Hive函数：rank()、dense_rank()

数据准备：

G1,KING,5000
G1,BING,5000
G2,FING,5000
G1,FORD,3000
G2,SCOTT,3000
G1,JONES,2975
G2,BLAKE,2850
G1,CLARK,2450
G1,ALLEN,1600
G1,CELL1,NULL
G2,CELL2,NULL

CREATE EXTERNAL TABLE test_data (
gid string,
ename STRING,
sal INT
) ROW FORMAT DELIMITED
FIELDS TERMINATED BY ','
stored as textfile location '/user/jc_rc_ftp/test_data';

select * from test_data l;
+--------+----------+--------+--+
| l.gid  | l.ename  | l.sal  |
+--------+----------+--------+--+
| G1     | CELL1    | NULL   |
| G2     | CELL2    | NULL   |
| G1     | KING     | 5000   |
| G1     | BING     | 5000   |
| G2     | FING     | 5000   |
| G1     | FORD     | 3000   |
| G2     | SCOTT    | 3000   |
| G1     | JONES    | 2975   |
| G2     | BLAKE    | 2850   |
| G1     | CLARK    | 2450   |
| G1     | ALLEN    | 1600   |
+--------+----------+--------+--+

rank

---rank()over(order by 列名排序)的结果是不连续的，如果有4个人，其中有3个是并列第1名，那么最后的排序结果结果如：1 1 1 4

SQL> select gid,ename,sal,rank() over(order by sal desc) from test_data;

---rank() over (partition by 分组字段 order by 排序字段 顺序)

注意：使用rank()over(order by 排序字段 顺序)排序的时候，空值是最大的

select gid,ename,sal,rank()over(partition by gid order by coalesce(sal,0) desc) from test_data;
+------+--------+-------+----------------+--+
| gid | ename | sal | rank_window_0 |
+------+--------+-------+----------------+--+
| G1 | BING | 5000 | 1 |
| G1 | KING | 5000 | 1 |
| G1 | FORD | 3000 | 3 |
| G1 | JONES | 2975 | 4 |
| G1 | CLARK | 2450 | 5 |
| G1 | ALLEN | 1600 | 6 |
| G1 | CELL1 | NULL | 7 |
| G2 | FING | 5000 | 1 |
| G2 | SCOTT | 3000 | 2 |
| G2 | BLAKE | 2850 | 3 |
| G2 | CELL2 | NULL | 4 |
+------+--------+-------+----------------+--+

dense_rank

---dense_rank()over(order by 列名排序)的结果是连续的，如果有4个人，其中有3个是并列第1名，那么最后的排序结果如：1 1 1 2
如果排序字段为null，可能造成在排序时将null字段排在最前面，影响排序的正确性。
所以建议将 dense_rank()over(order by 列名 排序)改为dense_rank()over(order by 列名排序 nulls last)

---rank()over(order by 列名排序)的结果是不连续的，如果有4个人，其中有3个是并列第1名，那么最后的排序结果结果如：1 1 1 4
SQL> select gid,ename,sal,rank() over(order by sal desc) from test_data;
+------+--------+-------+----------------+--+
| gid  | ename  |  sal  | rank_window_0  |
+------+--------+-------+----------------+--+
| G2   | FING   | 5000  | 1              |
| G1   | BING   | 5000  | 1              |
| G1   | KING   | 5000  | 1              |
| G2   | SCOTT  | 3000  | 4              |
| G1   | FORD   | 3000  | 4              |
| G1   | JONES  | 2975  | 6              |
| G2   | BLAKE  | 2850  | 7              |
| G1   | CLARK  | 2450  | 8              |
| G1   | ALLEN  | 1600  | 9              |
| G1   | CELL1  | NULL  | 10             |
| G2   | CELL2  | NULL  | 10             |
+------+--------+-------+----------------+--+

---rank() over (partition by 分组字段 order by 排序字段 顺序)
注意：使用rank()over(order by 排序字段 顺序)排序的时候，空值是最大的
select gid,ename,sal,rank()over(partition by gid order by coalesce(sal,0) desc) from test_data;
+------+--------+-------+----------------+--+
| gid  | ename  |  sal  | rank_window_0  |
+------+--------+-------+----------------+--+
| G1   | BING   | 5000  | 1              |
| G1   | KING   | 5000  | 1              |
| G1   | FORD   | 3000  | 3              |
| G1   | JONES  | 2975  | 4              |
| G1   | CLARK  | 2450  | 5              |
| G1   | ALLEN  | 1600  | 6              |
| G1   | CELL1  | NULL  | 7              |
| G2   | FING   | 5000  | 1              |
| G2   | SCOTT  | 3000  | 2              |
| G2   | BLAKE  | 2850  | 3              |
| G2   | CELL2  | NULL  | 4              |
+------+--------+-------+----------------+--+

---dense_rank()over(order by 列名排序)的结果是连续的，如果有4个人，其中有3个是并列第1名，那么最后的排序结果如：1 1 1 2
如果排序字段为null，可能造成在排序时将null字段排在最前面，影响排序的正确性。
所以建议将 dense_rank()over(order by 列名 排序)改为dense_rank()over(order by 列名排序 nulls last)

select gid,ename,sal,dense_rank()over(order by sal desc) from test_data;
+------+--------+-------+----------------------+--+
| gid  | ename  |  sal  | dense_rank_window_0  |
+------+--------+-------+----------------------+--+
| G2   | FING   | 5000  | 1                    |
| G1   | BING   | 5000  | 1                    |
| G1   | KING   | 5000  | 1                    |
| G2   | SCOTT  | 3000  | 2                    |
| G1   | FORD   | 3000  | 2                    |
| G1   | JONES  | 2975  | 3                    |
| G2   | BLAKE  | 2850  | 4                    |
| G1   | CLARK  | 2450  | 5                    |
| G1   | ALLEN  | 1600  | 6                    |
| G2   | CELL2  | NULL  | 7                    |
| G1   | CELL1  | NULL  | 7                    |
+------+--------+-------+----------------------+--+
select gid,ename,sal,dense_rank()over(partition by gid order by sal desc) from test_data;
+------+--------+-------+----------------------+--+
| gid  | ename  |  sal  | dense_rank_window_0  |
+------+--------+-------+----------------------+--+
| G1   | BING   | 5000  | 1                    |
| G1   | KING   | 5000  | 1                    |
| G1   | FORD   | 3000  | 2                    |
| G1   | JONES  | 2975  | 3                    |
| G1   | CLARK  | 2450  | 4                    |
| G1   | ALLEN  | 1600  | 5                    |
| G1   | CELL1  | NULL  | 6                    |
| G2   | FING   | 5000  | 1                    |
| G2   | SCOTT  | 3000  | 2                    |
| G2   | BLAKE  | 2850  | 3                    |
| G2   | CELL2  | NULL  | 4                    |
+------+--------+-------+----------------------+--+