Re: [icu-design] ICU converters: required to support more than 40, 000 mappings?

[Markus S. <mar...@gm...>]

:-)
Sorry, no dice.

The extension data structure itself uses a trie, and this trie is less
efficient than the base table trie because it always uses 4 bytes per
entry rather than 3 for the CNS table or 2 for Shift-JIS.

At least the part that would be put there is dense (not sparse), which
is good for the data structures.

On 7/12/06, George Rhoten <gr...@us...> wrote:
> You say that you're going to put the > 40,000 mappings in the extensions
> part of the table, right?  Isn't that part generally smaller than a sparse
> trie, and thus make cns-11643-1992.cnv smaller?  If so, that seems like a
> feature to me :-)  I'm hoping there are sparse parts of the tries in
> cns-11643-1992.

markus
-- 
Opinions expressed here may not reflect my company's positions unless
otherwise noted.