仅查询最早可能日期的数字值

Question

我必须从oracle 11 db查询。 使用下面的查询，我从我的数据库中获得所有最近的TAG_VALUE, TAG_DESC, INSERTION_DATE and PROJECT_ID。

SELECT * 
FROM (SELECT t.tag_value, 
       t.tag_desc, 
       u.update_as_of     AS INSERTION_DATE, 
       p.proj_id       AS PROJECT_ID, 
       Row_number() 
       over( 
        PARTITION BY p.proj_id 
        ORDER BY u.update_as_of DESC) RN 
     FROM project p 
       join update u 
       ON p.project_id = u.project_id 
       join tag t 
       ON t.tag_id = u.tag_id 
     WHERE t.tag_desc LIKE 'Equity%') 
WHERE rn = 1; 

不过，我来到了防空火炮的情况下，我的要求（不按日期排序的话）的回答可以看起来像：

+----------------------------------------------+ 
| TAG_VALUE TAG_DESC INSERTION_DATE PROJECT_ID | 
+----------------------------------------------+ 
| null  Equity 14-DEC-14  1  | 
| 0   Equity 14-DEC-14  1  | 
| 312   Equity 14-DEC-14  1  | 
| 23343  Equity 17-DEC-11  5  | 
| 1263  Equity 16-DEC-11  5  | 
| null  Equity 22-JÄN-14  2  | 
| null  Equity 11-JÄN-14  2  | 
| null  Equity 25-SEPT-13  2  | 
| 0   Equity 20-SEPT-13  2  | 
| 1234  Equity 19-SEPT-13  2  | 
| 13415  Equity 18-SEPT-13  2  | 
| 99999  Equity 16-OCT-10  9  | 
+----------------------------------------------+ 

我Result Set应该看起来像：

+----------------------------------------------+ 
| TAG_VALUE TAG_DESC INSERTION_DATE PROJECT_ID | 
+----------------------------------------------+ 
| 312   Equity 14-DEC-14  1  | 
| 23343  Equity 17-DEC-11  5  | 
| 1234  Equity 19-SEPT-13  2  | 
| 99999  Equity 16-OCT-10  9  | 
+----------------------------------------------+ 

有两种情况，基本上都集中在同一个问题上：

• 正如你所看到的，有两种情况，当插入日期总是相同时project_id = 1。但是，通过我的查询，我仍然得到null，因为ordering。我怎样才能得到312而没有得到null或0值？
• 如果projectID = 2有不同的插入日期，并且较早的日期有TAG_VALUEnull元素。但是，我想要tagValue的| 1234 Equity 19-SEPT-13 2 |，因为它是最新的价值？

如何，我可以基本忽略所有null也0值的值，只需要数字，这是大于0值与最早的日期？

我真的很感谢你的回答！

Answer 1

考虑到您的分组中PROJECT_ID发生了INSERTION_DATE DESC和积极的TAG_VALUE，我调整了分析功能实现的结果。这可能不是一个可靠的解决方案，但肯定会对你有所帮助。

数据设置：

CREATE TABLE Table1 
    ("TAG_VALUE" varchar2(5), "TAG_DESC" varchar2(6), "INSERTION_DATE" varchar2(10), "PROJECT_ID" int) 
; 

INSERT ALL 
    INTO Table1 ("TAG_VALUE", "TAG_DESC", "INSERTION_DATE", "PROJECT_ID") 
     VALUES (NULL, 'Equity', '14-DEC-14', 1) 
    INTO Table1 ("TAG_VALUE", "TAG_DESC", "INSERTION_DATE", "PROJECT_ID") 
     VALUES ('0', 'Equity', '14-DEC-14', 1) 
    INTO Table1 ("TAG_VALUE", "TAG_DESC", "INSERTION_DATE", "PROJECT_ID") 
     VALUES ('312', 'Equity', '14-DEC-14', 1) 
    INTO Table1 ("TAG_VALUE", "TAG_DESC", "INSERTION_DATE", "PROJECT_ID") 
     VALUES ('23343', 'Equity', '17-DEC-11', 5) 
    INTO Table1 ("TAG_VALUE", "TAG_DESC", "INSERTION_DATE", "PROJECT_ID") 
     VALUES ('1263', 'Equity', '16-DEC-11', 5) 
    INTO Table1 ("TAG_VALUE", "TAG_DESC", "INSERTION_DATE", "PROJECT_ID") 
     VALUES (NULL, 'Equity', '22-JÄN-14', 2) 
    INTO Table1 ("TAG_VALUE", "TAG_DESC", "INSERTION_DATE", "PROJECT_ID") 
     VALUES (NULL, 'Equity', '11-JÄN-14', 2) 
    INTO Table1 ("TAG_VALUE", "TAG_DESC", "INSERTION_DATE", "PROJECT_ID") 
     VALUES (NULL, 'Equity', '25-SEPT-13', 2) 
    INTO Table1 ("TAG_VALUE", "TAG_DESC", "INSERTION_DATE", "PROJECT_ID") 
     VALUES ('0', 'Equity', '20-SEPT-13', 2) 
    INTO Table1 ("TAG_VALUE", "TAG_DESC", "INSERTION_DATE", "PROJECT_ID") 
     VALUES ('1234', 'Equity', '19-SEPT-13', 2) 
    INTO Table1 ("TAG_VALUE", "TAG_DESC", "INSERTION_DATE", "PROJECT_ID") 
     VALUES ('13415', 'Equity', '18-SEPT-13', 2) 
    INTO Table1 ("TAG_VALUE", "TAG_DESC", "INSERTION_DATE", "PROJECT_ID") 
     VALUES ('99999', 'Equity', '16-OCT-10', 9) 
SELECT * FROM dual 
; 

查询：

SELECT tag_value, 
     tag_desc, 
     insertion_date, 
     project_id 
FROM (SELECT tag_value, 
       tag_desc, 
       insertion_date, 
       project_id, 
       Last_value(Decode(tag_value, 0, NULL, 
              tag_value) ignore nulls) 
       over ( 
        PARTITION BY project_id 
        ORDER BY insertion_date ROWS BETWEEN unbounded preceding AND 
       unbounded 
       following) new_tag_value 
     FROM table1) 
WHERE tag_value = new_tag_value; 

结果：

TAG_VALUE TAG_DESC INSERTION_DATE PROJECT_ID 
312   Equity  14-DEC-14  1 
1234  Equity  19-SEPT-13  2 
23343  Equity  17-DEC-11  5 
99999  Equity  16-OCT-10  9 

这里是fiddle

Answer 2

您可以从表格的内部查询中选择min（Insertion_Date）和项目ID，并通过Tag_Value <> Null对其进行过滤。然后你在外部查询中，内部将这个内部查询连接到project_id和insertion_date上。

Answer 3

使用案例1：

如果我理解正确的使用情况，您可以按如下通过您的分析函数的排序子句在非空非零tag_value -s的“优先级”做：

ROW_NUMBER() OVER (
    PARTITION BY p.proj_id 
    ORDER BY 
     CASE WHEN t.tag_value > 0 THEN 0 ELSE 1 END ASC, 
     u.update_as_of DESC 
) RN 

这当然会给你在你的输出null -s或零每次没有其他tag_value -s在你的数据分区由p.proj_id标识。

---

使用案例2：

如果你想摆脱零和null -s完全，你必须修改你的（内部）查询where条款改为：

WHERE t.tag_desc LIKE 'Equity%' 
    AND t.tag_value > 0 

Answer 4

你的问题是：“如何，可我基本上忽略所有空，也为0值的值”

简单的答案是：通过删除WHERE子句中的那些记录。

我用AND t.tag_value > 0这里。如果您想允许负值，您可以用AND t.tag_value <> 0 AND t.tag_value IS NOT NULL替换它。

SELECT * 
FROM 
(
    SELECT 
    t.tag_value, 
    t.tag_desc, 
    u.update_as_of AS INSERTION_DATE, 
    p.proj_id AS PROJECT_ID, 
    ROW_NUMBER() OVER(PARTITION BY p.proj_id ORDER BY u.update_as_of DESC) RN 
    FROM updated u 
    JOIN project p ON p.project_id = u.project_id 
    JOIN tag t ON t.tag_id = u.tag_id 
    WHERE t.tag_desc LIKE 'Equity%' AND t.tag_value > 0 
) 
WHERE RN = 1;