Question

问题：

我有一个名为item_tag_assn的表，它使用标签映射项目（多对多关联表）。我需要找出应用了一组标签的项目。例如，如果我的表有以下数据：

 item_id | tag_id 
------------------
     205 | 110
     206 | 120
     207 | 130
     205 | 130
     206 | 147
     210 | 110
     205 | 152
     209 | 111
     210 | 177
     205 | 147
     212 | 110
     212 | 135
     205 | 135
     212 | 147
------------------

如果我正在搜索

带有标签110,135和147的项目然后我期望结果集中的项目＃205和＃212。
带有代码110,130,135,147和152的商品，那么我应该只获得商品＃205，因为只有＃205有与其相关的所有商品。

环境：

PostgreSQL 9.5
我不允许在此表中添加第三列或完全创建新表。

到目前为止的进展：

我找到了这样的解决方案：

SELECT DISTINCT ita1.item_id 
FROM 
  item_tag_assn AS ita1 
  LEFT JOIN 
    item_tag_assn AS ita2 ON ita1.item_id = ita2.item_id 
  LEFT JOIN 
    item_tag_assn AS ita3 ON ita2.item_id = ita3.item_id 
GROUP BY ita1.item_id 
HAVING 
  sum((ita1.tag_id = 110 and ita2.tag_id = 135 and ita3.tag_id = 147)::integer) >= 1

和可行。

需要优化

关联表相当大。加入它本身是昂贵的并且速度慢，而且它的可扩展性也不高。我认为窗口功能可以帮助但我不知道如何使用它们。

有没有更好的方法来解决问题？

Answer 1

如果我理解正确，你需要这样的东西：

WITH search AS (
    SELECT '{110,130,135,147,152}'::int4[] as search
), searched AS (
    SELECT DISTINCT item_id,
           tag_id
      FROM item_tag_assn
      JOIN search ON (tag_id) = ANY(search)
  ORDER BY 1, 2
), aggregated AS (
    SELECT item_id,
           array_agg(tag_id) AS agg
      FROM searched
  GROUP BY 1
)
SELECT *
  FROM aggregated, search
 WHERE agg = search
;

search - 设置搜索数组（数组必须预先排序）。 searched - 所有行都没有搜索过标记 aggregated - 聚合在数组tag_id中，每个item_id

您可以将agg = search更改为agg @> search，然后在searched中不需要预先排序和ORDER BY。

从问题中添加数据集：

WITH item_tag_assn AS (
      SELECT   205 as item_id, 110 as tag_id
      UNION SELECT     206 , 120
      UNION SELECT     207 , 130
      UNION SELECT     205 , 130
      UNION SELECT     206 , 147
      UNION SELECT     210 , 110
      UNION SELECT     205 , 152
      UNION SELECT     209 , 111
      UNION SELECT     210 , 177
      UNION SELECT     205 , 147
      UNION SELECT     212 , 110
      UNION SELECT     212 , 135
      UNION SELECT     205 , 135
      UNION SELECT     212 , 147
),search AS (
    SELECT '{110,130,135,147,152}'::int4[] as search
), searched AS (
    SELECT DISTINCT item_id,
           tag_id
      FROM item_tag_assn
      JOIN search ON (tag_id) = ANY(search)
  ORDER BY 1, 2
), aggregated AS (
    SELECT item_id,
           array_agg(tag_id) AS agg
      FROM searched
  GROUP BY 1
)
SELECT *
  FROM aggregated, search
 WHERE agg = search
;

结果：

 item_id |          agg          |        search         
---------+-----------------------+-----------------------
     205 | {110,130,135,147,152} | {110,130,135,147,152}
(1 row)

如果将搜索更改为'{110,135,147}'：

 item_id |      agg      |    search     
---------+---------------+---------------
     212 | {110,135,147} | {110,135,147}
     205 | {110,135,147} | {110,135,147}
(2 rows)

要在产品上运行，您需要创建索引CREATE INDEX ON item_tag_assn (tag_id);

                                                               QUERY PLAN                                                               
----------------------------------------------------------------------------------------------------------------------------------------
 Hash Join  (cost=32.72..35.34 rows=1 width=68) (actual time=0.055..0.059 rows=3 loops=1)
   Hash Cond: (aggregated.agg = search.search)
   CTE search
     ->  Result  (cost=0.00..0.01 rows=1 width=32) (actual time=0.001..0.001 rows=1 loops=1)
   CTE searched
     ->  Unique  (cost=27.73..28.55 rows=110 width=8) (actual time=0.029..0.031 rows=3 loops=1)
           ->  Sort  (cost=27.73..28.00 rows=110 width=8) (actual time=0.029..0.029 rows=3 loops=1)
                 Sort Key: x.item_id, x.tag_id
                 Sort Method: quicksort  Memory: 25kB
                 ->  Nested Loop  (cost=10.40..24.00 rows=110 width=8) (actual time=0.013..0.014 rows=3 loops=1)
                       ->  CTE Scan on search search_1  (cost=0.00..0.02 rows=1 width=32) (actual time=0.002..0.002 rows=1 loops=1)
                       ->  Bitmap Heap Scan on x  (cost=10.40..22.88 rows=110 width=8) (actual time=0.009..0.009 rows=3 loops=1)
                             Recheck Cond: (tag_id = ANY (search_1.search))
                             Heap Blocks: exact=1
                             ->  Bitmap Index Scan on i1  (cost=0.00..10.38 rows=110 width=0) (actual time=0.002..0.002 rows=3 loops=1)
                                   Index Cond: (tag_id = ANY (search_1.search))
   CTE aggregated
     ->  HashAggregate  (cost=2.75..4.12 rows=110 width=36) (actual time=0.038..0.039 rows=3 loops=1)
           Group Key: searched.item_id
           ->  CTE Scan on searched  (cost=0.00..2.20 rows=110 width=8) (actual time=0.029..0.031 rows=3 loops=1)
   ->  CTE Scan on aggregated  (cost=0.00..2.20 rows=110 width=36) (actual time=0.040..0.043 rows=3 loops=1)
   ->  Hash  (cost=0.02..0.02 rows=1 width=32) (actual time=0.005..0.005 rows=1 loops=1)
         Buckets: 1024  Batches: 1  Memory Usage: 9kB
         ->  CTE Scan on search  (cost=0.00..0.02 rows=1 width=32) (actual time=0.000..0.001 rows=1 loops=1)
 Planning time: 0.309 ms
 Execution time: 0.115 ms

如何找到包含多个标签的商品？

问题：

环境：

到目前为止的进展：

需要优化

1 个答案: