def libjars(self):
"""Optional list of paths of jar files to run our job with using
Hadoop's ``-libjar`` option. Normally setting :py:attr:`LIBJARS`
is sufficient.
By default, this combines :option:`libjars` options from the command
lines with :py:attr:`LIBJARS`, with command line arguments taking
precedence. Paths from :py:attr:`LIBJARS` are interpreted as relative
to the the directory containing the script (paths from the
command-line are relative to the current working directory).
Note that ``~`` and environment variables in paths will always be
expanded by the job runner (see :mrjob-opt:`libjars`).
.. versionadded:: 0.5.3
"""
script_dir = os.path.dirname(self.mr_job_script())
paths_from_libjars = []
# libjar paths will eventually be combined with combine_path_lists,
# which will expand environment variables. We don't want to assume
# a path like $MY_DIR/some.jar is always relative ($MY_DIR could start
# with /), but we also don't want to expand environment variables
# prematurely.
for path in self.LIBJARS or []:
if os.path.isabs(expand_path(path)):
paths_from_libjars.append(path)
else:
paths_from_libjars.append(os.path.join(script_dir, path))
return combine_lists(paths_from_libjars, self.options.libjars)
def _job_kwargs(self):
"""Keyword arguments to the runner class that can be specified
by the job/launcher itself."""
# use the most basic combiners; leave magic like resolving paths
# and blanking out jobconf values to the runner
return dict(
# command-line has the final say on jobconf and libjars
jobconf=combine_dicts(
self.jobconf(), self.options.jobconf),
libjars=combine_lists(
self.libjars(), self.options.libjars),
partitioner=self.partitioner(),
sort_values=self.sort_values(),
# TODO: should probably put self.options last below for consistency
upload_archives=combine_lists(
self.options.upload_archives, self.archives()),
upload_dirs=combine_lists(
self.options.upload_dirs, self.dirs()),
upload_files=combine_lists(
self.options.upload_files, self.files()),
)
def emr_job_runner_kwargs(self):
args = super(DownloadToS3, self).emr_job_runner_kwargs()
# set up AWS credentials on EMR instances
access_key = os.environ['AWS_ACCESS_KEY_ID']
secret = os.environ['AWS_SECRET_ACCESS_KEY']
args['cmdenv'] = combine_dicts(args['cmdenv'], {'AWS_ACCESS_KEY_ID': access_key, 'AWS_SECRET_ACCESS_KEY': secret})
# install pip, aws-cli, and boto
args['bootstrap_cmds'] = combine_lists(args['bootstrap_cmds'],
['sysctl -w "net.ipv4.tcp_window_scaling=0"',
'sudo apt-get install python-pip',
'sudo pip install awscli',
'sudo pip install boto'])
return args
def test_concatenation(self):
assert_equal(combine_lists([1, 2], None, (3, 4)), [1, 2, 3, 4])
def test_empty(self):
assert_equal(combine_lists(), [])
def test_mix_lists_and_scalars(self):
self.assertEqual(combine_lists([1, 2], 3, (4, 5), 6),
[1, 2, 3, 4, 5, 6])
def test_scalars(self):
self.assertEqual(combine_lists(None, False, b'\x00', 42, 3.14),
[False, b'\x00', 42, 3.14])
def test_strings(self):
self.assertEqual(combine_lists('one', None, 'two', u'three'),
['one', 'two', u'three'])
def test_concatenation(self):
self.assertEqual(combine_lists([1, 2], None, (3, 4)), [1, 2, 3, 4])
def test_strings(self):
self.assertEqual(combine_lists("one", None, "two", u"three"), ["one", "two", u"three"])
def test_empty(self):
self.assertEqual(combine_lists(), [])
def test_dicts(self):
self.assertEqual(combine_lists({1: 2}, None, {}), [{1: 2}, {}])
def test_mix_lists_and_scalars(self):
self.assertEqual(combine_lists([1, 2], 3, (4, 5), 6),
[1, 2, 3, 4, 5, 6])
def test_scalars(self):
self.assertEqual(combine_lists(None, False, b'\x00', 42, 3.14),
[False, b'\x00', 42, 3.14])
def test_dicts(self):
self.assertEqual(combine_lists({1: 2}, None, {}),
[{1: 2}, {}])
def test_strings(self):
self.assertEqual(combine_lists('one', None, 'two', u'three'),
['one', 'two', u'three'])
def test_empty(self):
self.assertEqual(combine_lists(), [])
def test_empty(self):
assert_equal(combine_lists(), [])