Re: psqlODBC 08.01.0100 Released

On 9/11/05 8:28 pm, "Peter Eisentraut" <peter_e(at)gmx(dot)net> wrote:

> Dave Page wrote:
>> From this release, two versions of the driver are provided for
>> Windows, 'PostgreSQL ANSI' which supports single and multibyte
>> applications through the ANSI ODBC API, and 'PostgreSQL Unicode'
>> which provides Unicode support through the Unicode ODBC API. On Unix
>> systems, the driver type may be selected via a configure option.
>
> With my packager hat on, I'm puzzled by this. I am now supposed to
> produce two shared libraries? I thought the idea was to have both sets
> of functions in one binary and let the driver manager or application
> program figure out which to call.

This was discussed in depth a few months back. The Unicode driver will only
work with Unicode data or non-unicode data that maps directly into the first
unicode page (or whatever the correct term is). Non-ASCII charsets don't
work correctly with the Unicode driver as a result.

Drivers such as the Microsoft SQL Server driver handle this by having an
option to map strings on the fly from the client encoding into unicode. It
has to be done this way because if the Unicode functions exist, the driver
manager will *always* map all function calls to those instead of the ANSI
versions.

To get around this without providing our own remapping of non-unicode data,
the driver was reverted back to the way it was in the 07.xx series - ie.
Having a Unicode and ANSI version on Windows. On Unix, the
--enable/disable-unicode option that you added some time back was
reinstated.

The net result of this change was that a whole heap of complaints from
people using latin* as their client encoding went away overnight.

I've made no secret of the fact that I know little of encoding issues
however, so if you have a better way to handle this then please speak up!

Regards, Dave