如何ignorre來自相同組連續記錄與SQL

Question

，我有以下數據：

ItemID | ProdID | ItemTypeID 
=======+========+=========== 
1001 | 100 | A 
1002 | 100 | B 
1003 | 100 | A 
1004 | 100 | B 
1005 | 100 | B <- successive itemtype (should be excluded) 
1006 | 100 | C 
1007 | 200 | C 
1008 | 200 | A 

，我想每個組內，但沒有相同的ID連續列出ItemTypeIDs（LISTAGG）。所以我的結果應該是這樣的：

ProdID | ItemTypes 
=======+========== 
100 | A,B,A,B,C <- not A,B,A,B,B,C (successive B) 
200 | C,A 

Answer 1

這很棘手。 Listagg()甚至不允許distinct，所以所有的工作都需要在子查詢中完成。

您可以使用行號方法的差異來識別連續的ItemTypeId。然後，一旦羣體被識別，則可以通過組聚集和然後做listagg()：

select ProdId, 
     listagg(ItemId, ',') within group (order by seqnum) as items 
from (select ProdId, ItemId, count(*) as NumItems, 
      row_number() over (partition by ProdId order by min(ItemId)) as seqnum 
     from (select t.*, 
        (row_number() over (partition by ProdId order by ItemId) - 
        row_number() over (partition by ProdId, ItemTypeId order by ItemId) 
        ) as grp 
      from t 
      ) t 
     group by ProdId, ItemTypeId, grp 
    ) t 
group by ProdId; 

編輯：

另一種方式來處理這個使用lag()尋找到新的羣體開始。以前的方法可以讓你獲得計數。這種方法可能會更容易理解：

select ProdId, 
     listagg(ItemId, ',') within group (order by ItemId) as items 
from (select t.* 
     from (select t.*, 
        lag(ItemTypeId) over (partition by ProdId order by ItemId) as prev_ItemTypeId 
      from t 
      ) t 
     where prev_ItemTypeId is null or prev_ItemTypeId <> ItemTypeId 
    ) t 
group by ProdId;