Character sets requiring special string collating routines for sorting or multi-byte character support will need custom work a configuration file simply cannot handle. To accomplish this task, you will need to be familiar with C programming, as you will need to create several C source files. If you plan to attempt this, it is highly recommended that you study the existing ctype-*.c files and base yours on these.

As with simple character sets, you need to edit the sql/share/charsets/Index and configure.in files. Instead of a configuration file, however, you need to create a C source file called ctype -charset.c. In our case, this file will be ctype-elvish.c. You can get a head start with your character set file by copying the source file for a similar character set.

The first task in this source file is to define ctype , to_lower , to_upper , and sort_order . These are the same arrays you configured in the previous section, except they are defined as C arrays. You’ll need to create:

For our example, we need to create ctype_elvish, to_upper_elvish, to_lower_elvish, and sort_order_elvish.

If you need string collating functions, you must write your own custom string collating functions. These functions should be named:

We therefore need to create my_strnncoll_elvish, my_strnxfrm_elvish, my_strcoll_elvish, my_strxfrm_elvish, and my_like_range_elvish.

Multi-byte support requires the coding of special multi-byte functions:

We need to create ismbchar_elvish, ismbhead_elvish, and mbcharlen_elvish.

With your coding complete, you need to create a comment header for your ctype file. This header should look like this:

/*
 * This comment is parsed by configure to create ctype.c,
 * so don't change it unless you know what you are doing.
 *
 * .configure. number_charset=[character set index number from Index file]
 * .configure. strxfrm_multiply_charset=[max ratio that strings may grow
 *                                       during my_strxfrm_charset - a
 *                                       positive integer]
 * .configure. mbmaxlen_charset=[size in bytes of largest char in set]
 */

For our example, this might look like:

 * .configure. number_elvish=32
 * .configure. strxfrm_multiply_elvish=1
 * .configure. mbmaxlen_elvish=2

Finally, compile MySQL and install your new binaries supporting the custom character set.