FreeTDS Development Group

["Lowden, James K" <LowdenJK AT bernstein.com>]

> From: ZIGLIO Frediano [mailto:Frediano.Ziglio AT vodafoneomnitel.it]
> Sent: June 24, 2002 10:09 AM
> > 
> > The UG syntax is wrong; the correct syntax is "set TEXTSIZE 
> > 8000".  It would
> > be part of your SQL query.  The statement affects all queries 
> > for the spid.
> 
> I think a best solution should be use variable field for big 
> one (ex: field
> >1K should be reallocated if needed). This should solve the 
> missing TEXTSIZE
> problem. 

This for sure: blindly attempting to allocate 4 GB and dumbly suffering the
consequences is not useful.  ;)  

You're saying we can allocate space for the actual data as needed, per field
per row.  That's fine.  It might be a little too clever, though; I imagine
it would be simpler (if less memory efficient) to allocate the same size for
every row.  But I'm the sort who never writes "malloc" without having some
incense burning to keep away evil spirits.  

I think we have to assert a column width before any rows are retrieved, to
support binding and to answer dbcollen() (and friends) correctly.  If
dbcollen() returns 4 GB, client code is likely to segfault even if we don't.


> Also If I want to put DivX on a column I can. 

What's "DivX"?  

> Do you think this
> feature can be useful (I think it'll be a 0.7 feature...) ?

Yes.  Next release, definitely, IMO; we want to see Bill's work and the new
ODBC stuff settle in first.  I'd like to hear from Brian about what the
default column width ought to be, and how he'd like text fields to be
allocated.  

Regards, 

--jkl