Re: [icu-design] Collation Problem

[mehran m. <meh...@gm...>]

Four levels of comparison is just one of the problems.
By the way, would you please explain how can I assign more than three levels
of comparison to
a character with ICU's 32-bit collation elements?

I need to assign a collation element like <x,y,0,1> to U+0627 character, for
instance.

Is it possible to extend RuleBasedCollator syntax to support these kinds of
weight assignments?

Mehran

On 9/25/06, weiv <wei...@gm...> wrote:
>
> Merhan,
> it would probably be most beneficial if you could try to define the
> Persian collation. ICU does support collations that use more than three
> levels, such as Japanese JIS 4061X. Please provide more information on your
> requirements.
>
> Regards,
> v.
>
> On Sep 24, 2006, at 2:07 AM, mehran mehr wrote:
>
> Persian Collation needs three features available in glibc collation
> implementation:
>
> 1) At least four level of comparison (five is even better)
> 2) A method to ignore base characters at third-level of comparison (this
> is supported in glibc using 'IGNORE' keyword, which assigns a zero weight,
> strictly)
> 3) A method to ignore base characters at fourth-level of comparison (this
> feature is supported in glibc using 'position' keyword)
>
> but ICU Collation implementation doesn't support these features:
>
> 1) ICU has just three independent Collation Levels.  A 32-bit Collation
> Element (2 byte for first level, 1 byte for second level, and 1 byte for
> third level).
> the Case level and Tiebreaking level and even the Hiragana Case level are
> dependent to the character code-point and are not usable for other purposes.
> And as far as I know, shifted comparison for variables doesn't work in
> ICU.
> Is it possible to add more independent comparison levels to ICU in near
> future?
>
> 2) As stated in UTS#10:
> ---------------------------excerpts from
> UTS#10-------------------------------------
> 3.3 Well-Formed Collation Element Tables
>
> A well-formed Collation Element Table meets the following conditions:
> Except in special cases detailed in * Section 6.2, Large Weight Values*,
> no collation element can have a zero weight at Level N and a non-zero weight
> at Level N-1.
>
> ----------------------------------------------------------------------------------------------
> glibc collation tables may not be well-formed by this definition. But we
> actually need this feature in Persian Collation.
> Is it possible for ICU to support these kind of zero weights?
>
> 3) It is possible to implement positioning by assigning a sufficiently
> small weight to the base characters (for example 1) at fourth level. But
> strict weight assigning
> is not supported in RuleBasedCollation class.
> Is it possible to extend this class in order to support strict weight
> assigning?
>
> Mehran
>
> -------------------------------------------------------------------------
> Take Surveys. Earn Cash. Influence the Future of IT
> Join SourceForge.net's Techsay panel and you'll get the chance to share
> your
> opinions on IT & business topics through brief surveys -- and earn cash
>
> http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV_______________________________________________
> icu-design mailing list
> icu...@li...
> To Un/Subscribe: https://lists.sourceforge.net/lists/listinfo/icu-design
>
>
>
> -------------------------------------------------------------------------
> Take Surveys. Earn Cash. Influence the Future of IT
> Join SourceForge.net's Techsay panel and you'll get the chance to share
> your
> opinions on IT & business topics through brief surveys -- and earn cash
> http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
>
> _______________________________________________
> icu-design mailing list
> icu...@li...
> To Un/Subscribe: https://lists.sourceforge.net/lists/listinfo/icu-design
>
>