hivejdbc

Hive database driver via jdbc


Keywords
dbapi, jdbc, hive, hadoop
License
Apache-2.0
Install
pip install hivejdbc==0.2.1

Documentation

hivejdbc

hivejdbc is db-api-2.0 compliant Apache Hive driver that supports

  • kerberos
  • ssl
  • service discovery via zookeeper
  • host-connection list
  • and all other jdbc driver options

installation

pip3 install hivejdbc

Cursors

hivejdbc can use a dictonary cursor if desired.

from hivejdbc import connect, DictCursor
conn = connect('example.com', 'default', cursor=DictCursor)

Cursors support with

from hivejdbc import connect
conn = connect('example.com', database='default')
with conn.cursor() as cursor:
    cursor.execute('select * from test.persons')
    rows = cursor.fetchall()

Cursors are iterable

from hivejdbc import connect
conn = connect('example.com', database='default')
cursor = conn.cursor()
cursor.execute('select * from test.persons')
for row in cursor:
    print(row[0])
cursor.close()

Cursors Support

  • fetchone()
  • fetchmany()
  • fetchall()
from hivejdbc import connect
conn = connect('example.com', database='default')
cursor = conn.cursor()
cursor.execute('select * from test.persons')
cursor.fetchone() # fetch first row or None
cursor.fetchmany(5) # fetch next 5 rows
cursor.fetchall() # fetch remaining rows or empty list
cursor.close()

Connection Strings

hivejdbc features many connect function arguments. Many of these arguments can be ignored and are simply present to offer the full options provided by the Hive jdbc driver.

To import the hivejdbc connect function:

from hivejdbc import connect

Unsecured Hive Instance

to connect to an unsecured hive instance listening on the default port 10000, and the default database:

conn = connect('example.com', 'default')

unless all required hive-jars are on the classpath already you'll need to define the driver path
Java uses jar files to combine many libraries into one. We'll use our fatjar to provide all the required dependencies in one place.
Make sure you're using the correct driver for your Hive version.

conn = connect('example.com', 'default', driver='hive-client-hive-2.1.1-hdfs-3.0.3-fatjar.jar')

to connect with a custom port of 10015

conn = connect('example.com', 'default', port=10015)

Username and Password

conn = connect(host='example.com', 
               database='default', 
               port=10015, 
               user='hive_user', 
               password='secret')

SSL

If the hive-server has ssl enabled you'll need to provide a jks trust store that contains the servers public certificate.

conn = connect(host='hive2.example.com',
               port=10015,
               database='default',
               driver='hive-client-hive-2.1.1-hdfs-3.0.3-fatjar.jar',
               ssl=True,
               trust_store='./truststore.jks',
               trust_password='changeit',
               principal='hive/hive2.example.com@EXAMPLE.COM',
               user_principal='hive/hive2.example.com',
               user_keytab='hive.keytab',
               realm='EXAMPLE.COM',
               kdc='kerberosdc.example.com:88')

Kerberos

Authenticating with kerberos can be done a few ways:

  1. get valid kerberos credentials via kinit before running hivejdbc
  2. rely on hivejdbc to obtain kerberos credentials via a user-principal and user-keytab provided to the program.

Operating System kinit

Connect to...

  • a ssl enabled cluster
  • a secured cluster (kerberos)
  • using the operating systems kerberos configuration default locations are searched depending on platform
  • using the operating system kinit token
    default locations for the token-cache are searched
  • if kinit has not been performed, or a token-cache cannot be found an exception will be thrown
conn = connect(host='hive2.example.com',
               port=10015,
               database='default',
               driver='hive-client-hive-2.1.1-hdfs-3.0.3-fatjar.jar',
               ssl=True,
               trust_store='./truststore.jks',
               trust_password='changeit',
               principal='hive/hive2.example.com@EXAMPLE.COM')

hivejdbc does the kinit via keytab and a custom krb5.conf

connect to...

  • a ssl enabled cluster
  • a secured cluster (kerberos)
  • using the operating systems kerberos configuration krb5.conf
  • using a keytab for authentication
    the keytab will be used to login via java's built-in kerberos implementation avoiding the need for any operating system dependency
  • we will provide the kdc and realm via the krb5_conf argument if we didn't provide krb5_conf argument default locations would be searched within various system paths
conn = connect(host='hive2.example.com',
               port=10015,
               database='default',
               driver='hive-client-hive-2.1.1-hdfs-3.0.3-fatjar.jar',
               ssl=True,
               trust_store='./truststore.jks',
               trust_password='changeit',
               principal='hive/hive2.example.com@EXAMPLE.COM',
               krb5_conf='kerberos/custom_krb5.conf',
               user_principal='hive/hive2.example.com',
               user_keytab='user.keytab')

hivejdbc does the kinit via keytab with no krb5.conf

connect to...

  • an ssl enabled cluster
  • a secured cluster (kerberos)
  • not using the operating system or relying on any of its configurations
  • manually setting the realm, and the kerberos "kdc" to authenticate to
  • using a keytab for authentication
  • this configuration is the most portable...
conn = connect(host='hive2.example.com',
               port=10015,
               database='default',
               driver='hive-client-hive-2.1.1-hdfs-3.0.3-fatjar.jar',
               ssl=True,
               trust_store='./truststore.jks',
               trust_password='changeit',
               principal='hive/hive2.example.com@EXAMPLE.COM',
               user_principal='hive/hive2.example.com',
               user_keytab='hive.keytab',
               realm='EXAMPLE.COM',
               kdc='kerberosdc.example.com:88')

Queries and Parameters

For these examples we'll setup a test database with a persons table...

cursor = conn.cursor()
cursor.execute('CREATE DATABASE IF NOT EXISTS test')
cursor.execute('DROP TABLE IF EXISTS test.persons')
cursor.execute('CREATE TABLE test.persons (name VARCHAR(64), age INT, address STRING, '
               'first TIMESTAMP, balance DECIMAL(12,2))')

Our table sql will have 5 columns defined in the above statement:

CREATE TABLE test.persons (
    name VARCHAR(64), 
    age INT, 
    address STRING,
    first TIMESTAMP, 
    balance DECIMAL(12,2)
)

single insert

Let's insert a single record:

cursor.execute('''    
INSERT INTO TABLE test.persons (name, age, address, first, balance)
VALUES ('john doe', 35, '1583 Whistling Pines Dr, Redstone CO 80612', '08-22-1981 00:00:00', '100.10')
''')

positional parameterized sql query

Insert a single record, using paramterized arguments that will automatically be escaped.
This prevents sql injection as well

cursor.execute('''    
INSERT INTO TABLE test.persons (name, age, address, first, balance)
VALUES (%s, %s, %s, %s, %s)
''', ['Kevin Jones', 28, '802 1st st, Raleigh NC', '12-23-2020 00:00:00', 85.25])

The signature of execute is:

def execute(sql, params=None):
    ""
  • sql is the sql statement
  • params are named (dict) or positional (sequence) arguments used by the sql statement for variable substitution

named parameterized sql query

INSERT with named parameters

In addition to positional parameters using %s we support named parameters as well.

You can see the named arguments are defined below in the sql statement as: (:name, :age, :addr, :dt, :bal)

The second parameter to the execute method is a dictionary where the keys are equal to the parameters defined in the sql

cursor.execute('''
INSERT INTO TABLE test.persons (name, age, address, first, balance)
VALUES (:name, :age, :addr, :dt, :bal)
''', {'name': 'Bob Clark',
      'age': 41,
      'addr': '348 W Dickinson Rd, Norfolk VA',
      'dt': '12-23-2020 00:00:00',
      'bal': 200.20})

Using executemany

You can execute many queries in one python statement using executemany

Note that this is for programmer ease of use; hive's jdbc driver does not support batch-mode, so this functionality is faked and is no more efficient than executing 3 statements individually.

cursor.executemany('''
INSERT INTO TABLE test.persons (name, age, address, first, balance)
VALUES (%s, %s, %s, %s, %s)
''', [
    ('john doe', 35, '1583 Whistling Pines Dr, Redstone CO 80612', '08-22-1981 00:00:00', 100.10),
    ('Kevin Jones', 28, '802 1st st, Raleigh NC', '12-23-2020 00:00:00', 85.25),
    ('Bob Clark', 41, '348 W Dickinson Rd, Norfolk VA', '12-23-2020 00:00:00', 200.20)
])