Table output is empty intermittently when using fetchall() or cursor as iterator

[lucharo]

Hi, just started using impyla as an alternative to pyhive to connect to a HiveServer2 instance.

I've noticed that fetchall() and using the cursor as an iterator intermittently return empty data. I ran some experiments to test this:

My connection code:

configuration = {
    "hive.execution.engine": "tez",
}
conn = connect(
    host=HOSTNAME,
    port = 10000,
    user=USER,
    auth_mechanism="GSSAPI",
    kerberos_service_name="hive",
    database="mydb"
)

My benchmark code after setting a the connection using fetchall

%%time
empty = []
for _ in range(50):
    cur = conn.cursor()
    cur.execute('select * from db.my_table', configuration = configuration)
    output1 = []
    output1 = cur.fetchall()
    empty.append(output1 == [])
collections.Counter(empty)
## this returns Counter({False: 43, True: 7}), meaning 7 out the 50 times the output was empty

Then I run the same benchmark using the cursor as an iterator and a for loop:

empty = []
for _ in range(50):
    cur = conn.cursor()
    cur.execute('select * from db.my_table', configuration = configuration)
    output2 = []
    for row in cur:
        output2.append(row)
    output2 == []
    empty.append(output2 == [])
collections.Counter(empty)
## this returns Counter({False: 31, True: 19}), meaning 19 out the 50 times the output was empty

[csringhofer]

Hi! I have tried to reproduce this issue in the Impala development environment, but didn't notice any empty results sets.
Can you provide more info about the issue?

• What version of Impyla was used?
• What Hive version was used?
• What kind of table is db.my_table, e.g. number of rows in the table.