From your first message, I thought the structure was meant to be nnn.nnn.aaa.aaa, where nnn represent numbers and aaa represent characters, but where the users did not always enter the periods. But from the data you sent me, I see that this is not the case. I had a similar case many years ago myself. I ended up with making a tool for my user where I filtered out all the wrong records in a grid, and the user had to manually fix the wrong data. The user was mad at first, but very happy after he realized that it all was his own fault. In a later version of this user's application the data is separated into several fields.

>Good Morning Tore,
>
>no, this is not "real". The real, specified ones ore the properly separated digit and num words.
>But this is my customers spec. His next site might be on somebody who likes it mixed.
>
>But this does not mean one have not to think about worst case. And sometimes there are records similar to this. (In the end "20DP1" is not much different from an 40 char word mix.)
>
>I think I need three ways - I will do a other message with what I think after I have worked down the whole thread.
>
>Agnes
>
>>Hi Agnes,
>>
>>I have checked the file you sent me, and if that file is the real data, I see much less structure in the data than I saw in the data you showed in your first message. If the file contains mostly real data, I see no way to get it sorted properly.
>>
>>>Done
>>>
>>>>Email me some test data, and I will make an easier solution later today.