Hi Roger,
On Tue, Sep 23, 2014 at 07:34:19PM +0100, Roger Willcocks wrote:
> On Fri, 2014-09-19 at 11:06 -0500, Ben Myers wrote:
> > +#define AGE_NAME "DerivedAge.txt"
> > +#define CCC_NAME "DerivedCombiningClass.txt"
> > +#define PROP_NAME "DerivedCoreProperties.txt"
> > +#define DATA_NAME "UnicodeData.txt"
> > +#define FOLD_NAME "CaseFolding.txt"
> > +#define NORM_NAME "NormalizationCorrections.txt"
> > +#define TEST_NAME "NormalizationTest.txt"
>
> Is there a reason why you're using multiple text-based data files (and
> hand-parsing them) when there's an xml formatted flat file available ?
>
> http://www.unicode.org/Public/UCD/latest/ucdxml/
The UCD files being parsed are the authoritative source. Check out
ucdxml.readme.txt.
> And a 2nd question - why does the trie need to encode "the the unicode
> version in which the codepoint was assigned an interpretation" ?
You need to know whether a given code point is assigned in the version
of Unicode you're normalizing for. Unicode 8 is supposed to release
June/July 2015 (see http://www.unicode.org/versions/), but filesystems
you created this year will still need the version 7 normalization.
There is still some plumbing to do to pass the version along with the
string for normalization.
I think you bring up a good point, but we'll need to support multiple
versions in the long run.
Regards,
Ben
|