Since v3.6.1, SQLite3 amalgamation package includes the ICU (International Components for Unicode) source code to make it easier to build SDLite3 with the ICU extension enabled, see [check-in].

This change has broken sqlite3_unicode source code and could no longer be automatically loaded by hijacking the ICU start-up function. For this reason sqlite3_unicode has been updated (to build 10) which is fully compatible with SQLite v3.6.7, but must be manually loaded using custom exported functions, which are defined in a separate sqlite3_unicode.h header file, and explained below.

This version has been tested to build under Microsoft Windows and Ubuntu 8.10 Intrepid and work flawlessly either as a static or shared library.

A bug was fixed in the NOCASE collation source that prevented proper sorting of UTF-8 encoded strings.
NOCASE collation now works for both UTF-8 and UTF-16 encoded strings, but unlike ICU it does not take into account any language specific sorting logarithms.

To activate unicode functionality the following function must be called during application initialization sequence: sqlite3_unicode_load();
The function does not take any arguments and simply calls sqlite3_auto_extension(); function to load the extension.

Similarly before application exit the following function must be called to cleanup memory and unload extension: sqlite3_unicode_free();
The function does not take any arguments and simply calls sqlite3_reset_auto_extension(); function to unload the extension.

The library has been thouroughly tested for Greek language through a small greek-english dictionary project, designed to test:

1. the case folding conversion,
2. the LIKE operation in a SELECT statement, and
3. the NOCASE collation (using case folding and unaccenting of greek strings to enable proper sorting).

Download the implementation file sqlite3_unicode.c