Keld responded to David,
>
> On Tue, Mar 28, 2000 at 08:24:57AM -0800, david@oz.com wrote:
> > I was stepping through some code that did case insensitive comparison for
> > unicode, and noticed that the heart of the function converted each
> > character to lowercase before comparing them. If they matched, they were
> > considered equal, otherwise not.
>
> If you do case insensitive comparison, the right thing is to
> do it at the case insensitive level of the ISO/IEC ordering
> standard 14651. Don't do uppercase to lowercase mapping first,
> just compare directlye, case insensitive.
>
> Keld
>
This is, indeed, one way to do case insensitive comparison. If you
have access to an implementatation of string ordering according to
the forthcoming standard ISO/IEC 14651, or according to the corresponding
Unicode Standard: UTR #10 Unicode Collation Algorithm, and if that
implementation provides a good API that allows efficient, case insensitive
comparison of two strings according to a particular collation definition,
then this can be a good choice.
However, case folding does not necessarily depend on a collation
algorithm. See also the Unicode Technical Report #21, Case Mappings, for
discussion of case folding and a suggested data file for doing locale-independent
case folding.
There are circumstances under which one definitely does *not* want to
have case folding depend on particular collation tables or on locale
differences in comparison. The explanation and examples are provided
in the Case Mappings technical report.
--Ken
This archive was generated by hypermail 2.1.2 : Tue Jul 10 2001 - 17:21:00 EDT