MySQL数据透视/交叉表查询

Question

Question 1: I have a table with the below structure and data:

app_id  transaction_id  mobile_no   node_id  customer_attribute  entered_value 
100     111             9999999999  1        Q1                  2                             
100     111             9999999999  2        Q2                  1                             
100     111             9999999999  3        Q3                  4                             
100     111             9999999999  4        Q4                  3                             
100     111             9999999999  5        Q5                  2                             
100     222             8888888888  4        Q4                  1                             
100     222             8888888888  3        Q3                  2                             
100     222             8888888888  2        Q2                  1                             
100     222             8888888888  1        Q1                  3                             
100     222             8888888888  5        Q5                  4                             

I want to display these records in the below format:

app_id  |  transaction_id  | mobile     |  Q1  |  Q2  |  Q3  |  Q4 |  Q5  |
 100    |      111         | 9999999999 |   2  |   1  |   4  |  3  |  2   |
 100    |      222         | 8888888888 |   3  |   1  |   2  |  1  |  4   |

I know I need to use crosstab/pivot query to get this display. For this I tried it based on the limited knowledge that I have about it. Following is my query:

SELECT app_id, transaction_id, mobile_no,
  (CASE node_id WHEN 1 THEN entered_value ELSE '' END) AS user_input1,
  (CASE node_id WHEN 2 THEN entered_value ELSE '' END) AS user_input2,
  (CASE node_id WHEN 3 THEN entered_value ELSE '' END) AS user_input3,
  (CASE node_id WHEN 4 THEN entered_value ELSE '' END) AS user_input4,
  (CASE node_id WHEN 5 THEN entered_value ELSE '' END) AS user_input5
FROM trn_user_log 
GROUP BY app_id, transaction_id, mobile_no, node_id

And based on this query I got the below display:

app_id  transaction_id  mobile_no   user_input1  user_input2  user_input3  user_input4  user_input5  
100     111             9999999999  2                                                                
100     111             9999999999               1                                                   
100     111             9999999999                            4                                      
100     111             9999999999                                         3                         
100     111             9999999999                                                      2            
100     222             8888888888  3                                                                
100     222             8888888888               1                                                   
100     222             8888888888                            2                                      
100     222             8888888888                                         1                         
100     222             8888888888                                                      4            

Can anyone help me with the proper changes that I need to make to my query to get the records in one single row and not multiple rows as above.

Question 2: Also is there a way to get the value of a particular field as the NAME of the column. As you can see above I have user_input1 , user_input2 ,... as the header. Instead of that I want to have the values in customer_attribute as the header of the columns.

For this I checked NAME_CONST(name,value) as below:

SELECT app_id, transaction_id, mobile_no,
NAME_CONST(customer_attribute, (CASE node_id WHEN 1 THEN entered_value ELSE '' END))
FROM trn_user_log 

But it gives an error

Error Code : 1210 Incorrect arguments to NAME_CONST

Help required.

Answer 1

While @John's static answer works great, if you have an unknown number of columns that you want to transform, I would consider using prepared statements to get the results:

SET @sql = NULL;
SELECT
  GROUP_CONCAT(DISTINCT
    CONCAT(
      'GROUP_CONCAT((CASE node_id when ',
      node_id,
      ' then entered_value else NULL END)) AS user_input',
      node_id
    )
  ) INTO @sql
FROM trn_user_log;


SET @sql = CONCAT('SELECT app_id, transaction_id, mobile_no, ', @sql, ' 
                  FROM trn_user_log 
                  GROUP BY app_id, transaction_id, mobile_no');

PREPARE stmt FROM @sql;
EXECUTE stmt;
DEALLOCATE PREPARE stmt;

see SQL Fiddle with Demo

As far as your second, please clarify what you are trying to do it is not clear.

Answer 2

Add GROUP_CONCAT in your CASE clause

SELECT app_id, transaction_id, mobile_no,
  GROUP_CONCAT((CASE node_id WHEN 1 THEN entered_value ELSE NULL END)) AS user_input1,
  GROUP_CONCAT((CASE node_id WHEN 2 THEN entered_value ELSE NULL END)) AS user_input2,
  GROUP_CONCAT((CASE node_id WHEN 3 THEN entered_value ELSE NULL END)) AS user_input3,
  GROUP_CONCAT((CASE node_id WHEN 4 THEN entered_value ELSE NULL END)) AS user_input4,
  GROUP_CONCAT((CASE node_id WHEN 5 THEN entered_value ELSE NULL END)) AS user_input5
FROM trn_user_log 
GROUP BY app_id, transaction_id, mobile_no

SQLFiddle Demo

Answer 3

@DarkKnightFan, this was a very helpful question for a task I was working. I went ahead and modified the solution from @bluefin to solve your second question. The following code produces your originally requested format with the value of customer_attribute as the resulting column headings in the cross-tab.

The relevant change was to change:

' then entered_value else NULL END)) AS user_input',
      node_id

To this:

' then entered_value else NULL END)) AS ''',
          customer_attribute,''''

The full code:

SET @sql = NULL;
SELECT
  GROUP_CONCAT(DISTINCT
    CONCAT(
      'GROUP_CONCAT((CASE node_id when ',
      node_id,
      ' then entered_value else NULL END)) AS ''',
      customer_attribute,''''
    )
  ) INTO @sql
FROM trn_user_log;


SET @sql = CONCAT('SELECT app_id, transaction_id, mobile_no, ', @sql, ' 
                  FROM trn_user_log 
                  GROUP BY app_id, transaction_id, mobile_no');

PREPARE stmt FROM @sql;
EXECUTE stmt;
DEALLOCATE PREPARE stmt;

Also, for other users browsing this issue, if you have a lot of values that you are trying to cross-tab, you may run into error because GROUP_CONCAT() has a default max length of 1024 characters. To increase put this at the start of your prepared statement:

SET SESSION group_concat_max_len = value; -- replace value with an int