On Fri, Feb 09, 2007 at 08:13:22AM -0500, Grant Croker wrote:
> > -----Original Message-----
> > From: Henrik Tougaard [mailto:[EMAIL PROTECTED]
> > Sent: 07 February 2007 10:58
> > To: Viktor Hübner
> > Cc: [email protected]
> > Subject: Re: DBD.Ingres 0.51 unicode support
> >
> > Viktor Hübner wrote:
> > > sorry if this is the wrong address for my question, but I have not
> > > find related information on other way. I have a problem with
> > > DBD.Ingres 0.51 on SuSE Linux 9.0 with DB Ingres II 3.0.3 (int. lnx.
> > > 103). Are the unicode types (nvarchar/nchar) not supported in this
> > > version?
> > > An error message is shown "filed <Number> has unsupported
> > > type 27" if access a table with nvarchar. Have you any idea what can
> > > be this problem solved?
> >
> > Hi.
> >
> > This is the corect place for this kind of question.
> >
> > Unfortunately you are quite right - the current release of DBD::Ingres
> > does not support wide (unicode) character sets.
> >
> > As far as I can recall it should not be too difficult to add, but I
> > haven't got round to it yet.
> > And as I no longer use Ingres actively (I have switched jobs over a year
> > ago and now have to suffer under MySQL) the chances that I will add it
> > in the near furture are very slim.
> >
> > You are more than welcome to patch it yourself. I will be happy to
> > release a new version or add you as a co-maintainer of DBD::Ingres.
> >
> > Best of luck and sorry I cannot be more helpful
>
> Hi Henrik,
>
> I have managed to get Unicode fetches to work with Ingres. See
> http://ingres.co.uk/files/perl/unicode/dbd-ingres-unicode-fetch.patch
> for the patch (generated using diff -u).
>
> It has been tested on Windows XP as well as a 64 bit Linux system. If
> you feel it should be tested on another platform let me know.
>
> My test case
> (http://ingres.co.uk/files/perl/unicode/unicode-ingres.txt) uses the
> demodb database that is shipping with Ingres 2006 R2. Should you or
> anyone want to get a copy of this data and the schema please email me.
It's important that users are not burdened with having to add
decode("UTF-16LE", ...)
all over their code.
It's expected that drivers return non-iso-8859-1 character data as utf8
in scalars with the SvUTF8 flag set.
I see that the DBI docs don't mention that explicitly, so I'm adding it now.
Any driver not doing that is arguably broken, or at least painful to use
and the application is unlikely to port to other drivers easily.
Tim.
