Mysql行到列查询

Question

I am hoping this is a relatively simple question to answer. I have a list of transactions in a database table and I want to extract into columns an enum value (y/n) to each motorcycle manufacturer that exists in that table.

I have tried the following query:

SELECT
accCode,
rnFuncBool(accCode,'HON') purchasedHonda,
rnFuncBool(accCode,'YAM') purchasedYamaha,
rnFuncBool(accCode,'KAW') purchasedKawsaki,
rnFuncBool(accCode,'SUZ') purchasedSuzuki,
rnFuncBool(accCode,'DUC') purchasedDucati,
rnFuncBool(accCode,'KTM') purchasedKTM,
rnFuncBool(accCode,'SYM') purchasedSym,
rnFuncBool(accCode,'VIC') purchasedVictory
FROM _emarsys_vehiclessold WHERE accCode<>'' GROUP BY accCode

and the function rnFuncBool is as follows:

DELIMITER $$
USE `phcontacts`$$
DROP FUNCTION IF EXISTS `rnFuncBool`$$
CREATE DEFINER=`root`@`localhost` FUNCTION `rnFuncBool`(
fnAccCode VARCHAR(6),
fnAccMan VARCHAR(3)
) RETURNS VARCHAR(1) CHARSET utf8
BEGIN
IF (SELECT COUNT(*) FROM _emarsys_vehiclessold WHERE accCode=fnAccCode COLLATE utf8_unicode_ci AND vehichleManufacturer=fnAccMan COLLATE utf8_unicode_ci )>0 THEN
    RETURN 'y';
ELSE
    RETURN 'n';
END IF;
END$$

DELIMITER ;

Whilst this seems a really logical solution for a quick return the main table contains over 48,000 rows and so the execution time is in the minutes.

I did try working with a join on a temporary table but that only returned one manufacturer when grouped by the customer account code.

Of course I could try group_concat but that is not what I really want to achieve.

If anybody has any thoughts on how I can achieve this that would be brilliant.

As always many thanks in advance.

Cheers Graham

Answer 1

If you go for boolean 1/0 instead of y/n then you can greatly simplify things for yourself and do a cross-tab problem query which'll be way more efficient...

SELECT
accCode,
SUM(CASE WHEN accCode ='HON' THEN 1 ELSE 0 END) purchasedHonda,
SUM(CASE WHEN accCode='YAM' THEN 1 ELSE 0 END)  purchasedYamaha,
...
SUM(CASE WHEN accCode='VIC' THEN 1 ELSE 0 END)  purchasedVictory
FROM _emarsys_vehiclessold WHERE accCode<>'' 
GROUP BY accCode

That being said, your code and what I've shown an abridged amendment to is grouping by accCode, which is what you've said you want, but I'm not sure how much value there is for you unless you're looking to construct a lookup table or something similar in which case you could simply get a list of distinct accCode and do the rest by hand or in excel just as quickly as writing sql to do it.

Answer 2

Whenever you get "looping" like this in a design, there is frequently a much more efficient design.

It's hard to figure out why you would need a udf, and why that queries the same table, and why that query needs to be performed 8 times for every row in the table (or, more precisely every row where accCode<>''). If 48,000 rows match the criteria in that outer query, that's going to be 384,000 calls to the function, which is going to be a total 384,001 queries executed against the database.

I just can't fathom why you would need to do that, to get the specified resultset.

It would be much more efficient to get the information in just one pass through the table, with a query something like this:

SELECT accCode
     , MAX(IF(vehichleManufacturer='HON','y','n')) AS purchasedHonda
     , MAX(IF(vehichleManufacturer='YAM','y','n')) AS purchasedYamaha
     , MAX(IF(vehichleManufacturer='KAW','y','n')) AS purchasedKawasaki
     , ...
  FROM _emarsys_vehiclessold
 WHERE accCode<>''
 GROUP
    BY accCode

Your function includes a specification that the comparison should be done case insensitive collation. If you need to specify that, it can be done in the SQL

  , MAX(IF(vehichleManufacturer='HON' COLLATE utf8_unicode_ci,'y','n'))

That query is likely to benefit from a covering index on accCode and vehichleManufacturer .

(That's an odd spelling in that vehichleManufacturer column name.)

---

The MySQL-specific IF() function could be replaced with an equivalent ANSI CASE expression, eg:

  , MAX(CASE WHEN vehichleManufacturer='HON' THEN 'y' ELSE 'n' END)

Answer 3

Try:

SELECT accCode,
       case sum(CASE accCode WHEN 'HON' THEN 1 ELSE 0 END)
            when 0 then 'n' else 'y' 
       end purchasedHonda,
       case sum(CASE accCode WHEN 'YAM' THEN 1 ELSE 0 END)
            when 0 then 'n' else 'y' 
       end purchasedYamaha,
       case sum(CASE accCode WHEN 'KAW' THEN 1 ELSE 0 END)
            when 0 then 'n' else 'y' 
       end purchasedKawsaki,
       case sum(CASE accCode WHEN 'SUZ' THEN 1 ELSE 0 END)
            when 0 then 'n' else 'y' 
       end purchasedSuzuki,
       case sum(CASE accCode WHEN 'DUC' THEN 1 ELSE 0 END)
            when 0 then 'n' else 'y' 
       end purchasedDucati,
       case sum(CASE accCode WHEN 'KTM' THEN 1 ELSE 0 END)
            when 0 then 'n' else 'y' 
       end purchasedKTM,
       case sum(CASE accCode WHEN 'SYM' THEN 1 ELSE 0 END)
            when 0 then 'n' else 'y' 
       end purchasedSym,
       case sum(CASE accCode WHEN 'VIC' THEN 1 ELSE 0 END)
            when 0 then 'n' else 'y' 
       end purchasedVictory
FROM _emarsys_vehiclessold WHERE accCode<>'' GROUP BY accCode

Answer 4

Many thanks to Steph for your help on this. To explain the data is being exported to a CSV file for external use and so the system it is being imported into needs a flat file only.

I have modified your query slightly and hey presto it works:

SELECT accCode,
(IF (SUM(CASE WHEN `vehichleManufacturer` ='HON' THEN 1 ELSE 0 END)>0,'y','n')) purchasedHonda,
(IF (SUM(CASE WHEN `vehichleManufacturer` ='KAW' THEN 1 ELSE 0 END)>0,'y','n')) purchasedKawasaki,
(IF (SUM(CASE WHEN `vehichleManufacturer` ='SUZ' THEN 1 ELSE 0 END)>0,'y','n')) purchasedSuzuki,
(IF (SUM(CASE WHEN `vehichleManufacturer` ='DUC' THEN 1 ELSE 0 END)>0,'y','n')) purchasedDucati,
(IF (SUM(CASE WHEN `vehichleManufacturer` ='KTM' THEN 1 ELSE 0 END)>0,'y','n')) purchasedKTM,
(IF (SUM(CASE WHEN `vehichleManufacturer` ='SYM' THEN 1 ELSE 0 END)>0,'y','n')) purchasedSym,
(IF (SUM(CASE WHEN `vehichleManufacturer`='YAM' THEN 1 ELSE 0 END)>0,'y','n'))  purchasedYamaha,
(IF (SUM(CASE WHEN `vehichleManufacturer`='VIC' THEN 1 ELSE 0 END)>0,'y','n'))  purchasedVictory
FROM _emarsys_vehiclessold WHERE accCode<>'' 
GROUP BY accCode

Many thanks Graham