使用PostgreSQL数据库建立用户画像系统的方法

CREATE TABLE user_tag(uid serial primary key, tag jsonb);

CREATE OR REPLACE FUNCTioN f_random_attr(attr text[], max_attr int)
RETURNS text[] AS $$
DECLARE
     i integer := 0; 
     r integer := 0;
     res  text[];
     v text;
     l integer;
     num integer;
 BEGIN
     num := (random()*max_attr)::int;
     IF num < 1 THEN
         num := 1;
     END IF;
     l := array_length(attr, 1);
     WHILE i < num LOOP
          r := round(random()*l)::int + 1;
          v := attr[r];
          IF res @> array[v] THEN
             continue;
          ELSE 
             res := array_append(res, v);
             i := i + 1;
          END IF;
     END LOOP;
     return res;
 END;
$$ LANGUAGE plpgsql;

INSERT INTO user_tag(uid, tag) 
SELECT seq, 
       json_build_object(
         '职业',
         f_random_attr(array['农民','工人','IT工程师','理发师','医生','老师','美工','律师','公务员','官员'], 1),
         '爱好',
         f_random_attr(array['游泳','乒乓球','羽毛球','网球','爬山','高尔夫球','滑雪','爬山','旅游'], 5),
         '学历',
         f_random_attr(array['无学历','小学','初中','高中','中专','专科','本科','硕士','博士'], 1),
         '性格',
         f_random_attr(array['外向','内向','谨慎','稳重','细心','粗心','浮躁','自信'], 3))::jsonb
  FROM generate_series(1, 10000) as t(seq);

CREATE INDEX idx_user_tag_tag on user_tag using gin(tag);

osdba=# select * from user_tag where tag @> '{"性格":["外向","细心"]}' and tag @> '{"职业":["老师"]}';
 uid  |                                                         tag
------+----------------------------------------------------------------------------------------------------------------------
  881 | {"学历": ["中专"], "性格": ["细心", "内向", "外向"], "爱好": ["高尔夫球"], "职业": ["老师"]}
 1031 | {"学历": [null], "性格": ["外向", "细心"], "爱好": ["爬山"], "职业": ["老师"]}
 3313 | {"学历": [null], "性格": ["外向", "细心"], "爱好": ["乒乓球"], "职业": ["老师"]}
 4053 | {"学历": ["本科"], "性格": ["细心", "外向"], "爱好": ["爬山", "滑雪", "游泳"], "职业": ["老师"]}
 4085 | {"学历": ["初中"], "性格": ["外向", "细心"], "爱好": ["网球"], "职业": ["老师"]}
 4332 | {"学历": ["硕士"], "性格": ["外向", "细心"], "爱好": ["网球", "羽毛球"], "职业": ["老师"]}
 4997 | {"学历": ["小学"], "性格": ["外向", "细心", "浮躁"], "爱好": ["乒乓球"], "职业": ["老师"]}
 5231 | {"学历": ["本科"], "性格": ["外向", "细心"], "爱好": ["高尔夫球"], "职业": ["老师"]}
 5360 | {"学历": ["无学历"], "性格": ["浮躁", "外向", "细心"], "爱好": ["爬山", "网球", "旅游", "羽毛球"], "职业": ["老师"]}
 6281 | {"学历": ["专科"], "性格": ["细心", "外向", "自信"], "爱好": ["滑雪", null], "职业": ["老师"]}
 7681 | {"学历": ["小学"], "性格": ["粗心", "外向", "细心"], "爱好": ["乒乓球"], "职业": ["老师"]}
 8246 | {"学历": ["硕士"], "性格": ["外向", "细心"], "爱好": ["滑雪", "网球", "高尔夫球"], "职业": ["老师"]}
 8531 | {"学历": ["硕士"], "性格": ["细心", "外向", "粗心"], "爱好": ["滑雪", "爬山"], "职业": ["老师"]}
 8618 | {"学历": ["小学"], "性格": ["细心", "外向", "浮躁"], "爱好": ["乒乓球"], "职业": ["老师"]}
 9508 | {"学历": ["小学"], "性格": ["浮躁", "外向", "细心"], "爱好": ["爬山", "旅游", "高尔夫球"], "职业": ["老师"]}
(15 rows)

Time: 1.495 ms

osdba=# select * from user_tag where tag @> '{"性格":["外向","细心"]}' and tag @> '{"职业":["医生"]}' and tag @>'{"爱好":["滑雪", "游泳"]}';
 uid  |                                               tag
------+--------------------------------------------------------------------------------------------------
 4469 | {"学历": ["小学"], "性格": ["外向", "细心", "稳重"], "爱好": ["滑雪", "游泳"], "职业": ["医生"]}
(1 row)

Time: 2.139 ms

osdba=# explain select * from user_tag where tag @> '{"性格":["外向","细心"]}' and tag @> '{"职业":["医生"]}' and tag @>'{"爱好":["滑雪", "游泳"]}';
                                                                          QUERY PLAN
--------------------------------------------------------------------------------------------------------------------------------------------------------------
 Bitmap Heap Scan on user_tag  (cost=48.00..52.02 rows=1 width=153)
   Recheck Cond: ((tag @> '{"性格": ["外向", "细心"]}'::jsonb) AND (tag @> '{"职业": ["医生"]}'::jsonb) AND (tag @> '{"爱好": ["滑雪", "游泳"]}'::jsonb))
   ->  Bitmap Index Scan on idx_user_tag_tag  (cost=0.00..48.00 rows=1 width=0)
         Index Cond: ((tag @> '{"性格": ["外向", "细心"]}'::jsonb) AND (tag @> '{"职业": ["医生"]}'::jsonb) AND (tag @> '{"爱好": ["滑雪", "游泳"]}'::jsonb))
(4 rows)

Time: 1.736 ms

osdba=# select pg_relation_size('idx_user_tag_tag');
 pg_relation_size
------------------
           245760
(1 row)

Time: 0.522 ms
osdba=# CREATE INDEX idx_user_tag_tag2 on user_tag using gin(tag jsonb_path_ops);
CREATE INDEX
Time: 46.947 ms
osdba=# select pg_relation_size('idx_user_tag_tag2');
 pg_relation_size
------------------
           147456
(1 row)

osdba=# drop index idx_user_tag_tag;
DROP INDEX
Time: 2.833 ms
osdba=# select * from user_tag where tag @> '{"性格":["外向","细心"]}' and tag @> '{"职业":["医生"]}' and tag @>'{"爱好":["滑雪", "游泳"]}';
 uid  |                                               tag
------+--------------------------------------------------------------------------------------------------
 4469 | {"学历": ["小学"], "性格": ["外向", "细心", "稳重"], "爱好": ["滑雪", "游泳"], "职业": ["医生"]}
(1 row)

Time: 1.401 ms
osdba=# explain select * from user_tag where tag @> '{"性格":["外向","细心"]}' and tag @> '{"职业":["医生"]}' and tag @>'{"爱好":["滑雪", "游泳"]}';
                                                                          QUERY PLAN
--------------------------------------------------------------------------------------------------------------------------------------------------------------
 Bitmap Heap Scan on user_tag  (cost=24.00..28.02 rows=1 width=153)
   Recheck Cond: ((tag @> '{"性格": ["外向", "细心"]}'::jsonb) AND (tag @> '{"职业": ["医生"]}'::jsonb) AND (tag @> '{"爱好": ["滑雪", "游泳"]}'::jsonb))
   ->  Bitmap Index Scan on idx_user_tag_tag2  (cost=0.00..24.00 rows=1 width=0)
         Index Cond: ((tag @> '{"性格": ["外向", "细心"]}'::jsonb) AND (tag @> '{"职业": ["医生"]}'::jsonb) AND (tag @> '{"爱好": ["滑雪", "游泳"]}'::jsonb))
(4 rows)

Time: 0.634 ms