[unixODBC-support] what to do about unicode?
support at forceway.com
Tue May 27 22:34:48 BST 2008
Nick Gorham wrote:
> support wrote:
>> I spent a-lot of time googling and reading archives, but i'm still not
>> sure how to proceed. I'm writing some crossplatform odbc code. The UI is
>> unicode. Since it will run on Windows, it would be nice if I could just
>> use wchar_t and the 'wide' functions for everything. Obviously they work
>> on Windows, and if I could get them to work on UNIX i'd be done with no
>> changes to code. I'm mainly interested in targeting just MySQL right
>> now. It looks like I have 3 choices.
>> 1. forget it. use ANSI functions. The UI will have to do unicode to
>> multibyte type conversions when making odbc calls. To REALLY get this
>> right i'd have to support conversions to any and all known character
>> sets that might be used.. because i'm coming from unicode any character
>> is possible.
>> 2. perform wchar_t to UTF16 conversions when making odbc calls. On
>> Windows this would be sort of a noop, but on UNIX it is necessary
>> because wchar_t is UCS-4 and unixODBC is UTF16( or ucs-2..?). I see no
>> standard functions for this conversion on UNIX. Guess I would need to
>> use a library like iconv or ICU.
>> 3. I can make unixODBC take wchar_t/ucs-4 by using defining
>> SQL_WCHART_CONVERT. This would be great! ...but it will only work if my
>> drivers support it and if unixODBC has been built for it. How do I know?
>>> From what i've read it doesn't work or hasn't been tested.
>> unixODBC-support mailing list
>> unixODBC-support at easysoft.com
> Well, one way, is to write the code in the same way for Windows and
> unix, both then use UCS-2, all you then have to do, is add your own
> version of the missing calls that handle the 16 bit codes. You can
> certainly use iconv to convert to other encodings, but it depends on
> what you want to do with the data.
> I suspect that you will find you don't need that many conversion
> functions, and you might find some gotchas if you let the stdlib
> handle the codes, all it takes is someone to change the LANG and it
> all changes. Just look at some of the problems Java and Perl have had
> using/not using UTF8 LANGs
Ok. I was planning a thin wrapper around the odbc calls anyway. I can
put the wchar_t <--> utf16 conversion in there. I don't even know that
unicode in the database is really necessary, but the UI will be unicode
so I might as well be doing odbc unicode to UI unicode conversions
instead of odbc ascii to UI unicode.
More information about the unixODBC-support