MySQL计数，左连接，group by返回零行

Question

In the below sql statement:

 SELECT `keywords`.keyID, count(`keywords-occurencies`.keyID) as countOccurencies 
                    FROM `keywords-occurencies`  
                    LEFT JOIN `keywords` 
                    ON `keywords-occurencies`.keyID = `keywords`.keyID 
                    WHERE `keywords-occurencies`.`keyID` IN (1,2,3) AND date BETWEEN '2013/01/25' AND '2013/01/27'
                    GROUP BY `keywords`.`keyID`

If keyID 3 has no return values it is not counted as 0 and it is not included in the result set and a result like this is displayed

keyID countOccurencies
1       3
3       5

I would like to display the zero results like

keyID countOccurencies
1       3
2       0
3       5

Sample data to test with:

--
-- Table structure for table `keywords`
--

CREATE TABLE IF NOT EXISTS `keywords` (
  `keyID` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `keyName` varchar(40) NOT NULL,
  PRIMARY KEY (`keyID`)
) ENGINE=InnoDB  DEFAULT CHARSET=latin1 AUTO_INCREMENT=3 ;

--
-- Dumping data for table `keywords`
--

INSERT INTO `keywords` (`keyID`, `keyName`) VALUES
(1, 'testKey1'),
(2, 'testKey2');

-- --------------------------------------------------------

--
-- Table structure for table `keywords-occurencies`
--

CREATE TABLE IF NOT EXISTS `keywords-occurencies` (
  `occurencyID` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `keyID` int(10) unsigned NOT NULL,
  `date` date NOT NULL,
  PRIMARY KEY (`occurencyID`),
  KEY `keyID` (`keyID`)
) ENGINE=InnoDB  DEFAULT CHARSET=latin1 AUTO_INCREMENT=3 ;

--
-- Dumping data for table `keywords-occurencies`
--

INSERT INTO `keywords-occurencies` (`occurencyID`, `keyID`, `date`) VALUES
(1, 1, '2013-01-27'),
(2, 1, '2013-01-26');

--
-- Constraints for table `keywords-occurencies`
--
ALTER TABLE `keywords-occurencies`
  ADD CONSTRAINT `keywords@002doccurencies_ibfk_1` FOREIGN KEY (`keyID`) REFERENCES `keywords` (`keyID`) ON DELETE CASCADE ON UPDATE CASCADE;

Answer 1

things to do

• you should be grouping it with GROUP BY keywords-occurencies.keyID
• and you must be displaying keywords-occurencies.keyID not the keywords.keyID
• COUNT keywords.keyID
• ( optional ) use ALIAS so you can get rid of the backticks other than tableNames

query,

 
 
 
  
  SELECT a.keyID, count(b.keyID) AS countOccurencies FROM `keywords - occurencies` a LEFT JOIN `keywords` b ON a.keyID = b.keyID WHERE a.keyID IN ( 1, 2, 3 ) AND DATE BETWEEN '2013/01/25' AND '2013/01/27' GROUP BY a.keyID
 
  

UPDATE 1

Based on the example records, you need to do the following,

• interchange the tableNames
• put this condition DATE BETWEEN '2013-01-25' AND '2013-01-27' on the ON clause of join.
• ( optional ) use ALIAS so you can get rid of the backticks other than tableNames

query,

 SELECT a.keyID, count(b.keyID) AS countOccurencies FROM `keywords` a LEFT JOIN `keywords-occurencies` b ON a.keyID = b.keyID AND b.DATE BETWEEN '2013-01-25' AND '2013-01-27' WHERE a.keyID IN ( 1, 2, 3 ) GROUP BY a.keyID 

• SQLFiddle Demo

Answer 2

Put a back tick on date ::

SELECT 
`keywords`.keyID, 
count(`keywords-occurencies`.keyID) as countOccurencies 
FROM `keywords-occurencies`  
LEFT JOIN `keywords`  ON `keywords-occurencies`.keyID = `keywords`.keyID 
WHERE `keywords-occurencies`.`keyID` IN (1,2,3) AND `date` BETWEEN '2013/01/25' AND '2013/01/27'
GROUP BY `keywords-occurencies`.`keyID`

Answer 3

There are two things. You need to group by the id on the first part of the left outer join. You then need to count what is one the second side. For a right outer join, the order is the opposite:

SELECT k.keyID, count(ko.keyID) as countOccurencies 
FROM `keywords-occurencies` ko
      RIGHT JOIN `keywords` k
      ON ko.keyID = k.keyID 
WHERE k.`keyID` IN (1,2,3) AND date BETWEEN '2013/01/25' AND '2013/01/27'
GROUP BY k.`keyID`

The reason for this has to do with the left outer join. It keeps everything in the first table, even when there are no matches. So, that is where you get the complete list. As for the count, you want to count matches. If you count the id in the first table, you will always get at least 1. Counting the id in the second table allows you to get 0.

Notice I also added aliases to your table to make the query more readable.

Answer 4

Changing that to LEFT JOIN worked for me:

SELECT k.keyID, count(ko.keyID) as countOccurencies 
FROM `keywords-occurencies` ko
      LEFT JOIN `keywords` k
      ON ko.keyID = k.keyID 
WHERE k.`keyID` IN (1,2,3) AND date BETWEEN '2013/01/25' AND '2013/01/27'
GROUP BY k.`keyID`