Use Progress and ODBC on CentOS with Japanese

This is related to my previous blog about using Chinese characters with Progress, .Net and Mono: Allow Chinese characters on a Client-Server application

Now the next challenge was to get Japanese characters to work.

I first converted the Progress database to shift-jis.
I set the environment variable: SQL_CLIENT_CHARSET=SHIFT-JIS

I then experimented with several locales, e.g. ja_JP.UTF8 and ja_JP (which is an alias for ja_JP.eucjp).

I used Wikipedia to get some sample words in Japanese characters, see http://en.wikipedia.org/wiki/Japanese_writing_system.

I soon realised that with some words I would get an ODBC error:
eg. with Kanji script 金魚 (Goldfish)
and with Hiragina script, きんぎょ
If I left out the last letter, it worked.

The ODBC error was:

Saving of data failed!
Reasons:
* [[ODBC]] [unixODBC][DataDirect-Technologies][ODBC 20101
driver][20101]Internal error -4 (unknown nls services error) in SQL from
subsystem NLS SERVICES function nlsCountChars called from nc_get_strlen
on . for . Save log for Progress technical support.

A helpful person from Progress Support gave me some suggestions, and I got it to work with Progress 9.1D in the end!
I had to use another locale, ja_JP.SJIS, which did not exist by default on my CentOS 4.6 (locale -a | grep JP).

According to http://kbase.redhat.com/faq/FAQ_80_1256.shtm
you can create the locale ja_JP.SJIS with this command:
localedef -f SHIFT_JIS -i ja_JP ja_JP.SJIS

This resolved the internal ODBC error.

Another issue was that I was not aware how to properly set the font in the ini file. Finally I found a hint in the Progress manuals, to check the prolang directory for the file progress.ini which is available for each language on the client side; this has a sample font. For the Japanese, this is a font with Japanese characters. I just copied it, and it worked.