Possible bug in cursor iterator

Submitted by: Pietro Delsante <pietro.delsante at gmail.com>

Hello,

I have a script that uses psycopg2 2.4.5 (OS is Ubuntu 12.04, psycopg2 was installed from Ubuntu repository with apt-get) to open a connection to a local PostgreSQL database and another one to a remote PostgreSQL database, like that:

logger.info("Connecting to local database...")
try:

localconn = psycopg2.connect(host='localhost', dbname='somedb', user='someuser')
localcur = localconn.cursor(cursor_factory=psycopg2.extras.DictCursor)

except:

logger.critical("Could not connect to local database, aborting!")
sys.exit(0)

logger.info("Connecting to remote database...")
try:

remoteconn = psycopg2.connect(host='somehost', dbname='somedb', user='someuser', password='somepassword')
remotecur = remoteconn.cursor(cursor_factory=psycopg2.extras.DictCursor)

except:

logger.critical("Could not connect to remote database, aborting!")
sys.exit(0)

Then, I try to execute a SELECT query with one of the two cursors, then I start a cycle iterating on the results, like that:

remotecur.execute("SELECT * FROM sometable;")
for remoteresult in remotecur:

# Do something...

When I try to do that with the cursor pointing to the REMOTE database, everything seems to work pretty well; however, if I do the very same with the LOCAL database, like that;

localcur.execute("SELECT * FROM sometable;")
for localresult in localcur:

# Do something...

... I get this exception:

me@myhost:/path/to/# ./myscript.py
Traceback (most recent call last):
File "./myscript.py", line 37, in

for localsample in localcur:

File "/usr/lib/python2.7/dist-packages/psycopg2/extras.py", line 101, in iter

yield res.next()

psycopg2.ProgrammingError: no results to fetch

Of course, I checked that, in the LOCAL database, the query I am trying to execute does have a lot of results. In fact, if I modify my code to use fetchall() instead of using the iterator, like that:

localcur.execute("SELECT * FROM sometable;")
for localresult in localcur.fetchall():

# Do something...

... everything works pretty well, except that, of course, it's way slower.

Could this be a bug?

Thanks,
Pietro
pietro.delsante at gmail.com