FreeTDS Development Group

["ZIGLIO, Frediano, VF-IT" <Frediano.Ziglio AT vodafone.com>]

> 
> ZIGLIO, Frediano, VF-IT wrote:
> > MS driver NEVER 
> > > > returns SQL_NO_DATA !!!
> ...
> > 
> > Mmmm... as said if nobody complains I would commit... so I 
> complain :)
> > 
> > Yes, it seems that DELETE/UPDATE with no rows produce SQL_NO_DATA!!
> > However I must say that as documented only DELETE/UPDATE 
> with no rows
> > should return SQL_NO_DATA. The situaction is complicated by the fact
> > that only mssql (no Sybase) return proper informations to detect
> > UPDATE/DELETE execution!
> 
> I have been trying to test this for you with the Sybase ODBC 
> driver on my
> OS X box.  Last night I built FreeTDS on it for the first 
> time, but didn't
> get to testing anything.  
> 
> http://msdn2.microsoft.com/en-us/library/ms715424(VS.85).aspx
> 
> I'm sure you've read the above confusing page more than once.  It says

Agreed :)

> that whether SQLExecute returns SQL_NO_DATA or SQL_SUCCESS depends on:
> 
> 1.  Whether the *application* is 2.x or 3.x. 
> 2.  Whether the "searched update or delete" was part of a batch.  
> 3.  Whether any rows were affected.  
> 
> According to the documentation, then, SQLExecute returns 
> SQL_NO_DATA only
> if 
> 
> 1.  a "searched update or delete" is not part of a batch -- 
> or, I presume,
> a stored procedure -- and 

I think "not" here is only a typo error.

> 2.  is issued by a 3.x application, and
> 3.  affected zero rows.  
> 
> (I do not know what a "searched update" is.  I guess it means 
> one with a
> WHERE clause.  If I issue "UPDATE tablename SET columname = 
> 0" on a table
> with zero rows, it affects zero rows.  What should the driver return
> then?)  
> 

Yes, an update with a where.

> The simplest choice: always return SQL_SUCCESS from 
> SQLExecute and reserve
> SQL_NO_DATA for SQLFetch.  
> 

That is what the patch does... I applied on 0.82 branch.

> The "right" choice (apparently): Return SQL_NO_DATA from 
> SQLExecute *only*
> for a 3.x application and *only* if you get back a DONE 
> packet (0xFD) with
> rowcount == 0 immediately after issuing the query.  
> 
> How much simpler could it be?  ;-)
> 

I did some tests and it appear that driver check only for update/where.
For instance a create table or a begin transaction (which produce a done
with 0 rows). Or perhaps they produce a DONE without rows?? 

freddy77