Hugo,

thank you

INSERT INTO c_test (value) VALUES ('1.02.2')
INSERT INTO c_test (value) VALUES ('1.02.10')
INSERT INTO c_test (value) VALUES ('1.02.N')
INSERT INTO c_test (value) VALUES ('1.02.NR')
INSERT INTO c_test (value) VALUES ('1.02.N.R')

fails.
I think I do not need tho have any word that long.

Agnes

>Agnes,
>
>The problem I see with the idea of padding the words to get the correct order is that the string could become too large to use in the sort by, to minimize it you can count first what is the maximum number of words, something like this
>
>

>LOCAL laMax[1]
>CREATE CURSOR c_Test (PK int autoinc, value char(40))
>
>INSERT INTO c_test (value) VALUES ('1.1')
>INSERT INTO c_test (value) VALUES ('1.02')
>INSERT INTO c_test (value) VALUES ('10.1')
>INSERT INTO c_test (value) VALUES ('01.3')
>INSERT INTO c_test (value) VALUES ('1.02.N')
>INSERT INTO c_test (value) VALUES ('1.02.N.R')
>INSERT INTO c_test (value) VALUES ('1.10.R')
>INSERT INTO c_test (value) VALUES ('1.10.R.1')
>INSERT INTO c_test (value) VALUES ('20DP')
>INSERT INTO c_test (value) VALUES ('20')
>INSERT INTO c_test (value) VALUES ('1.02.R')
>
>SELECT MAX(GETWORDCOUNT(value, '.')) FROM c_Test INTO ARRAY laMax
>
>SELECT getorder(ALLTRIM(c_test.value), laMax[1]), c_test.* FROM c_test ORDER BY 1
>
>
>FUNCTION getOrder(tcValue, tnMaxWords)
>
>LOCAL lnWord, lnWords, lcValue
>
>lnWords = GETWORDCOUNT(tcValue, '.')
>lcValue = ''
>FOR lnWord = 1 TO tnMaxWords
>	IF lnWord <= lnWords
>		lcValue = lcValue + PADL(GETWORDNUM(tcValue, lnWord, '.'), 40, '0') + '.'
>	ELSE
>		lcValue = lcValue + REPLICATE('0', 40)
>	ENDIF
>NEXT lnWord
>
>RETURN lcValue
>

>
>Note that the code is using a word length of 40, which is the max length a single "word" can have, and in this simple sample you sent it will work, but if you have a combination of large words and small words, all the extra paddings will make the string huge and unsuitable for the order by clause.
>
>Which it makes me believe this is not the right answer :)