class TestTaskController():
def setup_method(self):
# provide mountable tmp directory for docker
tempfile.tempdir = "/tmp" if not platform.system() == "Windows" else None
test_datmo_dir = os.environ.get('TEST_DATMO_DIR',
tempfile.gettempdir())
self.temp_dir = tempfile.mkdtemp(dir=test_datmo_dir)
self.project = ProjectController(self.temp_dir)
self.project.init("test", "test description")
self.environment = EnvironmentController(self.temp_dir)
self.task = TaskController(self.temp_dir)
def teardown_method(self):
pass
def test_create(self):
task_command = ["sh", "-c", "echo accuracy:0.45"]
task_gpu = False
input_dict = {
"command": task_command,
"gpu": task_gpu
}
# Create task in the project
task_obj = self.task.create(input_dict)
assert task_obj
assert task_obj.command == task_command
assert task_obj.gpu == task_gpu
def test_run_helper(self):
# TODO: Try out more options (see below)
# Create environment_driver id
env_def_path = os.path.join(self.project.home,
"Dockerfile")
with open(env_def_path, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
environment_obj = self.environment.create({
"definition_filepath": env_def_path
})
# Set log filepath
log_filepath = os.path.join(self.task.home,
"test.log")
# create volume to mount
temp_test_dirpath = os.path.join(self.temp_dir, "temp")
os.makedirs(temp_test_dirpath)
# Test option set 1
random_name = ''.join([random.choice(string.ascii_letters + string.digits)
for _ in range(32)])
options_dict = {
"command": ["sh", "-c", "echo accuracy:0.45"],
"ports": ["8888:8888"],
"gpu": False,
"name": random_name,
"volumes": {
temp_test_dirpath: {
'bind': '/task/',
'mode': 'rw'
}
},
"detach": False,
"stdin_open": True,
"tty": False,
"api": False
}
return_code, run_id, logs = \
self.task._run_helper(environment_obj.id,
options_dict, log_filepath)
assert return_code == 0
assert run_id and \
self.task.environment_driver.get_container(run_id)
assert logs and \
os.path.exists(log_filepath)
self.task.environment_driver.stop_remove_containers_by_term(term=random_name)
# Test option set 2
random_name_2 = ''.join([random.choice(string.ascii_letters + string.digits)
for _ in range(32)])
options_dict = {
"command": ["sh", "-c", "echo accuracy:0.45"],
"ports": ["8888:8888"],
"gpu": False,
"name": random_name_2 ,
"volumes": {
temp_test_dirpath: {
'bind': '/task/',
'mode': 'rw'
}
},
"detach": False,
"stdin_open": True,
"tty": False,
"api": True
}
return_code, run_id, logs = \
self.task._run_helper(environment_obj.id,
options_dict, log_filepath)
assert return_code == 0
assert run_id and \
self.task.environment_driver.get_container(run_id)
assert logs and \
os.path.exists(log_filepath)
self.task.environment_driver.stop_remove_containers_by_term(term=random_name_2)
def test_parse_logs_for_results(self):
test_logs = """
this is a log
accuracy is good
accuracy : 0.94
this did not work
validation : 0.32
model_type : logistic regression
"""
result = self.task._parse_logs_for_results(test_logs)
assert isinstance(result, dict)
assert result['accuracy'] == "0.94"
assert result['validation'] == "0.32"
assert result['model_type'] == "logistic regression"
def test_run(self):
# 1) Test success case with default values and env def file
# 2) Test failure case if running same task (conflicting containers)
# 3) Test failure case if running same task with snapshot_dict (conflicting containers)
# 4) Test success case with snapshot_dict
# 5) Test success case with saved file during task run
# TODO: look into log filepath randomness, sometimes logs are not written
task_command = ["sh", "-c", "echo accuracy:0.45"]
input_dict = {
"command": task_command
}
# Create task in the project
task_obj = self.task.create(input_dict)
# Create environment definition
env_def_path = os.path.join(self.project.home,
"Dockerfile")
with open(env_def_path, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
# 1) Test option 1
updated_task_obj = self.task.run(task_obj.id)
assert task_obj.id == updated_task_obj.id
assert updated_task_obj.before_snapshot_id
assert updated_task_obj.ports == None
assert updated_task_obj.gpu == False
assert updated_task_obj.interactive == False
assert updated_task_obj.task_dirpath
assert updated_task_obj.log_filepath
assert updated_task_obj.start_time
assert updated_task_obj.after_snapshot_id
assert updated_task_obj.run_id
assert updated_task_obj.logs
assert "accuracy" in updated_task_obj.logs
assert updated_task_obj.results
assert updated_task_obj.results == {"accuracy": "0.45"}
assert updated_task_obj.status == "SUCCESS"
assert updated_task_obj.end_time
assert updated_task_obj.duration
# 2) Test option 2
failed = False
try:
self.task.run(task_obj.id)
except TaskRunException:
failed = True
assert failed
# 3) Test option 3
# Create files to add
self.project.file_driver.create("dirpath1", directory=True)
self.project.file_driver.create("dirpath2", directory=True)
self.project.file_driver.create("filepath1")
# Snapshot dictionary
snapshot_dict = {
"filepaths": [os.path.join(self.project.home, "dirpath1"),
os.path.join(self.project.home, "dirpath2"),
os.path.join(self.project.home, "filepath1")],
}
# Run a basic task in the project
failed = False
try:
self.task.run(task_obj.id,
snapshot_dict=snapshot_dict)
except TaskRunException:
failed = True
assert failed
# Test when the specific task id is already RUNNING
# Create task in the project
task_obj_1 = self.task.create(input_dict)
self.task.dal.task.update({"id": task_obj_1.id, "status": "RUNNING"})
# Create environment_driver definition
env_def_path = os.path.join(self.project.home,
"Dockerfile")
with open(env_def_path, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
failed = False
try:
self.task.run(task_obj_1.id)
except TaskRunException:
failed = True
assert failed
# 4) Test option 4
# Create a new task in the project
task_obj_2 = self.task.create(input_dict)
# Run another task in the project
updated_task_obj_2 = self.task.run(task_obj_2.id,
snapshot_dict=snapshot_dict)
assert task_obj_2.id == updated_task_obj_2.id
assert updated_task_obj_2.before_snapshot_id
assert updated_task_obj_2.ports == None
assert updated_task_obj_2.gpu == False
assert updated_task_obj_2.interactive == False
assert updated_task_obj_2.task_dirpath
assert updated_task_obj_2.log_filepath
assert updated_task_obj_2.start_time
assert updated_task_obj_2.after_snapshot_id
assert updated_task_obj_2.run_id
assert updated_task_obj_2.logs
assert "accuracy" in updated_task_obj_2.logs
assert updated_task_obj_2.results
assert updated_task_obj_2.results == {"accuracy": "0.45"}
assert updated_task_obj_2.status == "SUCCESS"
assert updated_task_obj_2.end_time
assert updated_task_obj_2.duration
# 5) Test option 5
# Create a basic script
# (fails w/ no environment)
test_filepath = os.path.join(self.temp_dir, "script.py")
with open(test_filepath, "w") as f:
f.write(to_unicode("import os\n"))
f.write(to_unicode("import numpy\n"))
f.write(to_unicode("import sklearn\n"))
f.write(to_unicode("print('hello')\n"))
f.write(to_unicode("print(' accuracy: 0.56 ')\n"))
f.write(to_unicode("with open(os.path.join('/task', 'new_file.txt'), 'a') as f:\n"))
f.write(to_unicode(" f.write('my test file')\n"))
task_command = ["python", "script.py"]
input_dict = {
"command": task_command
}
# Create task in the project
task_obj_2 = self.task.create(input_dict)
# Create environment definition
env_def_path = os.path.join(self.project.home,
"Dockerfile")
with open(env_def_path, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
updated_task_obj_2 = self.task.run(task_obj_2.id)
assert updated_task_obj_2.before_snapshot_id
assert updated_task_obj_2.ports == None
assert updated_task_obj_2.gpu == False
assert updated_task_obj_2.interactive == False
assert updated_task_obj_2.task_dirpath
assert updated_task_obj_2.log_filepath
assert updated_task_obj_2.start_time
assert updated_task_obj_2.after_snapshot_id
assert updated_task_obj_2.run_id
assert updated_task_obj_2.logs
assert "accuracy" in updated_task_obj_2.logs
assert updated_task_obj_2.results
assert updated_task_obj_2.results == {"accuracy": "0.56"}
assert updated_task_obj_2.status == "SUCCESS"
assert updated_task_obj_2.end_time
assert updated_task_obj_2.duration
# test if after snapshot has the file written
after_snapshot_obj = self.task.dal.snapshot.get_by_id(
updated_task_obj_2.after_snapshot_id
)
file_collection_obj = self.task.dal.file_collection.get_by_id(
after_snapshot_obj.file_collection_id
)
files_absolute_path = os.path.join(self.task.home, file_collection_obj.path)
assert os.path.isfile(os.path.join(files_absolute_path, "task.log"))
assert os.path.isfile(os.path.join(files_absolute_path, "new_file.txt"))
def test_list(self):
task_command = ["sh", "-c", "echo accuracy:0.45"]
input_dict = {
"command": task_command
}
# Create tasks in the project
task_obj_1 = self.task.create(input_dict)
task_obj_2 = self.task.create(input_dict)
# List all tasks regardless of filters
result = self.task.list()
assert len(result) == 2 and \
task_obj_1 in result and \
task_obj_2 in result
# List all tasks and filter by session
result = self.task.list(session_id=
self.project.current_session.id)
assert len(result) == 2 and \
task_obj_1 in result and \
task_obj_2 in result
def test_get_files(self):
task_command = ["sh", "-c", "echo accuracy:0.45"]
input_dict = {
"command": task_command
}
# Create task in the project
task_obj = self.task.create(input_dict)
# Create environment definition
env_def_path = os.path.join(self.project.home,
"Dockerfile")
with open(env_def_path, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
# Create file to add
self.project.file_driver.create("dirpath1", directory=True)
self.project.file_driver.create(os.path.join("dirpath1", "filepath1"))
# Snapshot dictionary
snapshot_dict = {
"filepaths": [os.path.join(self.project.home, "dirpath1", "filepath1")],
}
# Test the default values
updated_task_obj = self.task.run(task_obj.id,
snapshot_dict=snapshot_dict)
# TODO: Test case for during run and before_snapshot run
# Get files for the task after run is complete (default)
result = self.task.get_files(updated_task_obj.id)
after_snapshot_obj = self.task.dal.snapshot.get_by_id(
updated_task_obj.after_snapshot_id
)
file_collection_obj = self.task.dal.file_collection.get_by_id(
after_snapshot_obj.file_collection_id
)
assert len(result) == 2
assert isinstance(result[0], TextIOWrapper)
assert result[0].name == os.path.join(self.task.home, ".datmo",
"collections",
file_collection_obj.filehash,
"task.log")
assert result[0].mode == "r"
assert isinstance(result[1], TextIOWrapper)
assert result[1].name == os.path.join(self.task.home, ".datmo",
"collections",
file_collection_obj.filehash,
"filepath1")
assert result[1].mode == "r"
# Get files for the task after run is complete for different mode
result = self.task.get_files(updated_task_obj.id, mode="a")
assert len(result) == 2
assert isinstance(result[0], TextIOWrapper)
assert result[0].name == os.path.join(self.task.home, ".datmo",
"collections",
file_collection_obj.filehash,
"task.log")
assert result[0].mode == "a"
assert isinstance(result[1], TextIOWrapper)
assert result[1].name == os.path.join(self.task.home, ".datmo",
"collections",
file_collection_obj.filehash,
"filepath1")
assert result[1].mode == "a"
def test_delete(self):
task_command = ["sh", "-c", "echo accuracy:0.45"]
input_dict = {
"command": task_command
}
# Create tasks in the project
task_obj = self.task.create(input_dict)
# Delete task from the project
result = self.task.delete(task_obj.id)
# Check if task retrieval throws error
thrown = False
try:
self.task.dal.snapshot.get_by_id(task_obj.id)
except EntityNotFound:
thrown = True
assert result == True and \
thrown == True
def test_stop(self):
task_command = ["sh", "-c", "echo accuracy:0.45"]
input_dict = {
"command": task_command
}
# Create task in the project
task_obj = self.task.create(input_dict)
# Create environment driver definition
env_def_path = os.path.join(self.project.home,
"Dockerfile")
with open(env_def_path, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
# Test the default values
updated_task_obj = self.task.run(task_obj.id)
# Stop the task
task_id = updated_task_obj.id
result = self.task.stop(task_id)
# Check if task stop throws error when wrong task id is given
thrown = False
try:
self.task.dal.snapshot.get_by_id(task_obj.id)
except EntityNotFound:
thrown = True
assert result == True and \
thrown == True
class TestEnvironmentController():
def setup_method(self):
self.temp_dir = tempfile.mkdtemp(dir=test_datmo_dir)
Config().set_home(self.temp_dir)
self.project_controller = ProjectController()
def teardown_method(self):
pass
def __setup(self):
self.project_controller.init("test_setup", "test description")
self.environment_controller = EnvironmentController()
with open(os.path.join(self.temp_dir, "test.txt"), "wb") as f:
f.write(to_bytes("hello"))
self.random_filepath = os.path.join(
self.environment_controller.file_driver.environment_directory,
"test")
with open(self.random_filepath, "wb") as f:
f.write(to_bytes("cool"))
self.definition_filepath = os.path.join(
self.environment_controller.file_driver.environment_directory,
"Dockerfile")
with open(self.definition_filepath, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
def test_init_fail_project_not_init(self):
Config().set_home(self.temp_dir)
failed = False
try:
EnvironmentController()
except ProjectNotInitialized:
failed = True
assert failed
def test_get_supported_environments(self):
self.__setup()
result = self.environment_controller.get_supported_environments()
assert result
def test_setup(self):
self.project_controller.init("test_setup", "test description")
self.environment_controller = EnvironmentController()
# Test success setup once (no files present)
options = {"name": "xgboost:cpu"}
result = self.environment_controller.setup(options=options)
output_definition_filepath = os.path.join(
self.environment_controller.file_driver.environment_directory,
"Dockerfile")
assert isinstance(result, Environment)
assert result.name == options['name']
assert result.description == "supported base environment created by datmo"
assert os.path.isfile(output_definition_filepath)
assert "FROM datmo/xgboost:cpu" in open(output_definition_filepath,
"r").read()
# Test success setup again (files present, but staged)
options = {"name": "xgboost:cpu"}
result = self.environment_controller.setup(options=options)
output_definition_filepath = os.path.join(
self.environment_controller.file_driver.environment_directory,
"Dockerfile")
assert isinstance(result, Environment)
assert result.name == options['name']
assert result.description == "supported base environment created by datmo"
assert os.path.isfile(output_definition_filepath)
assert "FROM datmo/xgboost:cpu" in open(output_definition_filepath,
"r").read()
# Test failure in downstream function (e.g. bad inputs, no name given)
failed = False
try:
self.environment_controller.setup(options={})
except EnvironmentDoesNotExist:
failed = True
assert failed
# Change environment file
with open(output_definition_filepath, "wb") as f:
f.write(to_bytes("new content"))
# Test failure setup (unstaged changes)
failed = False
try:
self.environment_controller.setup(options=options)
except UnstagedChanges:
failed = True
assert failed
def test_create(self):
# 0) Test SUCCESS create when definition path exists in project environment directory (no input, no root) -- with hardware file
# 1) Test SUCCESS create when definition path exists in project environment directory (no input, no root)
# 5) Test SUCCESS when definition path exists in project environment directory and passed from input dict (takes input)
# 2) Test SUCCESS create when definition path exists in root project folder (no input, no project environment dir)
# 3) Test SUCCESS create when definition path is passed into input dict (takes input, no project environment dir)
# 4) Test SUCCESS create when definition path is passed into input dict along with expected filename to be saved
# 6) Test FAIL when passing same filepath with same filename into input dict
self.__setup()
input_dict_0 = {"name": "test", "description": "test description"}
# 0) Test option 0 (cannot test hash because hardware is machine-dependent)
environment_obj_0 = self.environment_controller.create(input_dict_0)
assert environment_obj_0
assert isinstance(environment_obj_0, Environment)
assert environment_obj_0.id
assert environment_obj_0.driver_type == "docker"
assert environment_obj_0.file_collection_id
assert environment_obj_0.definition_filename
assert environment_obj_0.hardware_info
assert environment_obj_0.unique_hash
assert environment_obj_0.name == "test"
assert environment_obj_0.description == "test description"
# Get file collection path
file_collection_obj = self.environment_controller.dal.file_collection. \
get_by_id(environment_obj_0.file_collection_id)
file_collection_dir = self.environment_controller.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert os.path.isfile(os.path.join(file_collection_dir, "test"))
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
output = open(os.path.join(file_collection_dir, "Dockerfile"),
"r").read()
print(repr(output))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
output = open(os.path.join(file_collection_dir, "datmoDockerfile"),
"r").read()
print(repr(output))
assert os.path.isfile(
os.path.join(file_collection_dir, "hardware_info"))
output = open(os.path.join(file_collection_dir, "hardware_info"),
"r").read()
print(repr(output))
# 1) Test option 1
environment_obj_0 = self.environment_controller.create(
input_dict_0, save_hardware_file=False)
assert environment_obj_0
assert isinstance(environment_obj_0, Environment)
assert environment_obj_0.id
assert environment_obj_0.driver_type == "docker"
assert environment_obj_0.file_collection_id
assert environment_obj_0.definition_filename
assert environment_obj_0.hardware_info
# Get file collection path
file_collection_obj = self.environment_controller.dal.file_collection. \
get_by_id(environment_obj_0.file_collection_id)
file_collection_dir = self.environment_controller.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert os.path.isfile(os.path.join(file_collection_dir, "test"))
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
output = open(os.path.join(file_collection_dir, "Dockerfile"),
"r").read()
print(repr(output))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
output = open(os.path.join(file_collection_dir, "datmoDockerfile"),
"r").read()
print(repr(output))
assert environment_obj_0.unique_hash == "c309ae4f58163693a91816988d9dc88b"
assert environment_obj_0.name == "test"
assert environment_obj_0.description == "test description"
# Files ["test", "Dockerfile", "datmoDockerfile"]
# 5) Test option 5
input_dict_1 = {
"name": "test",
"description": "test description",
"paths": [self.definition_filepath],
}
environment_obj = self.environment_controller.create(
input_dict_1, save_hardware_file=False)
assert environment_obj
assert isinstance(environment_obj, Environment)
assert environment_obj.id
assert environment_obj.driver_type == "docker"
assert environment_obj.file_collection_id
assert environment_obj.definition_filename
assert environment_obj.hardware_info
# Get file collection path
file_collection_obj = self.environment_controller.dal.file_collection. \
get_by_id(environment_obj.file_collection_id)
file_collection_dir = self.environment_controller.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
output = open(os.path.join(file_collection_dir, "Dockerfile"),
"r").read()
print(repr(output))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
output = open(os.path.join(file_collection_dir, "datmoDockerfile"),
"r").read()
print(repr(output))
assert environment_obj.unique_hash == "6e06d7c4d77cb6ae69e7e0efa883ef4b"
assert environment_obj.name == "test"
assert environment_obj.description == "test description"
# Files ["Dockerfile", "datmoDockerfile"]
# remove the project environment directory
shutil.rmtree(
self.environment_controller.file_driver.environment_directory)
# Create environment definition in root directory
home_definition_filepath = os.path.join(
self.environment_controller.home, "Dockerfile")
with open(home_definition_filepath, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
# 2) Test option 2
environment_obj_1 = self.environment_controller.create(
input_dict_0, save_hardware_file=False)
assert environment_obj_1
assert isinstance(environment_obj_1, Environment)
assert environment_obj_1.id
assert environment_obj_1.driver_type == "docker"
assert environment_obj_1.file_collection_id
assert environment_obj_1.definition_filename
assert environment_obj_1.hardware_info
assert environment_obj_1.unique_hash == file_collection_obj.filehash
# Get file collection path
file_collection_obj = self.environment_controller.dal.file_collection. \
get_by_id(environment_obj_1.file_collection_id)
file_collection_dir = self.environment_controller.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert environment_obj_1.name == "test"
assert environment_obj_1.description == "test description"
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
assert environment_obj_1.unique_hash == "6e06d7c4d77cb6ae69e7e0efa883ef4b"
# 3) Test option 3
input_dict_2 = {
"name": "test",
"description": "test description",
"paths": [home_definition_filepath],
}
# Create environment in the project
environment_obj_2 = self.environment_controller.create(
input_dict_2, save_hardware_file=False)
assert environment_obj_2
assert isinstance(environment_obj_2, Environment)
assert environment_obj_2.id
assert environment_obj_2.driver_type == "docker"
assert environment_obj_2.file_collection_id
assert environment_obj_2.definition_filename
assert environment_obj_2.hardware_info
assert environment_obj_2.unique_hash == file_collection_obj.filehash
# Get file collection path
file_collection_obj = self.environment_controller.dal.file_collection. \
get_by_id(environment_obj_2.file_collection_id)
file_collection_dir = self.environment_controller.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert environment_obj_2.name == "test"
assert environment_obj_2.description == "test description"
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
assert environment_obj_2.unique_hash == "6e06d7c4d77cb6ae69e7e0efa883ef4b"
# 4) Test option 4
input_dict_3 = {
"paths": [home_definition_filepath + ">Dockerfile"],
}
# Create environment in the project
environment_obj_3 = self.environment_controller.create(
input_dict_3, save_hardware_file=False)
assert environment_obj_3
assert isinstance(environment_obj_3, Environment)
assert environment_obj_3.id
assert environment_obj_3.driver_type == "docker"
assert environment_obj_3.file_collection_id
assert environment_obj_3.definition_filename
assert environment_obj_3.hardware_info
assert environment_obj_3.unique_hash == file_collection_obj.filehash
# Get file collection path
file_collection_obj = self.environment_controller.dal.file_collection. \
get_by_id(environment_obj_3.file_collection_id)
file_collection_dir = self.environment_controller.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert environment_obj_3.name == "test"
assert environment_obj_3.description == "test description"
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
assert environment_obj_3.unique_hash == "6e06d7c4d77cb6ae69e7e0efa883ef4b"
# 6) Test option 6
definition_filepath = os.path.join(self.environment_controller.home,
"Dockerfile")
input_dict = {
"paths": [
definition_filepath + ">Dockerfile",
definition_filepath + ">Dockerfile"
],
}
# Create environment in the project
failed = False
try:
_ = self.environment_controller.create(input_dict,
save_hardware_file=False)
except FileAlreadyExistsError:
failed = True
assert failed
@pytest_docker_environment_failed_instantiation(test_datmo_dir)
def test_build(self):
# 1) Test build when no environment given
# 2) Test build when definition path exists and given
# 3) Test build when NO file exists and definition path exists
# 4) Test build when file exists and definition path exists
# 5) Test build when file exists but NO definition path exists
self.project_controller.init("test5", "test description")
self.environment_controller = EnvironmentController()
# 1) Test option 1
failed = False
try:
_ = self.environment_controller.build("does_not_exist")
except EnvironmentDoesNotExist:
failed = True
assert failed
# Create environment definition
definition_filepath = os.path.join(self.environment_controller.home,
"Dockerfile")
random_text = str(uuid.uuid1())
with open(definition_filepath, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine" + "\n"))
f.write(to_bytes(str("RUN echo " + random_text)))
input_dict = {
"paths": [definition_filepath],
}
# 2) Test option 2
# Create environment in the project
environment_obj_1 = self.environment_controller.create(input_dict)
result = self.environment_controller.build(environment_obj_1.id)
assert result
# 3) Test option 3
# Create environment in the project
environment_obj_2 = self.environment_controller.create({})
result = self.environment_controller.build(environment_obj_2.id)
assert result
# Create script to test
test_filepath = os.path.join(self.environment_controller.home,
"script.py")
with open(test_filepath, "wb") as f:
f.write(to_bytes("import numpy\n"))
f.write(to_bytes("import sklearn\n"))
f.write(to_bytes("print('hello')\n"))
# 4) Test option 4
environment_obj_3 = self.environment_controller.create({})
result = self.environment_controller.build(environment_obj_3.id)
assert result
# test 2), 3), and 4) will result in the same environment
assert environment_obj_1.id == environment_obj_2.id
assert environment_obj_2.id == environment_obj_3.id
# Test for building dockerfile when there exists not
os.remove(definition_filepath)
# 5) Test option 5
# Create environment definition in project environment directory
definition_filepath = os.path.join(
self.environment_controller.file_driver.environment_directory,
"Dockerfile")
random_text = str(uuid.uuid1())
with open(definition_filepath, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine" + "\n"))
f.write(to_bytes(str("RUN echo " + random_text)))
environment_obj_4 = self.environment_controller.create({})
result = self.environment_controller.build(environment_obj_4.id)
assert result
# teardown
self.environment_controller.delete(environment_obj_1.id)
@pytest_docker_environment_failed_instantiation(test_datmo_dir)
def test_run(self):
# Test run simple command with simple Dockerfile
self.project_controller.init("test5", "test description")
self.environment_controller = EnvironmentController()
# 0) Test option 0
# Create environment definition in project environment directory
definition_filepath = os.path.join(
self.environment_controller.file_driver.environment_directory,
"Dockerfile")
random_text = str(uuid.uuid1())
with open(definition_filepath, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine" + "\n"))
f.write(to_bytes(str("RUN echo " + random_text)))
random_name = ''.join([
random.choice(string.ascii_letters + string.digits)
for _ in range(32)
])
run_options = {
"command": ["sh", "-c", "echo yo"],
"ports": ["8888:8888"],
"name": random_name,
"volumes": None,
"detach": True,
"stdin_open": False,
"mem_limit": "4g",
"tty": False,
"api": False
}
# Create environment in the project
environment_obj = self.environment_controller.create({})
log_filepath = os.path.join(self.project_controller.home, "task.log")
# Build environment in the project
_ = self.environment_controller.build(environment_obj.id)
# Run environment in the project
return_code, run_id, logs = \
self.environment_controller.run(environment_obj.id, run_options, log_filepath)
assert return_code == 0
assert run_id
assert logs
# teardown
self.environment_controller.delete(environment_obj.id)
shutil.rmtree(
self.environment_controller.file_driver.environment_directory)
# 1) Test option 1
# Create environment definition
definition_filepath = os.path.join(self.environment_controller.home,
"Dockerfile")
random_text = str(uuid.uuid1())
with open(definition_filepath, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine" + "\n"))
f.write(to_bytes(str("RUN echo " + random_text)))
random_name = ''.join([
random.choice(string.ascii_letters + string.digits)
for _ in range(32)
])
run_options = {
"command": ["sh", "-c", "echo yo"],
"ports": ["8888:8888"],
"name": random_name,
"volumes": None,
"mem_limit": "4g",
"detach": True,
"stdin_open": False,
"tty": False,
"api": False
}
input_dict = {
"paths": [definition_filepath],
}
# Create environment in the project
environment_obj = self.environment_controller.create(input_dict)
log_filepath = os.path.join(self.project_controller.home, "task.log")
# Build environment in the project
_ = self.environment_controller.build(environment_obj.id)
# Run environment in the project
return_code, run_id, logs = \
self.environment_controller.run(environment_obj.id, run_options, log_filepath)
assert return_code == 0
assert run_id
assert logs
# teardown
self.environment_controller.delete(environment_obj.id)
# 2) Test option 2
os.remove(definition_filepath)
# Create script to test
test_filepath = os.path.join(self.environment_controller.home,
"script.py")
with open(test_filepath, "wb") as f:
f.write(to_bytes("import os\n"))
f.write(to_bytes("import sys\n"))
f.write(to_bytes("print('hello')\n"))
# Create environment in the project
environment_obj = self.environment_controller.create({})
self.environment_controller.build(environment_obj.id)
random_name = ''.join([
random.choice(string.ascii_letters + string.digits)
for _ in range(32)
])
run_options = {
"command": ["sh", "-c", "echo yo"],
"ports": ["8888:8888"],
"name": random_name,
"volumes": {
self.environment_controller.home: {
'bind': '/home/',
'mode': 'rw'
}
},
"mem_limit": "4g",
"detach": False,
"stdin_open": False,
"tty": False,
"api": False
}
# Run environment in the project
return_code, run_id, logs = \
self.environment_controller.run(environment_obj.id, run_options, log_filepath)
assert return_code == 0
assert run_id
assert logs
# teardown
self.environment_controller.delete(environment_obj.id)
@pytest_docker_environment_failed_instantiation(test_datmo_dir)
def test_interactive_run(self):
# 1) Test run interactive terminal in environment
# 2) Test run jupyter notebook in environment
# Create environment definition
self.project_controller.init("test6", "test description")
self.environment_controller = EnvironmentController()
definition_filepath = os.path.join(self.environment_controller.home,
"Dockerfile")
random_text = str(uuid.uuid1())
with open(definition_filepath, "wb") as f:
f.write(to_bytes("FROM datmo/xgboost:cpu" + "\n"))
f.write(to_bytes(str("RUN echo " + random_text)))
input_dict = {
"paths": [definition_filepath],
}
# Create environment in the project
environment_obj = self.environment_controller.create(input_dict)
# 1) Test option 1
@timeout_decorator.timeout(10, use_signals=False)
def timed_run(container_name, timed_run):
run_options = {
"command": [],
"ports": ["8888:8888"],
"name": container_name,
"volumes": None,
"mem_limit": "4g",
"detach": True,
"stdin_open": True,
"tty": True,
"api": False
}
log_filepath = os.path.join(self.project_controller.home,
"task.log")
# Build environment in the project
_ = self.environment_controller.build(environment_obj.id)
# Run environment in the project
self.environment_controller.run(environment_obj.id, run_options,
log_filepath)
return timed_run
container_name = str(uuid.uuid1())
timed_run_result = False
try:
timed_run_result = timed_run(container_name, timed_run_result)
except timeout_decorator.timeout_decorator.TimeoutError:
timed_run_result = True
assert timed_run_result
# teardown
self.environment_controller.delete(environment_obj.id)
# 2) Test option 2
environment_obj = self.environment_controller.create(input_dict)
@timeout_decorator.timeout(10, use_signals=False)
def timed_run(container_name, timed_run):
run_options = {
"command": ["jupyter", "notebook"],
"ports": ["8888:8888"],
"name": container_name,
"volumes": None,
"mem_limit": "4g",
"detach": True,
"stdin_open": False,
"tty": False,
"api": False
}
log_filepath = os.path.join(self.project_controller.home,
"task.log")
# Build environment in the project
_ = self.environment_controller.build(environment_obj.id)
# Run environment in the project
self.environment_controller.run(environment_obj.id, run_options,
log_filepath)
return timed_run
container_name = str(uuid.uuid1())
timed_run_result = False
try:
timed_run_result = timed_run(container_name, timed_run_result)
except timeout_decorator.timeout_decorator.TimeoutError:
timed_run_result = True
assert timed_run_result
# Stop the running environment
# self.environment_controller.stop(container_name)
# teardown
self.environment_controller.delete(environment_obj.id)
def test_list(self):
self.project_controller.init("test4", "test description")
self.environment_controller = EnvironmentController()
# Create environment definition for object 1
definition_path_1 = os.path.join(self.environment_controller.home,
"Dockerfile")
with open(definition_path_1, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
input_dict_1 = {
"paths": [definition_path_1],
}
# Create environment in the project
environment_obj_1 = self.environment_controller.create(input_dict_1)
# Create environment definition for object 2
definition_path_2 = os.path.join(self.environment_controller.home,
"Dockerfile2")
with open(definition_path_2, "wb") as f:
f.write(to_bytes("FROM python:3.4-alpine"))
input_dict_2 = {
"paths": [definition_path_2 + ">Dockerfile"],
}
# Create second environment in the project
environment_obj_2 = self.environment_controller.create(input_dict_2)
# List all environments and ensure they exist
result = self.environment_controller.list()
assert len(result) == 2 and \
environment_obj_1 in result and \
environment_obj_2 in result
def test_update(self):
self.project_controller.init("test5", "test description")
self.environment_controller = EnvironmentController()
# Create environment definition
definition_filepath = os.path.join(self.environment_controller.home,
"Dockerfile")
with open(definition_filepath, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
input_dict = {
"paths": [definition_filepath],
}
# Create environment in the project
environment_obj = self.environment_controller.create(input_dict)
# Test success update
new_name = "test name"
new_description = "test description"
result = self.environment_controller.update(
environment_obj.id, name=new_name, description=new_description)
assert result
assert isinstance(result, Environment)
assert result.name == new_name
assert result.description == new_description
# Test failed update
failed = False
try:
self.environment_controller.update("random_id",
name=new_name,
description=new_description)
except EnvironmentDoesNotExist:
failed = True
assert failed
@pytest_docker_environment_failed_instantiation(test_datmo_dir)
def test_delete(self):
self.project_controller.init("test5", "test description")
self.environment_controller = EnvironmentController()
# Create environment definition
definition_filepath = os.path.join(self.environment_controller.home,
"Dockerfile")
with open(definition_filepath, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
input_dict = {
"paths": [definition_filepath],
}
# Create environment in the project
environment_obj = self.environment_controller.create(input_dict)
# Delete environment in the project
result = self.environment_controller.delete(environment_obj.id)
# Check if environment retrieval throws error
thrown = False
try:
self.environment_controller.dal.environment.get_by_id(
environment_obj.id)
except EntityNotFound:
thrown = True
assert result == True and \
thrown == True
@pytest_docker_environment_failed_instantiation(test_datmo_dir)
def test_stop_failure(self):
# 1) Test failure with RequiredArgumentMissing
# 2) Test failure with TooManyArgumentsFound
self.project_controller.init("test5", "test description")
self.environment_controller = EnvironmentController()
# 1) Test option 1
failed = False
try:
self.environment_controller.stop()
except RequiredArgumentMissing:
failed = True
assert failed
# 2) Test option 2
failed = False
try:
self.environment_controller.stop(run_id="hello",
match_string="there")
except TooManyArgumentsFound:
failed = True
assert failed
@pytest_docker_environment_failed_instantiation(test_datmo_dir)
def test_stop_success(self):
# TODO: test more run options
# 1) Test run_id input to stop
# 2) Test match_string input to stop
# 3) Test all input to stop
self.project_controller.init("test5", "test description")
self.environment_controller = EnvironmentController()
# Create environment definition
definition_filepath = os.path.join(self.environment_controller.home,
"Dockerfile")
with open(definition_filepath, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
run_options = {
"command": ["sh", "-c", "echo yo"],
"ports": ["8888:8888"],
"name": "datmo-task-" + self.environment_controller.model.id +
"-" + "test",
"volumes": None,
"mem_limit": "4g",
"detach": False,
"stdin_open": False,
"tty": False,
"api": False
}
# Create environment definition
env_def_path = os.path.join(self.project_controller.home, "Dockerfile")
random_text = str(uuid.uuid1())
with open(env_def_path, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine" + "\n"))
f.write(to_bytes(str("RUN echo " + random_text)))
input_dict = {
"paths": [definition_filepath],
}
# Create environment in the project
environment_obj = self.environment_controller.create(input_dict)
log_filepath = os.path.join(self.project_controller.home, "task.log")
# Build environment in the project
_ = self.environment_controller.build(environment_obj.id)
# 1) Test option 1
_, run_id, _ = \
self.environment_controller.run(environment_obj.id, run_options, log_filepath)
return_code = self.environment_controller.stop(run_id=run_id)
assert return_code
# 2) Test option 2
_, _, _ = \
self.environment_controller.run(environment_obj.id, run_options, log_filepath)
return_code = self.environment_controller.stop(
match_string="datmo-task-" + self.environment_controller.model.id)
assert return_code
# 3) Test option 3
_, _, _ = \
self.environment_controller.run(environment_obj.id, run_options, log_filepath)
run_options_2 = {
"command": ["sh", "-c", "echo yo"],
"ports": ["8888:8888"],
"name": "datmo-task-" + self.environment_controller.model.id +
"-" + "test2",
"volumes": None,
"mem_limit": "4g",
"detach": False,
"stdin_open": False,
"tty": False,
"api": False
}
_, _, _ = \
self.environment_controller.run(environment_obj.id, run_options_2, log_filepath)
return_code = self.environment_controller.stop(all=True)
assert return_code
# teardown
self.environment_controller.delete(environment_obj.id)
def test_exists_env(self):
# Test failure, not initialized
failed = False
try:
_ = self.environment_controller.create({})
except:
failed = True
assert failed
# Setup
self.__setup()
environment_obj = self.environment_controller.create({})
# Check by environment id
result = self.environment_controller.exists(
environment_id=environment_obj.id)
assert result
# Check by unique hash
result = self.environment_controller.exists(
environment_unique_hash=environment_obj.unique_hash)
assert result
# Test with wrong environment id
result = self.environment_controller.exists(
environment_id='test_wrong_env_id')
assert not result
def test_calculate_project_environment_hash(self):
# Setup
self.__setup()
# Test hashing the default (with hardware info)
result = self.environment_controller._calculate_project_environment_hash(
)
assert result
# Test hashing the default Dockerfile
result = self.environment_controller._calculate_project_environment_hash(
save_hardware_file=False)
assert result == "c309ae4f58163693a91816988d9dc88b"
# Test if hash is the same as that of create
environment_obj = self.environment_controller.create(
{}, save_hardware_file=False)
result = self.environment_controller._calculate_project_environment_hash(
save_hardware_file=False)
assert result == "c309ae4f58163693a91816988d9dc88b"
assert result == environment_obj.unique_hash
# Test if the hash is the same if the same file is passed in as an input
input_dict = {
"paths": [self.definition_filepath, self.random_filepath]
}
environment_obj_1 = self.environment_controller.create(
input_dict, save_hardware_file=False)
result = self.environment_controller._calculate_project_environment_hash(
save_hardware_file=False)
assert result == "c309ae4f58163693a91816988d9dc88b"
assert result == environment_obj_1.unique_hash
def test_has_unstaged_changes(self):
# Setup
self.__setup()
_ = self.environment_controller.create({})
# Check for no unstaged changes
result = self.environment_controller._has_unstaged_changes()
assert not result
# Make a change to the file (update python version)
with open(
os.path.join(
self.environment_controller.file_driver.
environment_directory, "Dockerfile"), "wb") as f:
f.write(to_bytes("FROM python:3.6-alpine"))
# Check again, should have unstaged changes
result = self.environment_controller._has_unstaged_changes()
assert result
def test_check_unstaged_changes(self):
# Setup
self.__setup()
obj = self.environment_controller.create({})
# 1) After commiting the changes
# Check for no unstaged changes because already committed
result = self.environment_controller.check_unstaged_changes()
assert not result
# Add a new file
with open(
os.path.join(
self.environment_controller.file_driver.
environment_directory, "test2"), "wb") as f:
f.write(to_bytes("cool"))
# 2) Not commiting the changes, should error and raise UnstagedChanges
failed = False
try:
self.environment_controller.check_unstaged_changes()
except UnstagedChanges:
failed = True
assert failed
# Remove new file
os.remove(
os.path.join(
self.environment_controller.file_driver.environment_directory,
"test2"))
# 3) Files are the same as before but no new commit, should have no unstaged changes
result = self.environment_controller.check_unstaged_changes()
assert not result
# 4) Remove another file, now it is different and should have unstaged changes
os.remove(
os.path.join(
self.environment_controller.file_driver.environment_directory,
"test"))
failed = False
try:
self.environment_controller.check_unstaged_changes()
except UnstagedChanges:
failed = True
assert failed
# 5) Remove the rest of the files, now it is empty and should return as already staged
os.remove(
os.path.join(
self.environment_controller.file_driver.environment_directory,
"Dockerfile"))
result = self.environment_controller.check_unstaged_changes()
assert not result
def test_checkout(self):
# Setup and create all environment files
self.__setup()
# Create environment to checkout to with defaults
environment_obj = self.environment_controller.create({})
# Checkout success with there are no unstaged changes
result = self.environment_controller.checkout(environment_obj.id)
assert result
current_hash = self.environment_controller._calculate_project_environment_hash(
)
assert environment_obj.unique_hash == current_hash
# Check the filenames as well because the hash does not take this into account
assert os.path.isfile(
os.path.join(
self.environment_controller.file_driver.environment_directory,
"test"))
assert os.path.isfile(
os.path.join(
self.environment_controller.file_driver.environment_directory,
"Dockerfile"))
assert not os.path.isfile(
os.path.join(
self.environment_controller.file_driver.environment_directory,
"datmoDockerfile"))
assert not os.path.isfile(
os.path.join(
self.environment_controller.file_driver.environment_directory,
"hardware_info"))
# Change file contents to make it unstaged
with open(self.definition_filepath, "wb") as f:
f.write(to_bytes("new content"))
# Checkout failure with unstaged changes
failed = False
try:
_ = self.environment_controller.checkout(environment_obj.id)
except UnstagedChanges:
failed = True
assert failed
# Create new environment to checkout to with defaults (no hardware)
environment_obj_1 = self.environment_controller.create(
{}, save_hardware_file=False)
# Checkout success with there are no unstaged changes
result = self.environment_controller.checkout(environment_obj.id)
assert result
current_hash = self.environment_controller._calculate_project_environment_hash(
)
assert environment_obj.unique_hash == current_hash
assert environment_obj_1.unique_hash != current_hash
# Check the filenames as well because the hash does not take this into account
assert os.path.isfile(
os.path.join(
self.environment_controller.file_driver.environment_directory,
"test"))
assert os.path.isfile(
os.path.join(
self.environment_controller.file_driver.environment_directory,
"Dockerfile"))
assert not os.path.isfile(
os.path.join(
self.environment_controller.file_driver.environment_directory,
"datmoDockerfile"))
assert not os.path.isfile(
os.path.join(
self.environment_controller.file_driver.environment_directory,
"hardware_info"))
class TestEnvironmentController():
def setup_method(self):
# provide mountable tmp directory for docker
tempfile.tempdir = "/tmp" if not platform.system(
) == "Windows" else None
test_datmo_dir = os.environ.get('TEST_DATMO_DIR',
tempfile.gettempdir())
self.temp_dir = tempfile.mkdtemp(dir=test_datmo_dir)
self.project = ProjectController(self.temp_dir)
self.environment = EnvironmentController(self.temp_dir)
def teardown_method(self):
pass
def test_create(self):
# 0) Test create when unsupported language given
# 1) Test create when NO file exists and NO definition path exists
# 2) Test create when NO file exists and definition path exists
# 3) Test create when definition path exists and given
# 4) Test create when file exists and definition path exists
# 5) Test create when file exists but NO definition path exists
# 6) Test create when definition path exists and given for NEW definition filepath
self.project.init("test3", "test description")
# 0) Test option 0
try:
self.environment.create({"language": "java"})
except EnvironmentDoesNotExist:
failed = True
assert failed
# 1) Test option 1
failed = False
try:
self.environment.create({})
except EnvironmentDoesNotExist:
failed = True
assert failed
# Create environment definition
definition_filepath = os.path.join(self.environment.home, "Dockerfile")
with open(definition_filepath, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
# 2) Test option 2
environment_obj_1 = self.environment.create({})
# Get file collection path
file_collection_obj = self.environment.dal.file_collection. \
get_by_id(environment_obj_1.file_collection_id)
file_collection_dir = self.environment.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert environment_obj_1
assert environment_obj_1.id
assert environment_obj_1.driver_type == "docker"
assert environment_obj_1.file_collection_id
assert environment_obj_1.definition_filename
assert environment_obj_1.hardware_info
assert environment_obj_1.unique_hash == file_collection_obj.filehash
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "hardware_info"))
# 3) Test option 3
input_dict = {
"definition_filepath": definition_filepath,
}
# Create environment in the project
environment_obj_2 = self.environment.create(input_dict)
# Get file collection path
file_collection_obj = self.environment.dal.file_collection. \
get_by_id(environment_obj_2.file_collection_id)
file_collection_dir = self.environment.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert environment_obj_2
assert environment_obj_2.id
assert environment_obj_2.driver_type == "docker"
assert environment_obj_2.file_collection_id
assert environment_obj_2.definition_filename
assert environment_obj_2.hardware_info
assert environment_obj_2.unique_hash == file_collection_obj.filehash
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "hardware_info"))
# Create script to test
test_filepath = os.path.join(self.environment.home, "script.py")
with open(test_filepath, "w") as f:
f.write(to_unicode("import numpy\n"))
f.write(to_unicode("import sklearn\n"))
f.write(to_unicode("print('hello')\n"))
# 4) Test option 4
environment_obj_3 = self.environment.create({})
# Get file collection path
file_collection_obj = self.environment.dal.file_collection. \
get_by_id(environment_obj_3.file_collection_id)
file_collection_dir = self.environment.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert environment_obj_3
assert environment_obj_3.id
assert environment_obj_3.driver_type == "docker"
assert environment_obj_3.file_collection_id
assert environment_obj_3.definition_filename
assert environment_obj_3.hardware_info
assert environment_obj_3.unique_hash == file_collection_obj.filehash
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "hardware_info"))
# Remove definition filepath
os.remove(definition_filepath)
assert environment_obj_1.id == environment_obj_2.id
assert environment_obj_2.id == environment_obj_3.id
# 5) Test option 5
environment_obj_4 = self.environment.create({})
# Get file collection path
file_collection_obj = self.environment.dal.file_collection. \
get_by_id(environment_obj_4.file_collection_id)
file_collection_dir = self.environment.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert environment_obj_4
assert environment_obj_4.id
assert environment_obj_4.driver_type == "docker"
assert environment_obj_4.file_collection_id
assert environment_obj_4.definition_filename
assert environment_obj_4.hardware_info
assert environment_obj_4.unique_hash == file_collection_obj.filehash
assert os.path.isfile(
os.path.join(file_collection_dir, "requirements.txt"))
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "hardware_info"))
assert environment_obj_1.id != environment_obj_4.id
# 6) Test option 6
# Create environment definition
definition_filepath = os.path.join(self.environment.home, "Dockerfile")
with open(definition_filepath, "w") as f:
f.write(to_unicode(str("FROM cloudgear/ubuntu:14.04")))
input_dict = {
"definition_filepath": definition_filepath,
}
# Create a new environment obj
environment_obj_5 = self.environment.create(input_dict)
# Get file collection path
file_collection_obj = self.environment.dal.file_collection. \
get_by_id(environment_obj_5.file_collection_id)
file_collection_dir = self.environment.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert environment_obj_5
assert environment_obj_5.id
assert environment_obj_5.driver_type == "docker"
assert environment_obj_5.file_collection_id
assert environment_obj_5.definition_filename
assert environment_obj_5.hardware_info
assert environment_obj_5.unique_hash == file_collection_obj.filehash
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "hardware_info"))
assert environment_obj_5.id != environment_obj_1.id
assert environment_obj_5.id != environment_obj_4.id
def test_build(self):
# 1) Test build when no environment given
# 2) Test build when definition path exists and given
# 3) Test build when NO file exists and definition path exists
# 4) Test build when file exists and definition path exists
# 5) Test build when file exists but NO definition path exists
self.project.init("test5", "test description")
# 1) Test option 1
failed = False
try:
_ = self.environment.build("does_not_exist")
except EntityNotFound:
failed = True
assert failed
# Create environment definition
definition_filepath = os.path.join(self.environment.home, "Dockerfile")
with open(definition_filepath, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
input_dict = {
"definition_filepath": definition_filepath,
}
# 2) Test option 2
# Create environment in the project
environment_obj_1 = self.environment.create(input_dict)
result = self.environment.build(environment_obj_1.id)
assert result
# 3) Test option 3
# Create environment in the project
environment_obj_2 = self.environment.create({})
result = self.environment.build(environment_obj_2.id)
assert result
# Create script to test
test_filepath = os.path.join(self.environment.home, "script.py")
with open(test_filepath, "w") as f:
f.write(to_unicode("import numpy\n"))
f.write(to_unicode("import sklearn\n"))
f.write(to_unicode("print('hello')\n"))
# 4) Test option 4
environment_obj_3 = self.environment.create({})
result = self.environment.build(environment_obj_3.id)
assert result
# test 2), 3), and 4) will result in the same environment
assert environment_obj_1.id == environment_obj_2.id
assert environment_obj_2.id == environment_obj_3.id
# Test for building dockerfile when there exists not
os.remove(definition_filepath)
# 5) Test option 5
environment_obj_4 = self.environment.create({})
result = self.environment.build(environment_obj_4.id)
assert result
assert environment_obj_4.id != environment_obj_1.id
# teardown
self.environment.delete(environment_obj_1.id)
self.environment.delete(environment_obj_4.id)
def test_run(self):
# 1) Test run simple command with simple Dockerfile
# 2) Test run script, with autogenerated definition
self.project.init("test5", "test description")
# 1) Test option 1
# Create environment definition
definition_filepath = os.path.join(self.environment.home, "Dockerfile")
with open(definition_filepath, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
run_options = {
"command": ["sh", "-c", "echo yo"],
"ports": ["8888:8888"],
"name": None,
"volumes": None,
"detach": False,
"stdin_open": False,
"tty": False,
"gpu": False,
"api": False
}
input_dict = {
"definition_filepath": definition_filepath,
}
# Create environment in the project
environment_obj = self.environment.create(input_dict)
log_filepath = os.path.join(self.project.home, "task.log")
# Build environment in the project
_ = self.environment.build(environment_obj.id)
# Run environment in the project
return_code, run_id, logs = \
self.environment.run(environment_obj.id, run_options, log_filepath)
assert return_code == 0
assert run_id
assert logs
# teardown
self.environment.delete(environment_obj.id)
# 2) Test option 2
os.remove(definition_filepath)
# Create script to test
test_filepath = os.path.join(self.environment.home, "script.py")
with open(test_filepath, "w") as f:
f.write(to_unicode("import numpy\n"))
f.write(to_unicode("import sklearn\n"))
f.write(to_unicode("print('hello')\n"))
# Create environment in the project
environment_obj = self.environment.create({})
self.environment.build(environment_obj.id)
run_options = {
"command": ["python", "script.py"],
"ports": ["8888:8888"],
"name": None,
"volumes": {
self.environment.home: {
'bind': '/home/',
'mode': 'rw'
}
},
"detach": False,
"stdin_open": False,
"tty": False,
"gpu": False,
"api": False
}
# Run environment in the project
return_code, run_id, logs = \
self.environment.run(environment_obj.id, run_options, log_filepath)
assert return_code == 0
assert run_id
assert logs
# teardown
self.environment.delete(environment_obj.id)
def test_list(self):
self.project.init("test4", "test description")
# Create environment definition for object 1
definition_path_1 = os.path.join(self.environment.home, "Dockerfile")
with open(definition_path_1, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
input_dict_1 = {
"definition_filepath": definition_path_1,
}
# Create environment in the project
environment_obj_1 = self.environment.create(input_dict_1)
# Create environment definition for object 2
definition_path_2 = os.path.join(self.environment.home, "Dockerfile2")
with open(definition_path_2, "w") as f:
f.write(to_unicode(str("FROM datmo/scikit-opencv")))
input_dict_2 = {
"definition_filepath": definition_path_2,
}
# Create second environment in the project
environment_obj_2 = self.environment.create(input_dict_2)
# List all environments and ensure they exist
result = self.environment.list()
assert len(result) == 2 and \
environment_obj_1 in result and \
environment_obj_2 in result
def test_delete(self):
self.project.init("test5", "test description")
# Create environment definition
definition_filepath = os.path.join(self.environment.home, "Dockerfile")
with open(definition_filepath, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
input_dict = {
"definition_filepath": definition_filepath,
}
# Create environment in the project
environment_obj = self.environment.create(input_dict)
# Delete environment in the project
result = self.environment.delete(environment_obj.id)
# Check if environment retrieval throws error
thrown = False
try:
self.environment.dal.environment.get_by_id(environment_obj.id)
except EntityNotFound:
thrown = True
assert result == True and \
thrown == True
def test_stop(self):
self.project.init("test5", "test description")
# Create environment definition
definition_filepath = os.path.join(self.environment.home, "Dockerfile")
with open(definition_filepath, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
run_options = {
"command": ["sh", "-c", "echo yo"],
"ports": ["8888:8888"],
"name": None,
"volumes": None,
"detach": False,
"stdin_open": False,
"tty": False,
"gpu": False,
"api": False
}
# Create environment_driver definition
env_def_path = os.path.join(self.project.home, "Dockerfile")
with open(env_def_path, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
input_dict = {
"definition_filepath": definition_filepath,
}
# Create environment in the project
environment_obj = self.environment.create(input_dict)
log_filepath = os.path.join(self.project.home, "task.log")
# Build environment in the project
_ = self.environment.build(environment_obj.id)
# Run environment in the project
_, run_id, _ = \
self.environment.run(environment_obj.id, run_options, log_filepath)
# Stop the running environment
return_code = self.environment.stop(run_id)
assert return_code
# teardown
self.environment.delete(environment_obj.id)
class SnapshotController(BaseController):
"""SnapshotController inherits from BaseController and manages business logic related to snapshots
Parameters
----------
home : str
home path of the project
Attributes
----------
code : datmo.core.controller.code.code.CodeController
file_collection : datmo.core.controller.file.file_collection.FileCollectionController
environment : datmo.core.controller.environment.environment.EnvironmentController
Methods
-------
create(dictionary)
Create a snapshot within the project
checkout(id)
Checkout to a specific snapshot within the project
list(session_id=None)
List all snapshots present within the project based on given filters
delete(id)
Delete the snapshot specified from the project
"""
def __init__(self):
super(SnapshotController, self).__init__()
self.code = CodeController()
self.file_collection = FileCollectionController()
self.environment = EnvironmentController()
if not self.is_initialized:
raise ProjectNotInitialized(
__("error", "controller.snapshot.__init__"))
def create(self, dictionary):
"""Create snapshot object
Parameters
----------
dictionary : dict
for each of the 5 key components, this function will search for
one of the variables below starting from the top. Default functionality
is described below for each component as well for reference if none
of the variables are given.
code :
code_id : str, optional
code reference associated with the snapshot; if not
provided will look to inputs below for code creation
commit_id : str, optional
commit id provided by the user if already available
Default
-------
commits will be taken and code created via the CodeController
and are added to the snapshot at the time of snapshot creation
environment :
environment_id : str, optional
id for environment used to create snapshot
environment_paths : list, optional
list of absolute or relative filepaths and/or dirpaths to collect with destination names
(e.g. "/path/to/file>hello", "/path/to/file2", "/path/to/dir>newdir")
Default
-------
default environment files will be searched and environment will
be created with the EnvironmentController and added to the snapshot
at the time of snapshot creation
file_collection :
file_collection_id : str, optional
file collection associated with the snapshot
paths : list, optional
list of absolute or relative filepaths and/or dirpaths to collect with destination names
(e.g. "/path/to/file:hello", "/path/to/file2", "/path/to/dir:newdir")
Default
-------
paths will be considered empty ([]), and the FileCollectionController
will create a blank FileCollection that is empty.
config :
config : dict, optional
key, value pairs of configurations
config_filepath : str, optional
absolute filepath to configuration parameters file
config_filename : str, optional
name of file with configuration parameters
Default
-------
config will be considered empty ({}) and saved to the snapshot
stats :
stats : dict, optional
key, value pairs of metrics and statistics
stats_filepath : str, optional
absolute filepath to stats parameters file
stats_filename : str, optional
name of file with metrics and statistics.
Default
-------
stats will be considered empty ({}) and saved to the snapshot
for the remaining optional arguments it will search for them
in the input dictionary
message : str
long description of snapshot
session_id : str, optional
session id within which snapshot is created,
will overwrite default if given
task_id : str, optional
task id associated with snapshot
label : str, optional
short description of snapshot
visible : bool, optional
True if visible to user via list command else False
Returns
-------
datmo.core.entity.snapshot.Snapshot
snapshot object with all relevant parameters
Raises
------
RequiredArgumentMissing
if required arguments are not given by the user
FileIOError
if files are not present or there is an error in File IO
"""
# Validate Inputs
create_dict = {
"model_id": self.model.id,
"session_id": self.current_session.id,
}
validate("create_snapshot", dictionary)
# Message must be present
if "message" in dictionary:
create_dict['message'] = dictionary['message']
else:
raise RequiredArgumentMissing(
__("error", "controller.snapshot.create.arg", "message"))
# Code setup
self._code_setup(dictionary, create_dict)
# Environment setup
self._env_setup(dictionary, create_dict)
# File setup
self._file_setup(dictionary, create_dict)
# Config setup
self._config_setup(dictionary, create_dict)
# Stats setup
self._stats_setup(dictionary, create_dict)
# If snapshot object with required args already exists, return it
# DO NOT create a new snapshot with the same required arguments
results = self.dal.snapshot.query({
"model_id":
create_dict["model_id"],
"code_id":
create_dict['code_id'],
"environment_id":
create_dict['environment_id'],
"file_collection_id":
create_dict['file_collection_id'],
"config":
create_dict['config'],
"stats":
create_dict['stats']
})
if results: return results[0]
# Optional args for Snapshot entity
optional_args = ["task_id", "label", "visible"]
for optional_arg in optional_args:
if optional_arg in dictionary:
create_dict[optional_arg] = dictionary[optional_arg]
# Create snapshot and return
return self.dal.snapshot.create(Snapshot(create_dict))
def create_from_task(self,
message,
task_id,
label=None,
config=None,
stats=None):
"""Create snapshot from a completed task.
# TODO: enable create from task DURING a run
Parameters
----------
message : str
long description of snapshot
task_id : str
task object to use to create snapshot
label: str, optional
short description of snapshot
config : dict, optional
key, value pairs of configurations
stats : dict, optional
key, value pairs of metrics and statistics
Returns
-------
datmo.core.entity.snapshot.Snapshot
snapshot object with all relevant parameters
Raises
------
TaskNotComplete
if task specified has not been completed
"""
validate(
"create_snapshot_from_task", {
"message": message,
"task_id": task_id,
"label": label,
"config": config,
"stats": stats
})
task_obj = self.dal.task.get_by_id(task_id)
if not task_obj.status and not task_obj.after_snapshot_id:
raise TaskNotComplete(
__("error", "controller.snapshot.create_from_task",
str(task_obj.id)))
after_snapshot_obj = self.dal.snapshot.get_by_id(
task_obj.after_snapshot_id)
snapshot_update_dict = {
"id": task_obj.after_snapshot_id,
"message": message,
"visible": True
}
if label:
snapshot_update_dict["label"] = label
if config:
snapshot_update_dict["config"] = config
if stats:
snapshot_update_dict["stats"] = stats
else:
# Append to any existing stats already present
snapshot_update_dict["stats"] = {}
if after_snapshot_obj.stats is not None:
snapshot_update_dict["stats"].update(after_snapshot_obj.stats)
if task_obj.results is not None:
snapshot_update_dict["stats"].update(task_obj.results)
if snapshot_update_dict["stats"] == {}:
snapshot_update_dict["stats"] = None
return self.dal.snapshot.update(snapshot_update_dict)
def checkout(self, snapshot_id):
# Get snapshot object
snapshot_obj = self.dal.snapshot.get_by_id(snapshot_id)
code_obj = self.dal.code.get_by_id(snapshot_obj.code_id)
file_collection_obj = self.dal.file_collection.\
get_by_id(snapshot_obj.file_collection_id)
environment_obj = self.dal.environment. \
get_by_id(snapshot_obj.environment_id)
# check for unstaged changes in code
self.code_driver.check_unstaged_changes()
# check for unstaged changes in environment
self.environment.check_unstaged_changes()
# check for unstaged changes in file
self.file_collection.check_unstaged_changes()
# Checkout code_driver to the relevant commit ref
code_checkout_success = self.code_driver.checkout_ref(
code_obj.commit_id)
# Checkout environment_driver to relevant environment id
environment_checkout_success = self.environment.checkout(
environment_obj.id)
# Checkout file_driver to relevant file collection id
file_checkout_success = self.file_collection.checkout(
file_collection_obj.id)
return (code_checkout_success and environment_checkout_success
and file_checkout_success)
def list(self,
session_id=None,
visible=None,
sort_key=None,
sort_order=None):
query = {}
if session_id:
try:
self.dal.session.get_by_id(session_id)
except EntityNotFound:
raise SessionDoesNotExist(
__("error", "controller.snapshot.list", session_id))
query['session_id'] = session_id
if visible is not None and isinstance(visible, bool):
query['visible'] = visible
return self.dal.snapshot.query(query, sort_key, sort_order)
def update(self,
snapshot_id,
config=None,
stats=None,
message=None,
label=None,
visible=None):
"""Update the snapshot metadata"""
if not snapshot_id:
raise RequiredArgumentMissing(
__("error", "controller.snapshot.delete.arg", "snapshot_id"))
update_snapshot_input_dict = {'id': snapshot_id}
validate(
"update_snapshot", {
"config": config,
"stats": stats,
"message": message,
"label": label,
"visible": visible
})
if config is not None:
update_snapshot_input_dict['config'] = config
if stats is not None:
update_snapshot_input_dict['stats'] = stats
if message is not None:
update_snapshot_input_dict['message'] = message
if label is not None:
update_snapshot_input_dict['label'] = label
if visible is not None:
update_snapshot_input_dict['visible'] = visible
return self.dal.snapshot.update(update_snapshot_input_dict)
def get(self, snapshot_id):
"""Get snapshot object and return
Parameters
----------
snapshot_id : str
id for the snapshot you would like to get
Returns
-------
datmo.core.entity.snapshot.Snapshot
core snapshot object
Raises
------
DoesNotExist
snapshot does not exist
"""
try:
return self.dal.snapshot.get_by_id(snapshot_id)
except EntityNotFound:
raise DoesNotExist()
def get_files(self, snapshot_id, mode="r"):
"""Get list of file objects for snapshot id.
Parameters
----------
snapshot_id : str
id for the snapshot you would like to get file objects for
mode : str
file open mode
(default is "r" to open file for read)
Returns
-------
list
list of python file objects
Raises
------
DoesNotExist
snapshot object does not exist
"""
try:
snapshot_obj = self.dal.snapshot.get_by_id(snapshot_id)
except EntityNotFound:
raise DoesNotExist()
file_collection_obj = self.dal.file_collection.get_by_id(
snapshot_obj.file_collection_id)
return self.file_driver.get_collection_files(
file_collection_obj.filehash, mode=mode)
def delete(self, snapshot_id):
"""Delete all traces of a snapshot
Parameters
----------
snapshot_id : str
id for the snapshot to remove
Returns
-------
bool
True if success
Raises
------
RequiredArgumentMissing
if the provided snapshot_id is None
"""
if not snapshot_id:
raise RequiredArgumentMissing(
__("error", "controller.snapshot.delete.arg", "snapshot_id"))
return self.dal.snapshot.delete(snapshot_id)
def _code_setup(self, incoming_dictionary, create_dict):
""" Set the code_id by using:
1. code_id
2. commit_id string, which creates a new code_id
3. create a new code id
Parameters
----------
incoming_dictionary : dict
dictionary for the create function defined above
create_dict : dict
dictionary for creating the Snapshot entity
"""
if "code_id" in incoming_dictionary:
create_dict['code_id'] = incoming_dictionary['code_id']
elif "commit_id" in incoming_dictionary:
create_dict['code_id'] = self.code.\
create(commit_id=incoming_dictionary['commit_id']).id
else:
create_dict['code_id'] = self.code.create().id
def _env_setup(self, incoming_dictionary, create_dict):
""" TODO:
Parameters
----------
incoming_dictionary : dict
dictionary for the create function defined above
create_dict : dict
dictionary for creating the Snapshot entity
"""
if "environment_id" in incoming_dictionary:
create_dict['environment_id'] = incoming_dictionary[
'environment_id']
elif "environment_paths" in incoming_dictionary:
create_dict['environment_id'] = self.environment.create({
"paths":
incoming_dictionary['environment_paths']
}).id
else:
# create some default environment
create_dict['environment_id'] = self.environment.\
create({}).id
def _file_setup(self, incoming_dictionary, create_dict):
""" Checks for user inputs and uses the file collection controller to
obtain the file collection id and create the necessary collection
Parameters
----------
incoming_dictionary : dict
dictionary for the create function defined above
create_dict : dict
dictionary for creating the Snapshot entity
"""
if "file_collection_id" in incoming_dictionary:
create_dict['file_collection_id'] = incoming_dictionary[
'file_collection_id']
elif "paths" in incoming_dictionary:
# transform file paths to file_collection_id
create_dict['file_collection_id'] = self.file_collection.\
create(incoming_dictionary['paths']).id
else:
# create some default file collection
create_dict['file_collection_id'] = self.file_collection.\
create([]).id
def _config_setup(self, incoming_dictionary, create_dict):
""" Fills in snapshot config by having one of the following:
1. config = JSON object
2. config_filepath = some location where a json file exists
3. config_filename = just the file name
Parameters
----------
incoming_dictionary : dict
dictionary for the create function defined above
create_dict : dict
dictionary for creating the Snapshot entity
Raises
------
FileIOError
"""
if "config" in incoming_dictionary:
create_dict['config'] = incoming_dictionary['config']
elif "config_filepath" in incoming_dictionary:
if not os.path.isfile(incoming_dictionary['config_filepath']):
raise FileIOError(
__("error", "controller.snapshot.create.file_config"))
# If path exists transform file to config dict
config_json_driver = JSONStore(
incoming_dictionary['config_filepath'])
create_dict['config'] = config_json_driver.to_dict()
elif "config_filename" in incoming_dictionary:
config_filename = incoming_dictionary['config_filename']
create_dict['config'] = self._find_in_filecollection(
config_filename, create_dict['file_collection_id'])
else:
config_filename = "config.json"
create_dict['config'] = self._find_in_filecollection(
config_filename, create_dict['file_collection_id'])
def _stats_setup(self, incoming_dictionary, create_dict):
"""Fills in snapshot stats by having one of the following:
1. stats = JSON object
2. stats_filepath = some location where a json file exists
3. stats_filename = just the file name
Parameters
----------
incoming_dictionary : dict
dictionary for the create function defined above
create_dict : dict
dictionary for creating the Snapshot entity
Raises
------
FileIOError
"""
if "stats" in incoming_dictionary:
create_dict['stats'] = incoming_dictionary['stats']
elif "stats_filepath" in incoming_dictionary:
if not os.path.isfile(incoming_dictionary['stats_filepath']):
raise FileIOError(
__("error", "controller.snapshot.create.file_stat"))
# If path exists transform file to config dict
stats_json_driver = JSONStore(
incoming_dictionary['stats_filepath'])
create_dict['stats'] = stats_json_driver.to_dict()
elif "stats_filename" in incoming_dictionary:
stats_filename = incoming_dictionary['stats_filename']
create_dict['stats'] = self._find_in_filecollection(
stats_filename, create_dict['file_collection_id'])
else:
stats_filename = "stats.json"
create_dict['stats'] = self._find_in_filecollection(
stats_filename, create_dict['file_collection_id'])
def _find_in_filecollection(self, file_to_find, file_collection_id):
""" Attempts to find a file within the file collection
Returns
-------
dict
output dictionary of the JSON file
"""
file_collection_obj = self.file_collection.dal.file_collection.\
get_by_id(file_collection_id)
file_collection_path = \
self.file_collection.file_driver.get_collection_path(
file_collection_obj.filehash)
# find all of the possible paths it could exist
possible_paths = [os.path.join(self.home, file_to_find)] + \
[os.path.join(self.home, item[0], file_to_find)
for item in os.walk(file_collection_path)]
existing_possible_paths = [
possible_path for possible_path in possible_paths
if os.path.isfile(possible_path)
]
if not existing_possible_paths:
# TODO: Add some info / warning that no file was found
# create some default stats
return {}
else:
# If any such path exists, transform file to stats dict
json_file = JSONStore(existing_possible_paths[0])
return json_file.to_dict()
class TestTaskController():
def setup_method(self):
self.temp_dir = tempfile.mkdtemp(dir=test_datmo_dir)
self.environment_ids = []
def teardown_method(self):
if not check_docker_inactive(test_datmo_dir,
Config().datmo_directory_name):
self.__setup()
self.environment_controller = EnvironmentController()
for env_id in list(set(self.environment_ids)):
if not self.environment_controller.delete(env_id):
raise Exception
def __setup(self):
Config().set_home(self.temp_dir)
self.project_controller = ProjectController()
self.project_controller.init("test", "test description")
self.environment_controller = EnvironmentController()
self.task_controller = TaskController()
def test_init_fail_project_not_init(self):
Config().set_home(self.temp_dir)
failed = False
try:
TaskController()
except ProjectNotInitialized:
failed = True
assert failed
def test_init_fail_invalid_path(self):
test_home = "some_random_dir"
Config().set_home(test_home)
failed = False
try:
TaskController()
except InvalidProjectPath:
failed = True
assert failed
def test_create(self):
self.__setup()
# Create task in the project
task_obj = self.task_controller.create()
assert isinstance(task_obj, Task)
assert task_obj.created_at
assert task_obj.updated_at
@pytest_docker_environment_failed_instantiation(test_datmo_dir)
def test_run_helper(self):
self.__setup()
# TODO: Try out more options (see below)
# Create environment_driver id
env_def_path = os.path.join(self.project_controller.home, "Dockerfile")
with open(env_def_path, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
paths = [env_def_path]
environment_obj = self.environment_controller.create({"paths": paths})
self.environment_ids.append(environment_obj.id)
# Set log filepath
log_filepath = os.path.join(self.task_controller.home, "test.log")
# create volume to mount
temp_test_dirpath = os.path.join(self.temp_dir, "temp")
os.makedirs(temp_test_dirpath)
# Test option set 1
random_name = str(uuid.uuid1())
options_dict = {
"command": ["sh", "-c", "echo accuracy:0.45"],
"ports": ["8888:8888"],
"name": random_name,
"volumes": {
temp_test_dirpath: {
'bind': '/task/',
'mode': 'rw'
}
},
"mem_limit": "4g",
"detach": False,
"stdin_open": False,
"tty": False,
"api": False,
"interactive": False
}
return_code, run_id, logs = \
self.task_controller._run_helper(environment_obj.id,
options_dict, log_filepath)
assert return_code == 0
assert run_id and \
self.task_controller.environment_driver.get_container(run_id)
assert logs and \
os.path.exists(log_filepath)
self.task_controller.environment_driver.stop_remove_containers_by_term(
term=random_name)
# Test option set 2
random_name_2 = str(uuid.uuid1())
options_dict = {
"command": ["sh", "-c", "echo accuracy:0.45"],
"ports": ["8888:8888"],
"name": random_name_2,
"volumes": {
temp_test_dirpath: {
'bind': '/task/',
'mode': 'rw'
}
},
"mem_limit": "4g",
"detach": True,
"stdin_open": False,
"tty": False,
"api": True,
"interactive": False
}
return_code, run_id, logs = \
self.task_controller._run_helper(environment_obj.id,
options_dict, log_filepath)
assert return_code == 0
assert run_id and \
self.task_controller.environment_driver.get_container(run_id)
assert logs and \
os.path.exists(log_filepath)
self.task_controller.environment_driver.stop_remove_containers_by_term(
term=random_name_2)
def test_parse_logs_for_results(self):
self.__setup()
test_logs = """
this is a log
accuracy is good
accuracy : 0.94
this did not work
validation : 0.32
model_type : logistic regression
"""
result = self.task_controller._parse_logs_for_results(test_logs)
assert isinstance(result, dict)
assert result['accuracy'] == "0.94"
assert result['validation'] == "0.32"
assert result['model_type'] == "logistic regression"
test_logs = """test"""
result = self.task_controller._parse_logs_for_results(test_logs)
assert result is None
def test_update_environment_run_options(self):
self.__setup()
environment_run_option = {
"command": ["python", "script.py"],
"volumes": {
os.path.join(self.temp_dir, "/temp_task"): {
'bind': '/task/',
'mode': 'rw'
},
self.temp_dir: {
'bind': '/home/',
'mode': 'rw'
}
}
}
# create data file
data_dirpath = os.path.join(self.temp_dir, "data")
data_file_dirpath = os.path.join(self.temp_dir, "data_folder")
data_filepath = os.path.join(data_file_dirpath, "data_file.txt")
os.mkdir(data_dirpath)
os.mkdir(data_file_dirpath)
with open(data_filepath, "wb") as f:
f.write(to_bytes("data file"))
data_file_path_map = [(data_filepath, "data_file.txt")]
data_directory_path_map = [(data_dirpath, "data_directory")]
environment_run_option = self.task_controller._update_environment_run_options(
environment_run_option, data_file_path_map,
data_directory_path_map)
assert environment_run_option["volumes"][data_file_dirpath] == {
'bind': '/data/',
'mode': 'rw'
}
assert environment_run_option["volumes"][data_dirpath] == {
'bind': '/data/data_directory',
'mode': 'rw'
}
# Error by passing directory which does not exist
data_dirpath = os.path.join(self.temp_dir, "data_dne")
data_filepath = os.path.join(self.temp_dir, "data_dne",
"data_file.txt")
data_file_path_map = [(data_filepath, "data_file.txt")]
data_directory_path_map = [(data_dirpath, "data_directory")]
failed = False
try:
self.task_controller._update_environment_run_options(
environment_run_option, data_file_path_map,
data_directory_path_map)
except TaskRunError:
failed = True
assert failed
@pytest_docker_environment_failed_instantiation(test_datmo_dir)
def test_run(self):
self.__setup()
# 0) Test failure case without command and without interactive
# 1) Test success case with default values and env def file
# 2) Test failure case if running same task (conflicting containers)
# 3) Test failure case if running same task with snapshot_dict (conflicting containers)
# 4) Test success case with snapshot_dict
# 5) Test success case with saved file during task run
# 6) Test success case with data file path being passed
# 7) Test success case with data directory path being passed
# TODO: look into log filepath randomness, sometimes logs are not written
# Create task in the project
task_obj = self.task_controller.create()
# Create environment definition
env_def_path = os.path.join(self.project_controller.home, "Dockerfile")
with open(env_def_path, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
# 0) Test option 0
failed = False
try:
self.task_controller.run(task_obj.id)
except RequiredArgumentMissing:
failed = True
assert failed
failed = False
try:
self.task_controller.run(
task_obj.id,
task_dict={
"command": None,
"interactive": False,
"ports": None
})
except RequiredArgumentMissing:
failed = True
assert failed
# Create task_dict
task_command = ["sh", "-c", "echo accuracy:0.45"]
task_dict = {"command_list": task_command}
# 1) Test option 1
updated_task_obj = self.task_controller.run(
task_obj.id, task_dict=task_dict)
after_snapshot_obj = self.task_controller.dal.snapshot.get_by_id(
updated_task_obj.after_snapshot_id)
environment_obj = self.task_controller.dal.environment.get_by_id(
after_snapshot_obj.environment_id)
self.environment_ids.append(environment_obj.id)
assert isinstance(updated_task_obj, Task)
assert task_obj.id == updated_task_obj.id
assert updated_task_obj.before_snapshot_id
assert updated_task_obj.ports == None
assert updated_task_obj.interactive == False
assert updated_task_obj.task_dirpath
assert updated_task_obj.log_filepath
assert updated_task_obj.start_time
assert updated_task_obj.after_snapshot_id
assert updated_task_obj.run_id
assert updated_task_obj.logs
assert "accuracy" in updated_task_obj.logs
assert updated_task_obj.results
assert updated_task_obj.results == {"accuracy": "0.45"}
assert after_snapshot_obj.stats == {"accuracy": "0.45"}
assert updated_task_obj.status == "SUCCESS"
assert updated_task_obj.end_time
assert updated_task_obj.duration
self.task_controller.stop(task_obj.id)
# 2) Test option 2
failed = False
try:
self.task_controller.run(task_obj.id)
except TaskRunError:
failed = True
assert failed
# 3) Test option 3
# Create files to add
self.project_controller.file_driver.create("dirpath1", directory=True)
self.project_controller.file_driver.create("dirpath2", directory=True)
self.project_controller.file_driver.create("filepath1")
# Snapshot dictionary
snapshot_dict = {
"paths": [
os.path.join(self.project_controller.home, "dirpath1"),
os.path.join(self.project_controller.home, "dirpath2"),
os.path.join(self.project_controller.home, "filepath1")
],
}
# Run a basic task in the project
failed = False
try:
self.task_controller.run(task_obj.id, snapshot_dict=snapshot_dict)
except TaskRunError:
failed = True
assert failed
# Test when the specific task id is already RUNNING
# Create task in the project
task_obj_1 = self.task_controller.create()
self.task_controller.dal.task.update({
"id": task_obj_1.id,
"status": "RUNNING"
})
# Create environment_driver definition
env_def_path = os.path.join(self.project_controller.home, "Dockerfile")
with open(env_def_path, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
failed = False
try:
self.task_controller.run(task_obj_1.id, task_dict=task_dict)
except TaskRunError:
failed = True
assert failed
# 4) Test option 4
# Create a new task in the project
task_obj_2 = self.task_controller.create()
# Run another task in the project
updated_task_obj_2 = self.task_controller.run(
task_obj_2.id, task_dict=task_dict, snapshot_dict=snapshot_dict)
after_snapshot_obj = self.task_controller.dal.snapshot.get_by_id(
updated_task_obj_2.after_snapshot_id)
environment_obj = self.task_controller.dal.environment.get_by_id(
after_snapshot_obj.environment_id)
self.environment_ids.append(environment_obj.id)
assert isinstance(updated_task_obj_2, Task)
assert task_obj_2.id == updated_task_obj_2.id
assert updated_task_obj_2.before_snapshot_id
assert updated_task_obj_2.ports == None
assert updated_task_obj_2.interactive == False
assert updated_task_obj_2.task_dirpath
assert updated_task_obj_2.log_filepath
assert updated_task_obj_2.start_time
assert updated_task_obj_2.after_snapshot_id
assert updated_task_obj_2.run_id
assert updated_task_obj_2.logs
assert "accuracy" in updated_task_obj_2.logs
assert updated_task_obj_2.results
assert updated_task_obj_2.results == {"accuracy": "0.45"}
assert updated_task_obj_2.status == "SUCCESS"
assert updated_task_obj_2.end_time
assert updated_task_obj_2.duration
self.task_controller.stop(task_obj_2.id)
# 5) Test option 5
# Create a basic script
test_filepath = os.path.join(self.temp_dir, "script.py")
with open(test_filepath, "wb") as f:
f.write(to_bytes("import os\n"))
f.write(to_bytes("import shutil\n"))
f.write(to_bytes("print('hello')\n"))
f.write(to_bytes("print(' accuracy: 0.56 ')\n"))
f.write(
to_bytes(
"with open(os.path.join('/task', 'new_file.txt'), 'a') as f:\n"
))
f.write(to_bytes(" f.write('my test file')\n"))
# Create task in the project
task_obj_2 = self.task_controller.create()
# Create task_dict
task_command = ["python", "script.py"]
task_dict = {"command_list": task_command}
# Create environment definition
env_def_path = os.path.join(self.project_controller.home, "Dockerfile")
with open(env_def_path, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
updated_task_obj_2 = self.task_controller.run(
task_obj_2.id, task_dict=task_dict)
after_snapshot_obj = self.task_controller.dal.snapshot.get_by_id(
updated_task_obj_2.after_snapshot_id)
environment_obj = self.task_controller.dal.environment.get_by_id(
after_snapshot_obj.environment_id)
self.environment_ids.append(environment_obj.id)
assert isinstance(updated_task_obj_2, Task)
assert updated_task_obj_2.before_snapshot_id
assert updated_task_obj_2.ports == None
assert updated_task_obj_2.interactive == False
assert updated_task_obj_2.task_dirpath
assert updated_task_obj_2.log_filepath
assert updated_task_obj_2.start_time
assert updated_task_obj_2.after_snapshot_id
assert updated_task_obj_2.run_id
assert updated_task_obj_2.logs
assert "accuracy" in updated_task_obj_2.logs
assert updated_task_obj_2.results
assert updated_task_obj_2.results == {"accuracy": "0.56"}
assert updated_task_obj_2.status == "SUCCESS"
assert updated_task_obj_2.end_time
assert updated_task_obj_2.duration
self.task_controller.stop(task_obj_2.id)
# test if after snapshot has the file written
after_snapshot_obj = self.task_controller.dal.snapshot.get_by_id(
updated_task_obj_2.after_snapshot_id)
file_collection_obj = self.task_controller.dal.file_collection.get_by_id(
after_snapshot_obj.file_collection_id)
files_absolute_path = os.path.join(self.task_controller.home,
file_collection_obj.path)
assert os.path.isfile(os.path.join(files_absolute_path, "task.log"))
assert os.path.isfile(
os.path.join(files_absolute_path, "new_file.txt"))
# 6) Test Option 6
self.project_controller.file_driver.create("dirpath1", directory=True)
self.project_controller.file_driver.create(
os.path.join("dirpath1", "file.txt"))
with open(
os.path.join(self.project_controller.home, "dirpath1",
"file.txt"), "wb") as f:
f.write(to_bytes('my initial line\n'))
test_filename = "script.py"
test_filepath = os.path.join(self.temp_dir, test_filename)
with open(test_filepath, "wb") as f:
f.write(to_bytes("import os\n"))
f.write(to_bytes("print('hello')\n"))
f.write(to_bytes("import shutil\n"))
f.write(
to_bytes(
"with open(os.path.join('/data', 'file.txt'), 'a') as f:\n"
))
f.write(to_bytes(" f.write('my test file')\n"))
# Create task in the project
task_obj_3 = self.task_controller.create()
# Create task_dict
task_command = ["python", test_filename]
task_dict = {
"command_list":
task_command,
"data_file_path_map": [(os.path.join(self.project_controller.home,
"dirpath1", "file.txt"),
'file.txt')]
}
# Create environment definition
env_def_path = os.path.join(self.project_controller.home, "Dockerfile")
with open(env_def_path, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
updated_task_obj_3 = self.task_controller.run(
task_obj_3.id, task_dict=task_dict)
after_snapshot_obj = self.task_controller.dal.snapshot.get_by_id(
updated_task_obj_3.after_snapshot_id)
environment_obj = self.task_controller.dal.environment.get_by_id(
after_snapshot_obj.environment_id)
self.environment_ids.append(environment_obj.id)
assert isinstance(updated_task_obj_3, Task)
assert updated_task_obj_3.before_snapshot_id
assert updated_task_obj_3.ports == None
assert updated_task_obj_3.interactive == False
assert updated_task_obj_3.task_dirpath
assert updated_task_obj_3.log_filepath
assert updated_task_obj_3.start_time
assert updated_task_obj_3.after_snapshot_id
assert updated_task_obj_3.run_id
assert updated_task_obj_3.logs
assert updated_task_obj_3.status == "SUCCESS"
assert updated_task_obj_3.end_time
assert updated_task_obj_3.duration
self.task_controller.stop(task_obj_3.id)
# test if after snapshot has the file written
after_snapshot_obj = self.task_controller.dal.snapshot.get_by_id(
updated_task_obj_3.after_snapshot_id)
file_collection_obj = self.task_controller.dal.file_collection.get_by_id(
after_snapshot_obj.file_collection_id)
files_absolute_path = os.path.join(self.task_controller.home,
file_collection_obj.path)
assert os.path.isfile(os.path.join(files_absolute_path, "task.log"))
assert os.path.isfile(
os.path.join(self.project_controller.home, "dirpath1", "file.txt"))
assert "my initial line" in open(
os.path.join(self.project_controller.home, "dirpath1", "file.txt"),
"r").read()
assert "my test file" in open(
os.path.join(self.project_controller.home, "dirpath1", "file.txt"),
"r").read()
# 7) Test Option 7
self.project_controller.file_driver.create("dirpath1", directory=True)
self.project_controller.file_driver.create(
os.path.join("dirpath1", "file.txt"))
with open(
os.path.join(self.project_controller.home, "dirpath1",
"file.txt"), "wb") as f:
f.write(to_bytes('my initial line\n'))
test_filename = "script.py"
test_filepath = os.path.join(self.temp_dir, test_filename)
with open(test_filepath, "wb") as f:
f.write(to_bytes("import os\n"))
f.write(to_bytes("print('hello')\n"))
f.write(to_bytes("import shutil\n"))
f.write(
to_bytes(
"with open(os.path.join('/data', 'dirpath1', 'file.txt'), 'a') as f:\n"
))
f.write(to_bytes(" f.write('my test file')\n"))
# Create task in the project
task_obj_4 = self.task_controller.create()
# Create task_dict
task_command = ["python", test_filename]
task_dict = {
"command_list":
task_command,
"data_directory_path_map": [(os.path.join(
self.project_controller.home, "dirpath1"), 'dirpath1')]
}
# Create environment definition
env_def_path = os.path.join(self.project_controller.home, "Dockerfile")
with open(env_def_path, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
updated_task_obj_4 = self.task_controller.run(
task_obj_4.id, task_dict=task_dict)
after_snapshot_obj = self.task_controller.dal.snapshot.get_by_id(
updated_task_obj_4.after_snapshot_id)
environment_obj = self.task_controller.dal.environment.get_by_id(
after_snapshot_obj.environment_id)
self.environment_ids.append(environment_obj.id)
assert isinstance(updated_task_obj_4, Task)
assert updated_task_obj_4.before_snapshot_id
assert updated_task_obj_4.ports == None
assert updated_task_obj_4.interactive == False
assert updated_task_obj_4.task_dirpath
assert updated_task_obj_4.log_filepath
assert updated_task_obj_4.start_time
assert updated_task_obj_4.after_snapshot_id
assert updated_task_obj_4.run_id
assert updated_task_obj_4.logs
assert updated_task_obj_4.status == "SUCCESS"
assert updated_task_obj_4.end_time
assert updated_task_obj_4.duration
self.task_controller.stop(task_obj_4.id)
# test if after snapshot has the file written
after_snapshot_obj = self.task_controller.dal.snapshot.get_by_id(
updated_task_obj_4.after_snapshot_id)
file_collection_obj = self.task_controller.dal.file_collection.get_by_id(
after_snapshot_obj.file_collection_id)
files_absolute_path = os.path.join(self.task_controller.home,
file_collection_obj.path)
assert os.path.isfile(os.path.join(files_absolute_path, "task.log"))
assert os.path.isfile(
os.path.join(self.project_controller.home, "dirpath1", "file.txt"))
assert "my initial line" in open(
os.path.join(self.project_controller.home, "dirpath1", "file.txt"),
"r").read()
assert "my test file" in open(
os.path.join(self.project_controller.home, "dirpath1", "file.txt"),
"r").read()
def test_list(self):
self.__setup()
# Create tasks in the project
task_obj_1 = self.task_controller.create()
task_obj_2 = self.task_controller.create()
# List all tasks regardless of filters
result = self.task_controller.list()
assert len(result) == 2 and \
task_obj_1 in result and \
task_obj_2 in result
# List all tasks regardless of filters in ascending
result = self.task_controller.list(
sort_key='created_at', sort_order='ascending')
assert len(result) == 2 and \
task_obj_1 in result and \
task_obj_2 in result
assert result[0].created_at <= result[-1].created_at
# List all tasks regardless of filters in descending
result = self.task_controller.list(
sort_key='created_at', sort_order='descending')
assert len(result) == 2 and \
task_obj_1 in result and \
task_obj_2 in result
assert result[0].created_at >= result[-1].created_at
# Wrong order being passed in
failed = False
try:
_ = self.task_controller.list(
sort_key='created_at', sort_order='wrong_order')
except InvalidArgumentType:
failed = True
assert failed
# Wrong key and order being passed in
failed = False
try:
_ = self.task_controller.list(
sort_key='wrong_key', sort_order='wrong_order')
except InvalidArgumentType:
failed = True
assert failed
# wrong key and right order being passed in
expected_result = self.task_controller.list(
sort_key='created_at', sort_order='ascending')
result = self.task_controller.list(
sort_key='wrong_key', sort_order='ascending')
expected_ids = [item.id for item in expected_result]
ids = [item.id for item in result]
assert set(expected_ids) == set(ids)
# List all tasks
result = self.task_controller.list()
assert len(result) == 2 and \
task_obj_1 in result and \
task_obj_2 in result
def test_get(self):
self.__setup()
# Test failure for no task
failed = False
try:
self.task_controller.get("random")
except DoesNotExist:
failed = True
assert failed
# Test success for task
task_obj = self.task_controller.create()
task_obj_returned = self.task_controller.get(task_obj.id)
assert task_obj == task_obj_returned
@pytest_docker_environment_failed_instantiation(test_datmo_dir)
def test_get_files(self):
self.__setup()
# Test failure case
failed = False
try:
self.task_controller.get_files("random")
except DoesNotExist:
failed = True
assert failed
# Create task in the project
task_obj = self.task_controller.create()
# Create environment definition
env_def_path = os.path.join(self.project_controller.home, "Dockerfile")
with open(env_def_path, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
# Create file to add
self.project_controller.file_driver.create("dirpath1", directory=True)
self.project_controller.file_driver.create(
os.path.join("dirpath1", "filepath1"))
# Snapshot dictionary
snapshot_dict = {
"paths": [
os.path.join(self.project_controller.home, "dirpath1",
"filepath1")
],
}
# Create task_dict
task_command = ["sh", "-c", "echo accuracy:0.45"]
task_dict = {"command_list": task_command}
# Test the default values
updated_task_obj = self.task_controller.run(
task_obj.id, task_dict=task_dict, snapshot_dict=snapshot_dict)
after_snapshot_obj = self.task_controller.dal.snapshot.get_by_id(
updated_task_obj.after_snapshot_id)
environment_obj = self.task_controller.dal.environment.get_by_id(
after_snapshot_obj.environment_id)
self.environment_ids.append(environment_obj.id)
# TODO: Test case for during run and before_snapshot run
# Get files for the task after run is complete (default)
result = self.task_controller.get_files(updated_task_obj.id)
after_snapshot_obj = self.task_controller.dal.snapshot.get_by_id(
updated_task_obj.after_snapshot_id)
file_collection_obj = self.task_controller.dal.file_collection.get_by_id(
after_snapshot_obj.file_collection_id)
file_names = [item.name for item in result]
assert len(result) == 2
for item in result:
assert isinstance(item, TextIOWrapper)
assert item.mode == "r"
assert os.path.join(self.task_controller.home, ".datmo", "collections",
file_collection_obj.filehash,
"task.log") in file_names
assert os.path.join(self.task_controller.home, ".datmo", "collections",
file_collection_obj.filehash,
"filepath1") in file_names
# Get files for the task after run is complete for different mode
result = self.task_controller.get_files(updated_task_obj.id, mode="a")
assert len(result) == 2
for item in result:
assert isinstance(item, TextIOWrapper)
assert item.mode == "a"
assert os.path.join(self.task_controller.home, ".datmo", "collections",
file_collection_obj.filehash,
"task.log") in file_names
assert os.path.join(self.task_controller.home, ".datmo", "collections",
file_collection_obj.filehash,
"filepath1") in file_names
self.task_controller.stop(task_obj.id)
def test_update(self):
self.__setup()
# Create task in the project
task_obj = self.task_controller.create()
assert isinstance(task_obj, Task)
# Test 1: When no meta data is passed
updated_task_obj = self.task_controller.update(task_obj.id)
assert updated_task_obj.workspace is None
# Test 2: When meta data for workspace is passed
test_workspace = "notebook"
test_command = "python script.py"
updated_task_obj = self.task_controller.update(
task_obj.id, workspace=test_workspace, command=test_command)
assert updated_task_obj.workspace == test_workspace
assert updated_task_obj.command == test_command
assert updated_task_obj.command_list == ["python", "script.py"]
# Test 3: When meta data for workspace is passed
test_interactive = True
updated_task_obj = self.task_controller.update(
task_obj.id, interactive=test_interactive)
assert updated_task_obj.interactive == test_interactive
# Test 4: When meta data for workspace is passed
test_command_list = ["python", "script.py"]
updated_task_obj = self.task_controller.update(
task_obj.id, command_list=test_command_list)
assert updated_task_obj.command_list == ["python", "script.py"]
@pytest_docker_environment_failed_instantiation(test_datmo_dir)
def test_delete(self):
self.__setup()
# Create tasks in the project
task_obj = self.task_controller.create()
# Delete task from the project
result = self.task_controller.delete(task_obj.id)
# Check if task retrieval throws error
thrown = False
try:
self.task_controller.dal.snapshot.get_by_id(task_obj.id)
except EntityNotFound:
thrown = True
assert result == True and \
thrown == True
@pytest_docker_environment_failed_instantiation(test_datmo_dir)
def test_stop_failure(self):
self.__setup()
# 1) Test required arguments not provided
# 2) Test too many arguments found
# 3) Test incorrect task id given
# 1) Test option 1
failed = False
try:
self.task_controller.stop()
except RequiredArgumentMissing:
failed = True
assert failed
# 2) Test option 2
failed = False
try:
self.task_controller.stop(task_id="test_task_id", all=True)
except TooManyArgumentsFound:
failed = True
assert failed
# 3) Test option 3
thrown = False
try:
self.task_controller.stop(task_id="incorrect_task_id")
except DoesNotExist:
thrown = True
assert thrown
@pytest_docker_environment_failed_instantiation(test_datmo_dir)
def test_stop_success(self):
self.__setup()
# 1) Test stop with task_id
# 2) Test stop with all given
# Create task in the project
task_obj = self.task_controller.create()
# Create environment driver definition
env_def_path = os.path.join(self.project_controller.home, "Dockerfile")
with open(env_def_path, "wb") as f:
f.write(to_bytes("FROM python:3.5-alpine"))
# Create task_dict
task_command = ["sh", "-c", "echo accuracy:0.45"]
task_dict = {"command_list": task_command}
# 1) Test option 1
updated_task_obj = self.task_controller.run(
task_obj.id, task_dict=task_dict)
after_snapshot_obj = self.task_controller.dal.snapshot.get_by_id(
updated_task_obj.after_snapshot_id)
environment_obj = self.task_controller.dal.environment.get_by_id(
after_snapshot_obj.environment_id)
self.environment_ids.append(environment_obj.id)
task_id = updated_task_obj.id
result = self.task_controller.stop(task_id=task_id)
after_task_obj = self.task_controller.dal.task.get_by_id(task_id)
assert result
assert after_task_obj.status == "STOPPED"
# 2) Test option 2
task_obj_2 = self.task_controller.create()
updated_task_obj = self.task_controller.run(
task_obj_2.id, task_dict=task_dict)
after_snapshot_obj = self.task_controller.dal.snapshot.get_by_id(
updated_task_obj.after_snapshot_id)
environment_obj = self.task_controller.dal.environment.get_by_id(
after_snapshot_obj.environment_id)
self.environment_ids.append(environment_obj.id)
result = self.task_controller.stop(all=True)
all_task_objs = self.task_controller.dal.task.query({})
assert result
for task_obj in all_task_objs:
assert task_obj.status == "STOPPED"
class TestEnvironmentController():
def setup_method(self):
# provide mountable tmp directory for docker
tempfile.tempdir = "/tmp" if not platform.system(
) == "Windows" else None
test_datmo_dir = os.environ.get('TEST_DATMO_DIR',
tempfile.gettempdir())
self.temp_dir = tempfile.mkdtemp(dir=test_datmo_dir)
self.project = ProjectController(self.temp_dir)
self.environment = EnvironmentController(self.temp_dir)
def teardown_method(self):
pass
def test_create(self):
# 0) Test create when unsupported language given
# 1) Test create when NO file exists and NO definition path exists
# 2) Test create when NO file exists and definition path exists
# 3) Test create when definition path exists and given
# 4) Test create when file exists and definition path exists
# 5) Test create when file exists but NO definition path exists
# 6) Test create when definition path exists and given for NEW definition filepath
self.project.init("test3", "test description")
# 0) Test option 0
try:
self.environment.create({"language": "java"})
except EnvironmentDoesNotExist:
failed = True
assert failed
# 1) Test option 1
# Creates environment with python3 based docker image
environment_obj_0 = self.environment.create({})
assert environment_obj_0
assert environment_obj_0.id
assert environment_obj_0.driver_type == "docker"
assert environment_obj_0.file_collection_id
assert environment_obj_0.definition_filename
assert environment_obj_0.hardware_info
# Create environment definition
definition_filepath = os.path.join(self.environment.home, "Dockerfile")
with open(definition_filepath, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
# 2) Test option 2
environment_obj_1 = self.environment.create({})
# Get file collection path
file_collection_obj = self.environment.dal.file_collection. \
get_by_id(environment_obj_1.file_collection_id)
file_collection_dir = self.environment.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert environment_obj_1
assert environment_obj_1.id
assert environment_obj_1.driver_type == "docker"
assert environment_obj_1.file_collection_id
assert environment_obj_1.definition_filename
assert environment_obj_1.hardware_info
assert environment_obj_1.unique_hash == file_collection_obj.filehash
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "hardware_info"))
# 3) Test option 3
input_dict = {
"definition_filepath": definition_filepath,
}
# Create environment in the project
environment_obj_2 = self.environment.create(input_dict)
# Get file collection path
file_collection_obj = self.environment.dal.file_collection. \
get_by_id(environment_obj_2.file_collection_id)
file_collection_dir = self.environment.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert environment_obj_2
assert environment_obj_2.id
assert environment_obj_2.driver_type == "docker"
assert environment_obj_2.file_collection_id
assert environment_obj_2.definition_filename
assert environment_obj_2.hardware_info
assert environment_obj_2.unique_hash == file_collection_obj.filehash
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "hardware_info"))
# Create script to test
test_filepath = os.path.join(self.environment.home, "script.py")
with open(test_filepath, "w") as f:
f.write(to_unicode("import numpy\n"))
f.write(to_unicode("import sklearn\n"))
f.write(to_unicode("print('hello')\n"))
# 4) Test option 4
environment_obj_3 = self.environment.create({})
# Get file collection path
file_collection_obj = self.environment.dal.file_collection. \
get_by_id(environment_obj_3.file_collection_id)
file_collection_dir = self.environment.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert environment_obj_3
assert environment_obj_3.id
assert environment_obj_3.driver_type == "docker"
assert environment_obj_3.file_collection_id
assert environment_obj_3.definition_filename
assert environment_obj_3.hardware_info
assert environment_obj_3.unique_hash == file_collection_obj.filehash
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "hardware_info"))
# Remove definition filepath
os.remove(definition_filepath)
assert environment_obj_1.id == environment_obj_2.id
assert environment_obj_2.id == environment_obj_3.id
# 5) Test option 5
environment_obj_4 = self.environment.create({})
# Get file collection path
file_collection_obj = self.environment.dal.file_collection. \
get_by_id(environment_obj_4.file_collection_id)
file_collection_dir = self.environment.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert environment_obj_4
assert environment_obj_4.id
assert environment_obj_4.driver_type == "docker"
assert environment_obj_4.file_collection_id
assert environment_obj_4.definition_filename
assert environment_obj_4.hardware_info
assert environment_obj_4.unique_hash == file_collection_obj.filehash
assert os.path.isfile(
os.path.join(file_collection_dir, "datmorequirements.txt"))
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "hardware_info"))
assert environment_obj_1.id != environment_obj_4.id
# 6) Test option 6
# Create environment definition
definition_filepath = os.path.join(self.environment.home, "Dockerfile")
with open(definition_filepath, "w") as f:
f.write(to_unicode(str("FROM cloudgear/ubuntu:14.04")))
input_dict = {
"definition_filepath": definition_filepath,
}
# Create a new environment obj
environment_obj_5 = self.environment.create(input_dict)
# Get file collection path
file_collection_obj = self.environment.dal.file_collection. \
get_by_id(environment_obj_5.file_collection_id)
file_collection_dir = self.environment.file_driver. \
get_collection_path(file_collection_obj.filehash)
assert environment_obj_5
assert environment_obj_5.id
assert environment_obj_5.driver_type == "docker"
assert environment_obj_5.file_collection_id
assert environment_obj_5.definition_filename
assert environment_obj_5.hardware_info
assert environment_obj_5.unique_hash == file_collection_obj.filehash
assert os.path.isfile(os.path.join(file_collection_dir, "Dockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "datmoDockerfile"))
assert os.path.isfile(
os.path.join(file_collection_dir, "hardware_info"))
assert environment_obj_5.id != environment_obj_1.id
assert environment_obj_5.id != environment_obj_4.id
def test_build(self):
# 1) Test build when no environment given
# 2) Test build when definition path exists and given
# 3) Test build when NO file exists and definition path exists
# 4) Test build when file exists and definition path exists
# 5) Test build when file exists but NO definition path exists
self.project.init("test5", "test description")
# 1) Test option 1
failed = False
try:
_ = self.environment.build("does_not_exist")
except EntityNotFound:
failed = True
assert failed
# Create environment definition
definition_filepath = os.path.join(self.environment.home, "Dockerfile")
with open(definition_filepath, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
input_dict = {
"definition_filepath": definition_filepath,
}
# 2) Test option 2
# Create environment in the project
environment_obj_1 = self.environment.create(input_dict)
result = self.environment.build(environment_obj_1.id)
assert result
# 3) Test option 3
# Create environment in the project
environment_obj_2 = self.environment.create({})
result = self.environment.build(environment_obj_2.id)
assert result
# Create script to test
test_filepath = os.path.join(self.environment.home, "script.py")
with open(test_filepath, "w") as f:
f.write(to_unicode("import numpy\n"))
f.write(to_unicode("import sklearn\n"))
f.write(to_unicode("print('hello')\n"))
# 4) Test option 4
environment_obj_3 = self.environment.create({})
result = self.environment.build(environment_obj_3.id)
assert result
# test 2), 3), and 4) will result in the same environment
assert environment_obj_1.id == environment_obj_2.id
assert environment_obj_2.id == environment_obj_3.id
# Test for building dockerfile when there exists not
os.remove(definition_filepath)
# 5) Test option 5
environment_obj_4 = self.environment.create({})
result = self.environment.build(environment_obj_4.id)
# 2) Test run script, with autogenerated definition
assert result
assert environment_obj_4.id != environment_obj_1.id
# teardown
self.environment.delete(environment_obj_1.id)
self.environment.delete(environment_obj_4.id)
def test_run(self):
# 1) Test run simple command with simple Dockerfile
self.project.init("test5", "test description")
# 1) Test option 1
# Create environment definition
definition_filepath = os.path.join(self.environment.home, "Dockerfile")
with open(definition_filepath, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
random_name = ''.join([
random.choice(string.ascii_letters + string.digits)
for _ in range(32)
])
run_options = {
"command": ["sh", "-c", "echo yo"],
"ports": ["8888:8888"],
"name": random_name,
"volumes": None,
"detach": True,
"stdin_open": False,
"tty": False,
"api": False
}
input_dict = {
"definition_filepath": definition_filepath,
}
# Create environment in the project
environment_obj = self.environment.create(input_dict)
log_filepath = os.path.join(self.project.home, "task.log")
# Build environment in the project
_ = self.environment.build(environment_obj.id)
# Run environment in the project
return_code, run_id, logs = \
self.environment.run(environment_obj.id, run_options, log_filepath)
assert return_code == 0
assert run_id
assert logs
# teardown
self.environment.delete(environment_obj.id)
# 2) Test option 2
os.remove(definition_filepath)
# Create script to test
test_filepath = os.path.join(self.environment.home, "script.py")
with open(test_filepath, "w") as f:
f.write(to_unicode("import numpy\n"))
f.write(to_unicode("import sklearn\n"))
f.write(to_unicode("print('hello')\n"))
# Create environment in the project
environment_obj = self.environment.create({})
self.environment.build(environment_obj.id)
random_name = ''.join([
random.choice(string.ascii_letters + string.digits)
for _ in range(32)
])
run_options = {
"command": ["python", "script.py"],
"ports": ["8888:8888"],
"name": random_name,
"volumes": {
self.environment.home: {
'bind': '/home/',
'mode': 'rw'
}
},
"detach": False,
"stdin_open": False,
"tty": False,
"api": False
}
# Run environment in the project
return_code, run_id, logs = \
self.environment.run(environment_obj.id, run_options, log_filepath)
assert return_code == 0
assert run_id
assert logs
# teardown
self.environment.delete(environment_obj.id)
def test_interactive_run(self):
# 1) Test run interactive terminal in environment
# 2) Test run jupyter notebook in environment
# Create environment definition
self.project.init("test6", "test description")
definition_filepath = os.path.join(self.environment.home, "Dockerfile")
with open(definition_filepath, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
input_dict = {
"definition_filepath": definition_filepath,
}
# Create environment in the project
environment_obj = self.environment.create(input_dict)
# 1) Test option 1
@timeout_decorator.timeout(10, use_signals=False)
def timed_run(container_name, timed_run):
run_options = {
"command": [],
"ports": ["8888:8888"],
"name": container_name,
"volumes": None,
"detach": True,
"stdin_open": True,
"tty": True,
"api": False
}
log_filepath = os.path.join(self.project.home, "task.log")
# Build environment in the project
_ = self.environment.build(environment_obj.id)
# Run environment in the project
self.environment.run(environment_obj.id, run_options, log_filepath)
return timed_run
container_name = str(uuid.uuid1())
timed_run_result = False
try:
timed_run_result = timed_run(container_name, timed_run_result)
except timeout_decorator.timeout_decorator.TimeoutError:
timed_run_result = True
assert timed_run_result
# teardown
self.environment.delete(environment_obj.id)
# 2) Test option 2
environment_obj = self.environment.create(input_dict)
@timeout_decorator.timeout(10, use_signals=False)
def timed_run(container_name, timed_run):
run_options = {
"command": ["jupyter", "notebook"],
"ports": ["8888:8888"],
"name": container_name,
"volumes": None,
"detach": True,
"stdin_open": False,
"tty": False,
"api": False
}
log_filepath = os.path.join(self.project.home, "task.log")
# Build environment in the project
_ = self.environment.build(environment_obj.id)
# Run environment in the project
self.environment.run(environment_obj.id, run_options, log_filepath)
return timed_run
container_name = str(uuid.uuid1())
timed_run_result = False
try:
timed_run_result = timed_run(container_name, timed_run_result)
except timeout_decorator.timeout_decorator.TimeoutError:
timed_run_result = True
assert timed_run_result
# Stop the running environment
# self.environment.stop(container_name)
# teardown
self.environment.delete(environment_obj.id)
def test_list(self):
self.project.init("test4", "test description")
# Create environment definition for object 1
definition_path_1 = os.path.join(self.environment.home, "Dockerfile")
with open(definition_path_1, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
input_dict_1 = {
"definition_filepath": definition_path_1,
}
# Create environment in the project
environment_obj_1 = self.environment.create(input_dict_1)
# Create environment definition for object 2
definition_path_2 = os.path.join(self.environment.home, "Dockerfile2")
with open(definition_path_2, "w") as f:
f.write(to_unicode(str("FROM datmo/scikit-opencv")))
input_dict_2 = {
"definition_filepath": definition_path_2,
}
# Create second environment in the project
environment_obj_2 = self.environment.create(input_dict_2)
# List all environments and ensure they exist
result = self.environment.list()
assert len(result) == 2 and \
environment_obj_1 in result and \
environment_obj_2 in result
def test_delete(self):
self.project.init("test5", "test description")
# Create environment definition
definition_filepath = os.path.join(self.environment.home, "Dockerfile")
with open(definition_filepath, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
input_dict = {
"definition_filepath": definition_filepath,
}
# Create environment in the project
environment_obj = self.environment.create(input_dict)
# Delete environment in the project
result = self.environment.delete(environment_obj.id)
# Check if environment retrieval throws error
thrown = False
try:
self.environment.dal.environment.get_by_id(environment_obj.id)
except EntityNotFound:
thrown = True
assert result == True and \
thrown == True
def test_stop_failure(self):
# 1) Test failure with RequiredArgumentMissing
# 2) Test failure with TooManyArgumentsFound
# 1) Test option 1
failed = False
try:
self.environment.stop()
except RequiredArgumentMissing:
failed = True
assert failed
# 2) Test option 2
failed = False
try:
self.environment.stop(run_id="hello", match_string="there")
except TooManyArgumentsFound:
failed = True
assert failed
def test_stop_success(self):
# TODO: test more run options
# 1) Test run_id input to stop
# 2) Test match_string input to stop
# 3) Test all input to stop
self.project.init("test5", "test description")
# Create environment definition
definition_filepath = os.path.join(self.environment.home, "Dockerfile")
with open(definition_filepath, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
run_options = {
"command": ["sh", "-c", "echo yo"],
"ports": ["8888:8888"],
"name": "datmo-task-" + self.environment.model.id + "-" + "test",
"volumes": None,
"detach": False,
"stdin_open": False,
"tty": False,
"api": False
}
# Create environment definition
env_def_path = os.path.join(self.project.home, "Dockerfile")
with open(env_def_path, "w") as f:
f.write(to_unicode(str("FROM datmo/xgboost:cpu")))
input_dict = {
"definition_filepath": definition_filepath,
}
# Create environment in the project
environment_obj = self.environment.create(input_dict)
log_filepath = os.path.join(self.project.home, "task.log")
# Build environment in the project
_ = self.environment.build(environment_obj.id)
# 1) Test option 1
_, run_id, _ = \
self.environment.run(environment_obj.id, run_options, log_filepath)
return_code = self.environment.stop(run_id=run_id)
assert return_code
# 2) Test option 2
_, _, _ = \
self.environment.run(environment_obj.id, run_options, log_filepath)
return_code = self.environment.stop(match_string="datmo-task-" +
self.environment.model.id)
assert return_code
# 3) Test option 3
_, _, _ = \
self.environment.run(environment_obj.id, run_options, log_filepath)
run_options_2 = {
"command": ["sh", "-c", "echo yo"],
"ports": ["8888:8888"],
"name": "datmo-task-" + self.environment.model.id + "-" + "test2",
"volumes": None,
"detach": False,
"stdin_open": False,
"tty": False,
"api": False
}
_, _, _ = \
self.environment.run(environment_obj.id, run_options_2, log_filepath)
return_code = self.environment.stop(all=True)
assert return_code
# teardown
self.environment.delete(environment_obj.id)
class EnvironmentCommand(ProjectCommand):
def __init__(self, cli_helper):
super(EnvironmentCommand, self).__init__(cli_helper)
def environment(self):
self.parse(["environment", "--help"])
return True
@Helper.notify_no_project_found
def setup(self, **kwargs):
self.environment_controller = EnvironmentController()
name = kwargs.get("name", None)
available_environments = self.environment_controller.get_supported_environments(
)
if not name:
name = self.cli_helper.prompt_available_environments(
available_environments)
try:
options = {"name": name}
environment_obj = self.environment_controller.setup(
options=options)
self.cli_helper.echo(
__("info", "cli.environment.setup.success",
(environment_obj.name, environment_obj.id)))
return environment_obj
except EnvironmentDoesNotExist:
self.cli_helper.echo(
__("error", "cli.environment.setup.argument", name))
@Helper.notify_no_project_found
def create(self, **kwargs):
self.environment_controller = EnvironmentController()
self.cli_helper.echo(__("info", "cli.environment.create"))
created_environment_obj = self.environment_controller.create(kwargs)
environments = self.environment_controller.list()
for environment_obj in environments:
if created_environment_obj == environment_obj:
self.cli_helper.echo(
__("info", "cli.environment.create.alreadyexist",
created_environment_obj.id))
return created_environment_obj
self.cli_helper.echo(
__("info", "cli.environment.create.success",
created_environment_obj.id))
return created_environment_obj
@Helper.notify_no_project_found
def update(self, **kwargs):
self.environment_controller = EnvironmentController()
environment_id = kwargs.get('id')
name = kwargs.get('name', None)
description = kwargs.get("description", None)
result = self.environment_controller.update(environment_id,
name=name,
description=description)
return result
@Helper.notify_environment_active(EnvironmentController)
@Helper.notify_no_project_found
def delete(self, **kwargs):
self.environment_controller = EnvironmentController()
environment_id = kwargs.get('id')
if self.environment_controller.delete(environment_id):
self.cli_helper.echo(
__("info", "cli.environment.delete.success", environment_id))
return True
@Helper.notify_no_project_found
def ls(self, **kwargs):
self.environment_controller = EnvironmentController()
print_format = kwargs.get('format', "table")
download = kwargs.get('download', None)
download_path = kwargs.get('download_path', None)
environment_objs = self.environment_controller.list()
header_list = ["id", "created at", "name", "description"]
item_dict_list = []
for environment_obj in environment_objs:
environment_obj_name = printable_object(environment_obj.name)
environment_obj_description = printable_object(
environment_obj.description)
item_dict_list.append({
"id":
environment_obj.id,
"created at":
prettify_datetime(environment_obj.created_at),
"name":
environment_obj_name,
"description":
environment_obj_description
})
if download:
if not download_path:
# download to current working directory with timestamp
current_time = datetime.utcnow()
epoch_time = datetime.utcfromtimestamp(0)
current_time_unix_time_ms = (
current_time - epoch_time).total_seconds() * 1000.0
download_path = os.path.join(
self.environment_controller.home,
"environment_ls_" + str(current_time_unix_time_ms))
self.cli_helper.print_items(header_list,
item_dict_list,
print_format=print_format,
output_path=download_path)
return environment_objs
self.cli_helper.print_items(header_list,
item_dict_list,
print_format=print_format)
return environment_objs
class SnapshotController(BaseController):
"""SnapshotController inherits from BaseController and manages business logic related to snapshots
Parameters
----------
home : str
home path of the project
Attributes
----------
code : CodeController
file_collection : FileCollectionController
environment : EnvironmentController
Methods
-------
create(dictionary)
Create a snapshot within the project
checkout(id)
Checkout to a specific snapshot within the project
list(session_id=None)
List all snapshots present within the project based on given filters
delete(id)
Delete the snapshot specified from the project
"""
def __init__(self, home):
super(SnapshotController, self).__init__(home)
self.code = CodeController(home)
self.file_collection = FileCollectionController(home)
self.environment = EnvironmentController(home)
if not self.is_initialized:
raise ProjectNotInitializedException(
__("error", "controller.snapshot.__init__"))
def create(self, incoming_dictionary):
"""Create snapshot object
Parameters
----------
dictionary : dict
for each of the 5 key components, this function will search for
one of the variables below starting from the top. Default functionality
is described below for each component as well for reference if none
of the variables are given.
code :
code_id : str, optional
code reference associated with the snapshot; if not
provided will look to inputs below for code creation
commit_id : str, optional
commit id provided by the user if already available
Default
-------
commits will be taken and code created via the CodeController
and are added to the snapshot at the time of snapshot creation
environment :
environment_id : str, optional
id for environment used to create snapshot
environment_definition_filepath : str, optional
absolute filepath for the environment definition file
(e.g. Dockerfile path for Docker)
Default
-------
default environment files will be searched and environment will
be created with the EnvironmentController and added to the snapshot
at the time of snapshot creation
file_collection :
file_collection_id : str, optional
file collection associated with the snapshot
filepaths : list, optional
list of files or folder paths to include within the snapshot
Default
-------
filepaths will be considered empty ([]), and the FileCollectionController
will create a blank FileCollection that is empty.
config :
config : dict, optional
key, value pairs of configurations
config_filepath : str, optional
absolute filepath to configuration parameters file
config_filename : str, optional
name of file with configuration parameters
Default
-------
config will be considered empty ({}) and saved to the snapshot
stats :
stats : dict, optional
key, value pairs of metrics and statistics
stats_filepath : str, optional
absolute filepath to stats parameters file
stats_filename : str, optional
name of file with metrics and statistics.
Default
-------
stats will be considered empty ({}) and saved to the snapshot
for the remaining optional arguments it will search for them
in the input dictionary
message : str
long description of snapshot
session_id : str, optional
session id within which snapshot is created,
will overwrite default if given
task_id : str, optional
task id associated with snapshot
label : str, optional
short description of snapshot
visible : bool, optional
True if visible to user via list command else False
Returns
-------
Snapshot
Snapshot object as specified in datmo.core.entity.snapshot
Raises
------
RequiredArgumentMissing
if required arguments are not given by the user
FileIOException
if files are not present or there is an error in File IO
"""
# Validate Inputs
create_dict = {
"model_id": self.model.id,
"session_id": self.current_session.id,
}
# Message must be present
if "message" in incoming_dictionary:
create_dict['message'] = incoming_dictionary['message']
else:
raise RequiredArgumentMissing(
__("error", "controller.snapshot.create.arg", "message"))
# Code setup
self._code_setup(incoming_dictionary, create_dict)
# Environment setup
self._env_setup(incoming_dictionary, create_dict)
# File setup
self._file_setup(incoming_dictionary, create_dict)
# Config setup
self._config_setup(incoming_dictionary, create_dict)
# Stats setup
self._stats_setup(incoming_dictionary, create_dict)
# If snapshot object with required args already exists, return it
# DO NOT create a new snapshot with the same required arguments
results = self.dal.snapshot.query({
"model_id":
create_dict["model_id"],
"code_id":
create_dict['code_id'],
"environment_id":
create_dict['environment_id'],
"file_collection_id":
create_dict['file_collection_id'],
"config":
create_dict['config'],
"stats":
create_dict['stats']
})
if results: return results[0]
# Optional args for Snapshot entity
optional_args = ["task_id", "label", "visible"]
for optional_arg in optional_args:
if optional_arg in incoming_dictionary:
create_dict[optional_arg] = incoming_dictionary[optional_arg]
# Create snapshot and return
return self.dal.snapshot.create(Snapshot(create_dict))
def create_from_task(self, message, task_id):
"""Create snapshot from a completed task.
# TODO: enable create from task DURING a run
Parameters
----------
message : str
long description of snapshot
task_id : str
task object to use to create snapshot
Returns
-------
Snapshot
Snapshot object as specified in datmo.core.entity.snapshot
Raises
------
TaskNotComplete
if task specified has not been completed
"""
task_obj = self.dal.task.get_by_id(task_id)
if not task_obj.status and not task_obj.after_snapshot_id:
raise TaskNotComplete(
__("error", "controller.snapshot.create_from_task",
str(task_obj.id)))
return self.dal.snapshot.update({
"id": task_obj.after_snapshot_id,
"message": message,
"stats": task_obj.results,
"visible": True
})
def checkout(self, snapshot_id):
# Get snapshot object
snapshot_obj = self.dal.snapshot.get_by_id(snapshot_id)
code_obj = self.dal.code.get_by_id(snapshot_obj.code_id)
file_collection_obj = self.dal.file_collection.\
get_by_id(snapshot_obj.file_collection_id)
# Create new code_driver ref to revert back (if needed)
# TODO: Save this to be reverted to
current_code_obj = self.code.create()
# Checkout code_driver to the relevant commit ref
self.code_driver.checkout_ref(code_obj.commit_id)
# Pull file collection to the project home
dst_dirpath = os.path.join("datmo_snapshots", snapshot_id)
abs_dst_dirpath = self.file_driver.create(dst_dirpath, directory=True)
self.file_driver.transfer_collection(file_collection_obj.filehash,
abs_dst_dirpath)
return True
def list(self, session_id=None, visible=None):
query = {}
if session_id:
try:
self.dal.session.get_by_id(session_id)
except EntityNotFound:
raise SessionDoesNotExistException(
__("error", "controller.snapshot.list", session_id))
query['session_id'] = session_id
if visible is not None and isinstance(visible, bool):
query['visible'] = visible
return self.dal.snapshot.query(query)
def delete(self, snapshot_id):
if not snapshot_id:
raise RequiredArgumentMissing(
__("error", "controller.snapshot.delete.arg", "snapshot_id"))
return self.dal.snapshot.delete(snapshot_id)
def _code_setup(self, incoming_dictionary, create_dict):
""" Set the code_id by using:
1. code_id
2. commit_id string, which creates a new code_id
3. create a new code id
Parameters
----------
incoming_dictionary : dict
dictionary for the create function defined above
create_dict : dict
dictionary for creating the Snapshot entity
"""
if "code_id" in incoming_dictionary:
create_dict['code_id'] = incoming_dictionary['code_id']
elif "commit_id" in incoming_dictionary:
create_dict['code_id'] = self.code.\
create(commit_id=incoming_dictionary['commit_id']).id
else:
create_dict['code_id'] = self.code.create().id
def _env_setup(self, incoming_dictionary, create_dict):
""" TODO:
Parameters
----------
incoming_dictionary : dict
dictionary for the create function defined above
create_dict : dict
dictionary for creating the Snapshot entity
"""
language = incoming_dictionary.get("language", None)
if "environment_id" in incoming_dictionary:
create_dict['environment_id'] = incoming_dictionary[
'environment_id']
elif "environment_definition_filepath" in incoming_dictionary:
create_dict['environment_id'] = self.environment.create({
"definition_filepath":
incoming_dictionary['environment_definition_filepath']
}).id
elif language:
create_dict['environment_id'] = self.environment.\
create({"language": language}).id
else:
# create some default environment
create_dict['environment_id'] = self.environment.\
create({}).id
def _file_setup(self, incoming_dictionary, create_dict):
""" TODO:
Parameters
----------
incoming_dictionary : dict
dictionary for the create function defined above
create_dict : dict
dictionary for creating the Snapshot entity
"""
if "file_collection_id" in incoming_dictionary:
create_dict['file_collection_id'] = incoming_dictionary[
'file_collection_id']
elif "filepaths" in incoming_dictionary:
# transform file paths to file_collection_id
create_dict['file_collection_id'] = self.file_collection.\
create(incoming_dictionary['filepaths']).id
else:
# create some default file collection
create_dict['file_collection_id'] = self.file_collection.\
create([]).id
def _config_setup(self, incoming_dictionary, create_dict):
""" Fills in snapshot config by having one of the following:
1. config = JSON object
2. config_filepath = some location where a json file exists
3. config_filename = just the file nam
Parameters
----------
incoming_dictionary : dict
dictionary for the create function defined above
create_dict : dict
dictionary for creating the Snapshot entity
Raises
------
FileIOException
"""
if "config" in incoming_dictionary:
create_dict['config'] = incoming_dictionary['config']
elif "config_filepath" in incoming_dictionary:
if not os.path.isfile(incoming_dictionary['config_filepath']):
raise FileIOException(
__("error", "controller.snapshot.create.file_config"))
# If path exists transform file to config dict
config_json_driver = JSONStore(
incoming_dictionary['config_filepath'])
create_dict['config'] = config_json_driver.to_dict()
elif "config_filename" in incoming_dictionary:
config_filename = incoming_dictionary['config_filename'] \
if "config_filename" in incoming_dictionary else "config.json"
create_dict['config'] = self._find_in_filecollection(
config_filename, create_dict['file_collection_id'])
else:
create_dict['config'] = {}
def _stats_setup(self, incoming_dictionary, create_dict):
"""Fills in snapshot stats by having one of the following:
1. stats = JSON object
2. stats_filepath = some location where a json file exists
3. stats_filename = just the file name
Parameters
----------
incoming_dictionary : dict
dictionary for the create function defined above
create_dict : dict
dictionary for creating the Snapshot entity
Raises
------
FileIOException
"""
if "stats" in incoming_dictionary:
create_dict['stats'] = incoming_dictionary['stats']
elif "stats_filepath" in incoming_dictionary:
if not os.path.isfile(incoming_dictionary['stats_filepath']):
raise FileIOException(
__("error", "controller.snapshot.create.file_stat"))
# If path exists transform file to config dict
stats_json_driver = JSONStore(
incoming_dictionary['stats_filepath'])
create_dict['stats'] = stats_json_driver.to_dict()
elif "stats_filename" in incoming_dictionary:
stats_filename = incoming_dictionary['stats_filename'] \
if "stats_filename" in incoming_dictionary else "stats.json"
create_dict['stats'] = self._find_in_filecollection(
stats_filename, create_dict['file_collection_id'])
else:
create_dict['stats'] = {}
def _find_in_filecollection(self, file_to_find, file_collection_id):
""" Attempts to find a file within the file collection
Returns
-------
dict
output dictionary of the JSON file
"""
file_collection_obj = self.file_collection.dal.file_collection.\
get_by_id(file_collection_id)
file_collection_path = \
self.file_collection.file_driver.get_collection_path(
file_collection_obj.filehash)
# find all of the possible paths it could exist
possible_paths = [os.path.join(self.home, file_to_find)] + \
[os.path.join(self.home, item[0], file_to_find)
for item in os.walk(file_collection_path)]
existing_possible_paths = [
possible_path for possible_path in possible_paths
if os.path.isfile(possible_path)
]
if not existing_possible_paths:
# TODO: Add some info / warning that no file was found
# create some default stats
return {}
else:
# If any such path exists, transform file to stats dict
json_file = JSONStore(existing_possible_paths[0])
return json_file.to_dict()