FreeTDS Development Group

[Michael Peppler <mpeppler AT peppler.org>]

James K. Lowden writes:
 > Michael Peppler wrote:
 > 
 > >  >From what I'm seeing, MS is, by default under TDS 7.0, actually sending
 > >  > the entire text clob all in one go.  So technically they are sticking 
 > > to
 > >  > the protocol. It just makes the maxlength value useless for 
 > > preallocating
 > >  > memory!
 > >
 > > What?!?!
 > >
 > > Don't they honour the default TEXTSIZE value?
 > >
 > > (oh, and of course they don't allow setting CS_OPT_foo via
 > > ct_options() either, so you have to execute it as a language statement
 > > for that connection)
 > 
 > Well... MS doesn't offer a ct-lib client, so we can't really say what they 
 > do. :-{  Would
 > better ct_option support in FreeTDS make DBD::Sybase better?

Maybe. But none of the ct_options() calls work when using the Sybase
libs and MS-SQL, so I suspect that this might be another protocol
glitch. 

 > 
 > >  > If freetds were to allow a "set textsize" to be executed 
 > > automatically on
 > >  > connect, ct_describe would return a sane value and everyone should be
 > >  > happy.
 > 
 > If I read Brian's message right, ct_describe doesn't return a value 
 > representative of
 > TEXTSIZE until it's been explicitly set, which is (as Mr.Friesen observes) 
 > less than
 > helpful.  A pain in the tookus, my grandfather would have said.
 > 
 > The documentation http://larr.unm.edu/~owen/SQLBOL70/html/globals_26.htm  
 > says:
 > 
 > "The default size is 4096 bytes (4 KB)."
 > 
 > Houston, we have a bug.

:-)


 > 
 > May I offer a suggestion?
 > 
 > If you're going to send "set textsize" on connect, maybe the default value 
 > (in the absence of
 > a config file value) should be 32 KB cf. Sybase or 4 KB cf. Microsoft?  
 > Principle of least
 > astonishment and all that?  I think that would cut down on the number of 
 > questions.  Better
 > truncated data (which the calling application can should test for in any 
 > case) than a broken
 > query.

I think that that sounds good. That is definitely what I would have
expected to happen based on my (limited) knowledge of the MS-SQL
values for @@textsize.

 > There's a fly in the ointment, too: unicode.  "Textsize" is measured in 
 > bytes, not
 > characters.  Just when you thought you knew what text is, the world goes 
 > multilingual on you.
 > 

Indeed. But for the purpose of allocating memory I suspect that the
@@textsize values should work OK (I hope!)

Michael
-- 
Michael Peppler - Data Migrations Inc. - mpeppler AT peppler.org
http://www.mbay.net/~mpeppler - mpeppler AT mbay.net
International Sybase User Group - http://www.isug.com
Sybase on Linux mailing list: ase-linux-list AT isug.com