At 02:12 PM 11-08-01 +1000, you wrote:

> > On the other hand Win2K (Locale English (Australia)) gives:
> > A NAME
> > A$NAME
> > A_NAMe
> > AANAME
> > A-NAME
> > AZNAME
><...>
> > Further experiments shows windows windows sorting the '-'
> > between M and N. I really dont know what the
> > logic is behind this.
>
>I finally worked it out (I can be slow sometimes). '-' is being
>treated almost as though it does not exist. That is '-' has less
>significance than even a space. eg:
>
> 'AA AA'
> 'AAAA'
> 'AA-AA'
> 'AAAA '
> 'AAAAA'
>
>Which seems to indicate that 'AA-AA' is treated as almost the same as
>'AAAA' in a sort.
>
>Fascinating stuff. :-)
>
>I guess those that deal with languages a lot may understand why this
>happens, for me (english only and not that good at that), this seems
>strange.

It has to do with the priority given to the non-characters in the sort-order defined for character sets used by different languages. For example, a hyphen is a more significant character in English and French orthographies than it is in German, for example, where compounded words rarely have hyphens.

However, I can't see what sort order and character sets have to do with the problem described, which is TIB_LookupCombo feeding an altered string to the Keysource. I think the problem might be more closely related to what is being used as the keylinks of the lookup dataset. There ARE problems when the list value in the lookup list is being matched as the KeyLink to the Keysource column.

I would want to run the SQL for the lookup dataset in IB_SQL and look at what is being returned there. Also, since an atomic key isn't being used for the linkage, I would be wondering whether the link is being corrupted by a size or character type mismatch.

regards,
Helen


All for Open and Open for All
InterBase Developer Initiative · http://www.interbase2000.org
_______________________________________________________