Clean duplicate rows on custom conditions - mysql

Question

I have this problem managing notes. I started with the strategy to always INSERT new notes and SELECT the last one. Please don't laugh, I must have thought it was a good idea, but right now, the system is not even in all-out production and there's been 300k rows inserted in about a month. In two years, my system will fail. I need to merge duplicate lines. Here is the structure of my notes table:
CREATE TABLE IF NOT EXISTS `ps_notes` (
`CodeNTE` int(11) NOT NULL AUTO_INCREMENT,
`CodePRS` int(11) NOT NULL,
`CodeXYZ` int(11) NOT NULL,
`Type` char(3) NOT NULL,
`Focus` char(3) NOT NULL,
`Texte` tinytext NOT NULL,
`Date` datetime NOT NULL,
PRIMARY KEY (`CodeNTE`),
KEY `CodeXYZ` (`CodeXYZ`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 AUTO_INCREMENT=335068 ;
Notes can be related to a person CodePRS, are necessarily related to a Type, Focus and CodeXYZ. They have a Texte entry and sometime I want to know the Date.
CodeXYZ is a unique identifier for the entity to which the note is attached. This identifier can come from any table and therefore is not absolutely unique, hence comes the Type field. This field specifies from which table the parent row comes. The focus field distincts notes that refer to a same CodeXYZ and Type.
Here some sample lines:
+---------+------+-------+-------------+------------+
| CodeXYZ | Type | Focus | Texte | Date |
+---------+------+-------+-------------+------------+
| 30008 | ctr | adm | Whatever | 2013-05-09 |
| 30008 | ctr | adm | Whatever | 2013-06-10 |
| 30008 | ctr | adm | Lorem ipsum | 2013-06-11 |
| 30008 | ctr | clt | He's cool | 0000-00-00 |
| 2546 | ctr | sup | Another | 2013-02-11 |
| 2546 | ctr | sup | Another | 2013-02-11 |
| 2546 | ctr | sup | Another | 2013-02-19 |
+---------+------+-------+-------------+------------+
this is the output I'd like to have:
+---------+------+-------+-------------+-----------------------------------------+
| CodeXYZ | Type | Focus | Texte | Date |
+---------+------+-------+-------------+-----------------------------------------+
| 30008 | ctr | adm | Lorem ipsum | 2013-06-11 (I want the most recent one) |
| 30008 | ctr | clt | He's cool | 0000-00-00 |
| 2546 | ctr | sup | Another | 2013-02-11 |
| 2546 | ctr | sup | Another | 2013-02-19 |
+---------+------+-------+-------------+-----------------------------------------+
Conditions for merging
I want to merge lines that have the same CodeXYZ,Type and Focus when Focus is not 'sup'.
When Focus is 'sup' I want to merge the lines that have the same CodeXYZ,Type,Focus and Date
Always I want to keep the most recent one
So I ran this query to merge rows in a temporary table:
INSERT INTO notes_tmp (CodePRS,CodeXYZ,Type,Focus,Texte,Date)
SELECT CodePRS,CodeXYZ,Type,Focus,Texte,Date
FROM notes
GROUP BY CodeXYZ,Type,Focus
But that way, all lines will be merged even the last ones.
So I thought of this:
INSERT INTO notes_tmp (CodePRS,CodeXYZ,Type,Focus,Texte,Date)
SELECT CodePRS,CodeXYZ,Type,Focus,Texte,Date
FROM notes
WHERE Focus<>'sup'
GROUP BY CodeXYZ,Type,Focus
ORDER BY Date DESC
UNION
SELECT CodePRS,CodeXYZ,Type,Focus,Texte,Date
FROM notes
WHERE Focus='sup'
GROUP BY CodeXYZ,Type,Focus,Date
ORDER BY Date DESC
but UNION is not at the right place, I don't think I can use it in INSERT INTO ... SELECT sql syntax
Is there a way to manage copying those lines over in a single mysql call with multiple sub queries all ending up in the same table acording to separate conditions

Answer 1

you can use group_concat to merge text field and make other columns unique with group by. try this:
INSERT INTO notes_temp
SELECT CodeXYZ,Type, Focus,GROUP_CONCAT(Texte),Date
FROM notes WHERE Focus = 'sup'
GROUP BY CodeXYZ,Type, Focus,Date;
INSERT INTO notes_temp
SELECT CodeXYZ,Type, Focus,GROUP_CONCAT(Texte),MAX(Date)
FROM notes WHERE Focus <> 'sup'
GROUP BY CodeXYZ,Type, Focus;
check sqlfiddle

Answer 2

So with part of #Volkan answer, I could come up with this somehow strangely working sql to get the correct note out of my GROUP_CONCAT()
The case will get the last entry of the group concat. I used another separator (,,,) because commas do happen often in text. three in a row a little bit less.
INSERT INTO notes_temp
SELECT CodeXYZ,Type, Focus,Texte,Date
FROM notes WHERE Focus = 'sup'
GROUP BY CodeXYZ,Type, Focus,Date;
INSERT INTO notes_temp
SELECT
CodeXYZ,
Type,
Focus,
CASE
WHEN COUNT(Texte) > 1
THEN SUBSTR(GROUP_CONCAT(Texte SEPARATOR ",,,"),((LENGTH(GROUP_CONCAT(Texte SEPARATOR ",,,"))+2) - INSTR(REVERSE(GROUP_CONCAT(Texte SEPARATOR ",,,")),",,,")))
ELSE
Texte
END
AS Texte,
MAX(Date)
FROM notes WHERE Focus <> 'sup'
GROUP BY CodeXYZ,Type, Focus;