[freetds] Set up client charset with DSN-less configuration

Johnny Yan jyan at tableau.com
Wed Jul 6 13:28:29 EDT 2016

Hi Frediano,

Thanks for the reply. Actually we use *W version of APIs internally. I think the sample was over-simplified. Sorry about that.

Attached please find the updated sample with SQLDriverConnectW. However with or without ClientCharset, we didn't see the client charset is set up correspondingly.

Also about UTF-8, I think you implied this is your internal default? Is there any way to change this?

If my understanding is correct, FreeTDS is doing something like <server charset> <-> UCS-2LE <-> <client charset>. Let me know otherwise though. What we want is to set the <client charset> to UTF-16 if possible.

Hope this doesn't make you more confused. But let me know if it does. We could provide more info.

From: FreeTDS <freetds-bounces at lists.ibiblio.org> on behalf of Frediano Ziglio <freddy77 at gmail.com>
Sent: Wednesday, July 6, 2016 4:34:01 AM
To: FreeTDS Development Group
Subject: Re: [freetds] Set up client charset with DSN-less configuration

2016-07-05 22:16 GMT+01:00 Johnny Yan <jyan at tableau.com>:
> Hi,
> We're writing data analytic application on Linux using FreeTDS(0.95.81) with unixODBC(2.3.4) to connect to SQL Server databases. Since we need to programmatically decide the hosts and other info provided by our customers, we decided to use the DSN-less configuration and pass the attributes through the connection string. However it seems that we can't set up the client side charset correctly.
> From the attached simple sample code, you could see that we could connect with that connection string(you need to change the server and username/password if you want to test it). And we get the following head in the freetds.log file,
> log.c:167:Starting log file for FreeTDS 0.95.81
>     on 2016-06-30 18:17:32 with debug flags 0x4fff.
> iconv.c:328:tds_iconv_open(0xff47a0, UTF-8)
> iconv.c:187:local name for ISO-8859-1 is ISO-8859-1
> iconv.c:187:local name for UTF-8 is UTF-8
> iconv.c:187:local name for UCS-2LE is UCS-2LE
> iconv.c:187:local name for UCS-2BE is UCS-2BE
> iconv.c:346:setting up conversions for client charset "UTF-8"
> iconv.c:348:preparing iconv for "UTF-8" <-> "UCS-2LE" conversion
> iconv.c:395:preparing iconv for "ISO-8859-1" <-> "UCS-2LE" conversion
> It seems that the client charset is set to UTF-8. Am I correct? My understanding is that this is from the LANG env variable since we didn't specify it anywhere. Right?
> However when we try to specify the charset in the connection string, that is appending ";ClientCharset=UTF-16" to the connection string in the sample, we didn't observe the expected behavior. In another words, we still see the above bold line which implies that the ClientCharset attr in the connection string didn't make any difference. Is that expected? Did we do it correctly? Probably not. If so, how can we make it right?
> Let me know if you need any other info. Thank you very much!
> Johnny

  ClientCharset setting is for multi-byte, not wide. To use wide use
*W versions of the APIs. Internally utf-8 is used so this is normal.
ODBC does not force libTDS to do conversion to multi-byte so only one
conversion is done (beside metadatas).

FreeTDS mailing list
FreeTDS at lists.ibiblio.org

-------------- next part --------------
A non-text attachment was scrubbed...
Name: sampleW.cpp
Type: text/x-c++src
Size: 1117 bytes
Desc: sampleW.cpp
URL: <http://lists.ibiblio.org/pipermail/freetds/attachments/20160706/1b27a46d/attachment.bin>

More information about the FreeTDS mailing list