Re: [mdb-dev] convert_to_ucs2

[<br...@br...>]

Ok, a better fix is in for this I moved mdb_unicode2ascii from
libmdb/data.c to libmdb/iconv.c and added a corresponding
mdb_ascii2unicode function.  This also gets rid of the need to
malloc/free.  oops, forgot to remove the old function...will do that
today.

Brian

On Fri, 05 Mar 2004 07:45:52 -0800 (PST), br...@br... wrote:

> 
> Yeah that was a quick fix when I noticed the results of the patch as
> is.  I want to re-visit it today or tomorrow and move it down into
> libmdb too.  The longer term solution is to use iconv for the
> conversion, but I wanted to fix the bug quickly and then have some
> time to think about fixing it right.  It will be more elegant in the
> final version, I think I need to clean up some of the building of
temp
> table columns too....it's too verbose.  Should have it all done by
> Monday.
> 
> Please check your patch against the tree, I used the row position
> number from the tdef page to get the postion of the null bit instead
> of a simple ordering of columns, and I consolidated the building of
> the mask between pack_row[3|4] since it's identical.  So, please
> verify that I didn't screw up the effects of your patch.
> 
> I'm having some problems running "list tables" in odbc which I need
to
> look at too.
> 
> Brian
> 
> On Fri, 5 Mar 2004 04:58:18 -0800 (PST), Jeff Smith wrote:
> 
> > 
> > I would like to suggest a parameter change in convert_to_ucs2
> > recently added to src/sql/mdbsql.c.  Add an additional
> > parameter of int * into which the final size is stored.  That
> > way, 1. you don't have to calculate it after you return, and
> > 2. if logic for convert_to_ucs2 changes, you don't only have
> > to update the logic for the string size in one place (high
> > cohesion.
> > 
> >  -- Jeff S
> >