[egenix-users] Hi all, bug or feature?

[Rafael Pinilla]

hi all.

i'm in the process to make SQL requests in an automatic way. Right now, 
the data are in a Sybase SQL anywhere 7 database providing native ODBC 
driver as System DSN on Win32 platform.

The base is quite complex, with more than 20 tables and more than 
800.000 records in one table, 5 years old, with integrity reference 
checking from beginning.

I installed mxODBC, with test license, and I tried to reach my data.

After I figured out how to firt create an instance of the connexion, 
then an instance of the cursor, I tried hard coded SQL statement.

I gracefully had a sequence of tuples result. As I wanted to count some 
results, i tried cursor.rowcount() Afortunately the results were very 
different from direct SQL based request (like isql statements).

A deeper look showed me that the sequence of tuples were right until the 
last row, and then, a variable amount of null tuples were added.

So, cursor.rowcount return the good number of rows, but there are more 
empty (null) rows than expected. I really can't figure why.

I found a workaound using
     i=1
     courant = list(cur.fetchone())
     while courant[1] <> () :
         print i
         courant = cur.fetchone()
         print courant
         i = i+1

In these lines, I convert the tuple in a list to let python make a test 
on the first (second, yes I figure) item. This is a right 
cursor.rowcount workaround for suspicious dummies.

So, this bug dont stop me. Perhaps is it a bug, perhaps an unknown 
feature. I can give more information, perhaps a shelve of the resulting 
tuple. Ask me if you want. Perhaps the cursor.rowcount method should 
make a test to stop inserting null tuples at the end of the sequence of 
(good) tuples.

Thanks for your reaction.

Rafael