!!!NOTE!!!  The implementation had been improved without updating this
document, so some part of it is obsolete.


* CharUnicodeInfo internal design

** Normalization information structures

	From	DerivedNormalizationProps.txt
		NormalizationCorrections.txt
		UnicodeData.txt
	To Normalization.cs

	1) byte [] NormalizationPropIndexes (length = 65536)

	2) int [] NormalizationPropValues (length = property value count)

	A value consists of those flags:

	bit	semantics
	0	NFD_QC
	1	NFKD_QC
	2	NFC_QC : NO
	3	NFC_QC : Maybe
	4	NFKC_QC : NO
	5	NFKC_QC : Maybe
	6	Expands_On_NFD
	7	Expands_On_NFC
	8	Expands_On_NKFD
	9	Expands_On_NKFC
	10	Full_Composition_Exclusion
	11-12	index to NormalizationCorrections
	16-31	FC_NFKC : short[]
		0 : no replacement;
		plus - index to SingleNormalization
		minus - 0xF0000000 and index to MultiNormalization

	3) char [] SingleNormalization
		FIXME: it should consider NormalizationCorrections.txt

	4) char [] MultiNormalization (index for each 4 bytes, zero padded)

	5) int [] canonicalEq
	6) int [] canonEquiv : index to canonicalEq


** Default Collation Element Table

	From allkeys.txt from UCA
	To CollationElementTable.cs

	struct SortKeyValue {
		bool Alt;
		int Primary;
		int Secondary;
		int Thirtiary;
		int Quartenary;
	}

	SortKeyValue [] keyValues; of length count of defined values

	struct CollationElementEntry {
		short Index;
		short Count;
	}

	int [] collElem; of length char.MaxValue
	bits:
	 0-15 : index of top entry to keyValues
	16-31 : count or 0 (count = 1)

	Indexes are computed for compression (eliminating vast zero-arrays).
	See CollationElementTableUtil.cs for the implementation.


** Combining class information

	From DerivedCombiningClass.txt
	To CombiningClass.cs

	It generates one simple GetCombiningClass () method.

** Width information

	From EastAsianWidth.txt

	1) WidthType : enum ( : short)
		values are: N/A/H/W/F/Na (FIXME: descriptive names!)


** Case folding information

	FromCaseFolding.txt

	(SpecialCasing.txt is not considered. We could use TextInfo.)

	Common case folding is handled in Char.ToLower(InvariantCulture).

	1) CaseFoldingIndexes : short index to CaseFolding

	2) CaseFoldingDetails : array of CaseFolding structure defined below:

	struct CaseFolding {
		char Simple;
		char Full1;
		char Full2;
		char Full3;
	}


** Properties which do not consist of tables

	- Hex_Digit, ASCII_Hex_Digit : too few to create tables.
	- Logical_Order_Exception : too few to create table.