed_words_id_seq'::regclass),
keyword character varying(300) COLLATE pg_catalog."default",
);
insert into words (keyword)
VALUES ('while swam is interesting');
CREATE TABLE IF NOT EXISTS trademarks
(
id bigint NOT NULL DEFAULT nextval('trademarks_id_seq'::regclass),
trademark character varying(300) COLLATE pg_catalog."default",
);
Into table trademarks I will have thousands of registered trademarks names.
I want to compare words stored into words table keyword, do they match not only for a words but also for word which is in a group of words. For example:
I have a keyword while swam is interesting stored into words.keyword. I also have a trademark swam located in trademarks.trademark I have a word match, so I want to detect this using SQL.
I tried this:
select
w.id,
w.keyword,
t.trademark
from words w
join trademarks t on t.trademark =
any(string_to_array(w.keyword, ' '))
where 'all' = any(string_to_array(w.keyword, ' '))
The SQL query is implemented properly but execution time is too much. The exception time for table with 30 million of records is 10 seconds. Is there some way to speedup the execution time?

Answer 1

You can create an index on words.keyword which supports the queries with regular expressions :
CREATE INDEX IF NOT EXISTS words_keyword ON words USING GIN (keyword gin_trgm_ops);
This index should be used by a query like :
select
w.id,
w.keyword,
t.trademark
from words w
inner join trademarks t
on w.keyword ~ t.trademark
where w.keyword ~ 'all'
This should be checked with EXPLAIN ANALYSE for this query.