Encoding logic for MS Access "General Legacy" (Access 2000–2007) text index entries.

MS Access stores text index values in a proprietary, order-preserving byte format rather than using standard Unicode sort keys. This class implements that format for the "General Legacy" sort order (LCID 1033, version 0). GeneralIndexCodes (Access 2010+) and General97IndexCodes (Access 97) extend or mirror this class for their respective format variants.

Byte-format overview

Each character in the input string contributes zero or more bytes to up to four output sections, which are written sequentially:

1. Inline bytes – the primary sort key bytes for the character. Written immediately into the main output stream. Most printable ASCII characters contribute exactly one byte here.
2. 1 (0x01) – a terminator byte written after all inline bytes.
3. Extra / international bytes (optional) – diacritic / accent weights for characters that have a base character plus a modifier. Each normal char contributes a placeholder byte (INTERNATIONAL_EXTRA_PLACEHOLDER = 0x02); each accented char contributes its accent code. Written after the terminator if any accented characters were present.
4. Unprintable bytes (optional) – position-encoded bytes for control characters and other non-printable code points. Written after extra bytes if present.
5. Crazy codes (optional) – additional flags for a small set of special characters that affect sort order in complex ways. Written last.

Character handlers (CharHandler)

Each of the 65 536 BMP characters is mapped to a CharHandler instance that supplies the byte values for each of the four sections above. Handlers for code points U+0000–U+00FF are loaded from index_codes_genleg.txt; handlers for U+0100–U+FFFF from index_codes_ext_genleg.txt. Both files live in the Jackcess classpath resources under DatabaseImpl#RESOURCE_PATH.

Relevance for non-General sort orders

Any sort order other than the three General variants (LCID 1033) currently lacks a corresponding implementation in Jackcess. Attempting to write an index on a text column with, for example, Turkish sort order (LCID 1055) will cause IndexData#setUnsupportedReason to be called and the index to become read-only. Adding support for a new locale requires reverse-engineering the MS Access byte tables for that locale and providing a new *IndexCodes class plus a corresponding IndexData.ColumnDescriptor subclass — not using java.text.Collator#getCollationKey, whose byte format is JVM-internal and incompatible with the MS Access format.