Am 24.04.2006 um 14:51 schrieb COS:
Hi Jens,
----- Original Message -----
From: "Jens Miltner" <[EMAIL PROTECTED]>
To: <sqlite-users@sqlite.org>
Sent: Monday, April 24, 2006 5:30 AM
Subject: Re: [sqlite] Using sqlite3_open or sqlite3_open16?
Am 22.04.2006 um 15:48 schrieb COS:
Thanks for the info. I did have found in the manual that UTF16 is
converted
to UTF8 on Windows environments.
But one little question: How would one know if the filename is in
UTF8 or
UTF16?
You see, my little application is installed in PC's all around the
world. I
don't know in advance if the filename is UTF8 or UTF16. I noticed
that
always using UTF16 to open the database can lead to an exception in
filenames with UTF8.
Hmmh - I'm not sure I entirely understand your problem: usually
strings passed in UTF-16 are passed in 'wide' strings, i.e. each
character is 16 bit as opposed to UTF-8 strings, which are usually
passed in strings where each character is 8 bit.
I don't know the Windows APIs very well, but from what I remember,
you build your app either as Unicode app or as "ASCII" app and only
unicode apps get their filenames as UTF-16, "ASCII" apps get whatever
code page is installed/activated on the system, usually not UTF-8,
though...
My application is not built using Unicode. But I use a lot of MFC and
CString is aware of International characters. The problem is a little
different. I don't need to use UTF16 strings at all. The problem
comes up
when the application is started and it needs to open the database
from the
My documents folder. In Windows NT/2000/XP and probably others too,
the
userid is a string to compose the personal folder location (i.e.
C:\Documents and Settings\userid\My Documents). As you can see this
location
can change a lot. It may even work for a particular user and don't for
another.
Where do you get your filenames from so you don't know the encoding?
I get this information directly from Windows, using the Win32 SDK
API. The
string is correct, even for UTF16 strings. The Windows SDK API
allows to
open the file directly without problems or any conversion (maybe some
conversion is done internally by the Win32 SDK).
If you're coding against the 'standard' Win32 APIs, Windows will send
you the strings in the character encoding specified by the either
the System or the current user settings. This has nothing to do with
unicode. However, for many roman characters, the standard Windows
code pages and UTF-8 have a fair amount of overlap, so you can get
away just passing those strings to sqlite3_open() for ASCII names and
_some_ localized names, but in general, this won't work if the
filename is not pure ASCII.
If you can't use the Win32 unicode APIs (in which case you'd call
sqlite3_open16), you'll need to make sure you convert from the
Windows encoding to UTF-8 before passing the filenames.
Unfortunately, I don't know enough details about the Windows APIs to
tell you exactly how to do this :( You'll have to look up on MSDN -
searching for "MultiByteToWideChar" might get you started...
(I'm lucky enough to code on Mac OS X, where filenames are generally
passed in UTF-8 ;-)
HTH,
</jum>
