查找条件满足x在mysql / sql中的次数的记录

Question

I'm struggling with what is a complicated SQL query for me, although I believe that it is not particularly complicated. I'm close to the right answer, but not quite there yet.

My database represents a criminal abstract. I have three tables in my database (I've simplified my schema enormously for the purposes of this question): arrest, arrestcharges, and dispositions.

Each defendant can have multiple arrests (defendant table not included for simplification). Each arrest can have multiple charges, which are in the arrestcharges table. And each charge has a grade and is associated with a disposition (guilty, not guilty, etc...). The dispositions are categorized so that 0=some form of guilt disposition, 1=a non-guilty disposition.

I want to find individuals who have been convicted of a charge graded as "M1" on more than one case. If an individual has been convicted of more than M1, but they are in the same case, that person shouldn't be returned (unless they have another case with an M1 conviction).

A sqlfiddle link and the SQL to create and populate the table is below.

I believe that this query should work, but it doesn't:

select a.defendantid, count(a.id)  
    FROM `arrest` AS a LEFT JOIN `arrestcharges` AS ac 
          ON a.id=ac.arrestid LEFT JOIN `dispositions` AS d 
          ON ac.dispositionid=d.id 
    WHERE d.dispocategory=0 AND ac.grade="M1"  
        GROUP BY a.id HAVING COUNT(a.id) > 1 ORDER BY a.defendantid;

Based on the sql below, I expect that defendant IDs 1 and 5 should be returned since they are the only two defendants with an M1 conviction in more than one arrest. But the actual response I am getting is 2 and 5. 2 should not be returned b/c defendant 2 only has one arrest in the database.

Any thoughts on what I am doing wrong?

SQLFiddle

CREATE TABLE IF NOT EXISTS `arrest` (
  `id` int(6) unsigned NOT NULL,
  `defendantid` int(6) unsigned NOT NULL,
  `docketno` varchar(21) NOT NULL,
  PRIMARY KEY (`id`)
) DEFAULT CHARSET=utf8;

CREATE TABLE IF NOT EXISTS `arrestcharges` (
  `id` int(6) unsigned NOT NULL,
  `arrestid` int(6) unsigned NOT NULL,
  `grade` varchar(2) NOT NULL,
  `dispositionid` int(6) NOT NULL,
  PRIMARY KEY (`id`)
) DEFAULT CHARSET=utf8;

CREATE TABLE IF NOT EXISTS `dispositions` (
  `id` int(6) unsigned NOT NULL,
  `disposition` varchar(30) NOT NULL,
  `dispoCategory` int(1) unsigned NOT NULL,
  PRIMARY KEY (`id`)
) DEFAULT CHARSET=utf8;

INSERT INTO `arrest` (`id`, `defendantid`, `docketno`) VALUES
  ('1', '1', 'MC-51-CR-0000222-1999'),
  ('2', '1', 'MC-51-CR-0000223-1999'),
  ('3', '1', 'MC-51-CR-0000224-1999'),
  ('4', '2', 'MC-51-CR-0002343-2000'),
  ('5', '3', 'MC-51-CR-0002349-2000'),
  ('6', '3', 'MC-51-CR-0002350-2000'),
  ('7', '3', 'MC-51-CR-0002351-2010'),
  ('8', '3', 'MC-51-CR-0002352-2013'),
  ('9', '4', 'MC-51-CR-1209293-2011'),
  ('10', '5', 'MC-51-CR-2389848-1999'),
  ('11', '5', 'MC-51-CR-3893923-1999'),
  ('12', '5', 'MC-51-CR-2393912-1999');

INSERT INTO `dispositions` (`id`, `disposition`, `dispoCategory`) VALUES
  ('1', 'Guilty', '0'),
  ('2', 'Not Guilty', '1'),
  ('3', 'Guilty Plea', '0'),
  ('4', 'Dismissed', '1');

INSERT INTO `arrestcharges` (`id`, `arrestid`, `grade`, `dispositionid`)     VALUES
  ('1', '1', 'M1', '1'),
  ('2', '1', 'M', '2'),
  ('3', '2', 'F', '2'),
  ('4', '2', 'M1', '3'),
  ('5', '3', 'M1', '1'),
  ('6', '4', 'M2', '4'),
  ('7', '4', 'M1', '3'),
  ('8', '4', 'M1', '3'),
  ('9', '4', 'M1', '1'),
  ('10', '5', 'M1', '2'),
  ('11', '6', 'M1', '2'),
  ('12', '7', 'F2', '1'),
  ('13', '8', 'F3', '1'),
  ('14', '9', 'M1', '2'),
  ('15', '9', 'M1', '2'),
  ('16', '9', 'M1', '2'),
  ('17', '9', 'M1', '2'),
  ('18', '10', 'M1', '1'),
  ('19', '10', 'M1', '1'),
  ('20', '11', 'M2', '3'),
  ('21', '12', 'M1', '4'),
  ('22', '12', 'M1', '3');  

Answer 1

Try this query:

select a.defendantid, count(distinct(ac.arrestid)) as count
    FROM `arrest` AS a LEFT JOIN `arrestcharges` AS ac 
          ON a.id=ac.arrestid LEFT JOIN `dispositions` AS d 
          ON ac.dispositionid=d.id
    WHERE d.dispocategory=0 AND ac.grade="M1"
    GROUP BY a.defendantid HAVING count>1;

Answer 2

You should count the distinct number as distinct_count of rows you need and make use of having filter such as having distinct_count>1 . This way you can ensure that the count are not getting repeated.

Answer 3

You appear to be aggregating by the wrong column. You need a.defendantid in the group by :

SELECT a.defendantid, count(*)  
FROM `arrest` a JOIN
     `arrestcharges` ac 
      ON a.id = ac.arrestid JOIN
     `dispositions` d 
      ON ac.dispositionid = d.id 
WHERE d.dispocategory = 0 AND ac.grade = 'M1'  
GROUP BY a.defendantid
HAVING COUNT(DISTINCT a.id) > 1
ORDER BY a.defendantid;

Note that I also changed the outer joins to inner joins. If charges and dispositions are not available, then your filtering conditions cannot be met. Hence, the appropriate join is an inner join.