Re: Proposal: add simpler case closure

I like the idea, though it does add API complexity. I don't think 
exemplar sets are that performance-sensitive, but other applications of 
the new capability might well be. The constants would then be:

USET_UPPERCASE_CLOSURE
USET_LOWERCASE_CLOSURE
USET_TITLECASE_CLOSURE

I will have to do two implementations, as I need this work on 3.0 as 
well...

Deborah

On Oct 4, 2004, at 3:29 PM, Markus Scherer wrote:

> Deborah Goldsmith wrote:
>> The simple case closure will take the existing items in the set, and 
>> perform uppercase, lowercase, and titlecase (only relevant for 
>> strings) operations on them, then add them back in. Therefore [a S 
>> DZ] would become [a A s S DZ Dz dz].
>
> Possible alternative:
>
> You could add three options - one to add the simple-lowercase forms of 
> the existing characters, one for upper, one for title. Then you would 
> use only the upper+title for exemplar characters because we store them 
> in lowercase.
>
> This would get proper behavior, add some useful functionality, and 
> would also help with performance because it would not have to look up 
> the lowercase form of already-lowercase characters.
>
> You could implement these three with one single loop.
>
> Note: I recently munged the ICU4C case mapping code a lot, and 
> modified UnicodeSet some, too. You will want to take a look at the 
> new, internal ucase.h and use its functions, starting from 
> ucase_getSingleton(). Makes it faster and more modular. Please keep 
> such code in uniset_props.cpp, for modularization (uniset.cpp itself 
> is now properties-free).
>
> markus
> _______________________________________________
> icu mailing list
> [email protected]
> http://oss.software.ibm.com/developerworks/oss/mailman/listinfo/icu