Mám to takto:

SELECT products_id, additional_name, min( additional_value ) , max( additional_value ) as val, additional_text
FROM products_additional
WHERE TYPE = 'ADI' AND (additional_value regexp '^[0-9]' = 1)
GROUP BY additional_name

UNION

SELECT products_id, additional_name, additional_value, additional_value as val, additional_text
FROM products_additional
WHERE TYPE = 'ADI' AND (additional_value regexp '^[0-9]' = 0)

Ale nepracuje to, jak si představuju.
Nenumerické hodnoty jsou stále v jednom sloupečku:



Hádám, že to bude v tom substitutu za numeric.

Tady kdyžtak přidávám tabulku:

-- Struktura tabulky `products_additional`
--

CREATE TABLE IF NOT EXISTS `products_additional` (
`products_additional_id` int(11) NOT NULL auto_increment,
`type` char(3) collate cp1250_czech_cs NOT NULL,
`products_id` int(11) NOT NULL default '0',
`additional_name` text collate cp1250_czech_cs,
`language_id` int(11) NOT NULL default '0',
`additional_value` varchar(255) collate cp1250_czech_cs NOT NULL,
`additional_text` varchar(128) collate cp1250_czech_cs default NULL,
PRIMARY KEY (`products_additional_id`),
KEY `products_id` (`products_id`)
) ENGINE=MyISAM DEFAULT CHARSET=cp1250 COLLATE=cp1250_czech_cs AUTO_INCREMENT=43 ;

--
-- Vypisuji data pro tabulku `products_additional`
--

INSERT INTO `products_additional` (`products_additional_id`, `type`, `products_id`, `additional_name`, `language_id`, `additional_value`, `additional_text`) VALUES
(34, 'ADI', 61500, 'Spotreba', 0, '20', 'Spotřeba'),
(35, 'ADI', 61500, 'Rozliseni', 0, '1280x1024', 'Rozlišení'),
(36, 'ADI', 61507, 'ScanerDPI', 0, '2000', '2000 DPI'),
(37, 'ADI', 61512, 'CPU', 0, 'CPU 4MB', 'CPU 4MB'),
(38, 'ADI', 61500, 'Spotreba', 0, '40', 'Spotřeba'),
(39, 'ADI', 61500, 'Vaha', 0, '10', 'Váha'),
(40, 'ADI', 61500, 'Vaha', 0, '15', 'Váha'),
(41, 'ADI', 61500, 'Rozliseni', 0, '1680x1050', 'Rozlišení'),
(42, 'ADI', 61500, 'Rozliseni', 0, '1920x1080', 'Rozlišení');



Problém byl jen v regulárním výrazu:

SELECT products_id, additional_name, min( additional_value ) , max( additional_value ) as val, additional_text
FROM products_additional
WHERE TYPE = 'ADI' AND (additional_value regexp( '^[0-9]+$') = 1)
GROUP BY additional_name

UNION

SELECT products_id, additional_name, additional_value, additional_value as val, additional_text
FROM products_additional
WHERE TYPE = 'ADI' AND (additional_value regexp ('^[0-9]+$') = 0 )


Díky Spidere !