PostgreSQL-模糊查询

drop function if exists clear_punctuation(text);
create or replace function clear_punctuation(text) 
  returns text
as $$
    select regexp_replace($1,
	'[\ |\~|\`|\!|\@|\#|\$|\%|\^|\&|\*|\(|\)|\-|\_|\+|\=|\||\\|\[|\]|\{|\}|\;|\:|\"|\''|\,|\<|\.|\>|\/|\?|\：|\。|\；|\，|\：|\“|\”|\（|\）|\、|\？|\《|\》]'
	,'','g');
$$ language sql strict immutable;

drop function if exists str_to_tsvector(text);
create or replace function str_to_tsvector(text) 
returns tsvector
as $$      
	declare
		v_count integer;
		v_txt text;
		v_txts text[];
		v_result tsvector;
	begin
		v_txt := clear_punctuation($1);
		--数组大小为字符数量-1
		v_count := length(v_txt)-1;
		if( v_count < 1 ) then
			raise exception '输入参数("%")去除标点符号后至少需要2个字符',$1;
		end if;
		for i in 1..v_count loop      
			v_txts := array_append(v_txts, substring(v_txt,i,2));      
		end loop;
		--tsvector类型要求去除重复并排序
		with cte1 as(
			select f from unnest(v_txts) as f group by f
		),cte2 as(
			select f from cte1 order by f
		)select array_to_tsvector(array_agg(f)) into v_result from cte2;
		return v_result;
	end;      
$$ language plpgsql strict immutable;

drop table if exists test_cond;
drop table if exists test;
create table test(
	objectid bigserial not null,					--唯一编号
	name text not null,								--名称
	newtime timestamptz default(now()) not null,	--首次添加时间
	lastime timestamptz,							--最后一次修改时间	
	constraint pk_test_objectid primary key (objectid)
);
--rum需要大量的计算和写入,保存速度非常慢,因此创建表时设置unlogged标记
--unlogged标记不是绝对安全的,因此和主表分开
--虽然浪费了一定的磁盘空间,但可以最大程度保证写入速度
--test_cond表损坏后,通过主表可以很容易的重建,不过机率非常小
--test_cond中的数据通过触发器更新,不需要操作这个表
create unlogged table test_cond(
	objectid bigserial not null,					--唯一编号
	keys tsvector not null,							--关键字
	constraint pk_test_cond_objectid primary key (objectid),
	constraint fk_test_cond_objectid foreign key(objectid) references test(objectid) on delete cascade
);
create index idx_test_cond_keys on test_cond using rum(keys rum_tsvector_ops);

--test_cond表设置了级联删除,所以不需要delete触发
drop trigger if exists tri_test_change on test;
drop function if exists tri_test_trigger();
create or replace function tri_test_trigger() 
returns trigger as $$ 
        begin 
			if (TG_OP = 'INSERT') then					
				insert into test_cond(objectid,keys) values(NEW.objectid,str_to_tsvector(NEW.name));
				return NEW;
			elsif (TG_OP = 'UPDATE') then
				update test_cond set keys=str_to_tsvector(NEW.name) where objectid=NEW.objectid;
				return NEW;			
			end if;
			 RETURN NULL; 
        end; 
$$ language 'plpgsql' SECURITY DEFINER;
--test_cond依赖test表,因此只能在test成功后再触好,只能设置为after
create trigger tri_test_change after INSERT or UPDATE on test 
	for each ROW EXECUTE PROCEDURE tri_test_trigger();

--test_cond表设置了级联删除,所以不需要delete触发
drop trigger if exists tri_test_change on test;
drop function if exists tri_test_trigger();
create or replace function tri_test_trigger() 
returns trigger as $$ 
        begin 
			if (TG_OP = 'INSERT') then					
				insert into test_cond(objectid,keys) values(NEW.objectid,dichotomy_split_tsv(NEW.name));
				return NEW;
			elsif (TG_OP = 'UPDATE') then
				update test_cond set keys=dichotomy_split_tsv(NEW.name) where objectid=NEW.objectid;
				return NEW;			
			end if;
			 RETURN NULL; 
        end; 
$$ language 'plpgsql' SECURITY DEFINER;
--test_cond依赖test表,因此只能在test成功后再触好,只能设置为after
create trigger tri_test_change after INSERT or UPDATE on test 
	for each ROW EXECUTE PROCEDURE tri_test_trigger();

drop function if exists gen_random_zh(int,int);
create or replace function gen_random_zh(int,int)
	returns text
as $$
	select string_agg(chr((random()*(20901-19968)+19968 )::integer) , '')  from generate_series(1,(random()*($2-$1)+$1)::integer);
$$ language sql;

drop function if exists ins_test();
create or replace function ins_test() 
	returns void 
as $$
	declare 
		v_index integer;
	begin
		v_index := 0;
		for i in 1..1000 loop	
			v_index := v_index + 1;
			insert into test(name)
				select gen_random_zh(8,32) as name from generate_series(1,1000);
			raise notice  '%', v_index;
		end loop;
	end;
$$ language plpgsql;
--每调一次ins_test插入100万数据
select ins_test();

update test set name='哈哈,我来验证了' where objectid=10000;

select count(*) from test;
select count(*) from test_cond;

drop function if exists str_to_tsquery(text,boolean);
create or replace function str_to_tsquery(text,boolean default true) 
returns tsquery
as $$      
	declare
		v_count integer;
		v_txt text;
		v_txts text[];
		v_result tsquery;
	begin
		v_txt := clear_punctuation($1);
		--数组大小为字符数量-1
		v_count := length(v_txt)-1;
		if( v_count < 1 ) then
			raise exception '输入参数("%")去除标点符号后至少需要2个字符',$1;
		end if;
		for i in 1..v_count loop      
			v_txts := array_append(v_txts, substring(v_txt,i,2));      
		end loop;
		--tsquery类型要求去除重复并排序
		with cte1 as(
			select f from unnest(v_txts) as f group by f
		),cte2 as(
			select f from cte1 order by f
		)select string_agg(f, (case when $2 then '&' else '|' end ) )::tsquery into v_result from cte2;
		return v_result;
	end;      
$$ language plpgsql strict immutable;

--因优先级问题,此语句可能会不走rum索引
explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where name @@ (str_to_tsquery('価仴'))

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select str_to_tsquery('哈哈'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select str_to_tsquery('哈我'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select str_to_tsquery('我来'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select str_to_tsquery('来验'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select str_to_tsquery('验证'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select str_to_tsquery('证了'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select str_to_tsquery('哈哈,我来验证了'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select str_to_tsquery('侒亩'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select dichotomy_split_tsq('哈哈'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select dichotomy_split_tsq('哈我'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select dichotomy_split_tsq('我来'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select dichotomy_split_tsq('来验'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select dichotomy_split_tsq('验证'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select dichotomy_split_tsq('证了'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select dichotomy_split_tsq('哈哈,我来验证了'));

explain (analyze,verbose,costs,buffers,timing) 
select f.* from test as f left join test_cond as s
on f.objectid=s.objectid
where s.keys @@ (select dichotomy_split_tsq('侒亩'));

delete from test where objectid=10000;
``