Mocking HappyBase for unit testing HBase code

HappyBase is a friendly interface to interact with HBase from Python. It lets you perform basic HBase operations like get, put and scan. But say you have a bunch of puts littered around your code. How do you unit test that? One method would be to mock out the happy base calls themselves, and just assert that they are called with specific parameters. But what if you want to test the final state of the HBase tables after a series of operations?

For that, you can replace the HappyBase Table class with a version that keeps the data in memory.

from collections import defaultdict
import unittest
import happybase

class MockTable(object):

def __init__(self, table_name):
self.table_name = table_name = defaultdict(dict)

def put(self, row, data):[row].update(data)

def row(self, row, columns=None):

def scan(self, **options):
''' does not respect any options like start/stop row '''

class MockConnection(object):
''' singleton object, so that multiple HBaseTables can collaborate '''

_instance = None
tables = dict()

def __new__(cls, *args, **kwargs):
if not cls._instance:
cls._instance = super(MockConnection, cls).__new__(cls, *args, **kwargs)
return cls._instance

def __init__(self, *args, **kwargs):

def is_table_enabled(self, name):
return True

def table(self, name):
table = self.tables.get(name)
if not table:
table = MockTable(name)
self.tables[name] = table
return table

def flush(self):
self.tables = dict()

class HBaseTestCase(unittest.TestCase):
''' mock out calls to hbase
if you over-ride setUp(), make sure to call super '''

def setUp(self):
happybase._Connection = happybase.Connection
happybase.Connection = MockConnection

def tearDown(self):
happybase.Connection = happybase._Connection

To use it, just have your unit test class extend HBaseTestCase.

import happybase

class HappybaseMockTests(HBaseTestCase):

def setUp(self):
connection = happybase.Connection()
self.table = connection.table('table-name')
super(HappybaseMockTests, self).setUp()

def test_put(self):
self.table.put('row1', {'cf1:col1': '1', 'cf1:col2': '2', 'cf2:col1': '3'})
self.assertEquals(self.table.row('row1'), {'cf1:col1': '1', 'cf1:col2': '2', 'cf2:col1': '3'})

def test_scan(self):
self.table.put('row1', {'cf1:col1': '1'})
self.table.put('row1', {'cf1:col2': '2'})
self.table.put('row2', {'cf2:col1': '3'})
self.assertEquals(self.table.scan(), [('row1', {'cf1:col1': '1', 'cf1:col2': '2'}), ('row2', {'cf2:col1': '3'})]

Note: this version does not force you to use valid column families, it just created families on the fly as you put columns in. It also does not support any of the options on scan, such as filtering by a range of rowkeys. Some of those options could be added easily. Others would be very difficult, such as java based column filters. Hopefully this is enough you get you started testing your HBase code.