def _get_confusion_matrix_action(self, **kwargs):
from tasks import calculate_confusion_matrix
parser = reqparse.RequestParser()
parser.add_argument('weights', type=str)
args = parser.parse_args()
test = self._get_details_query(None, **kwargs)
if not test:
raise NotFound('Test not found')
model = Model.query.get(kwargs.get('model_id'))
if not model:
raise NotFound('Model not found')
try:
arg = args.get("weights")
import json
json_weights = json.loads(arg)
if "weights_list" not in json_weights or \
not json_weights["weights_list"]:
raise ValueError("Weights list is empty")
weights = []
for w in json_weights["weights_list"]:
if not ("label" in w and "value" in w):
raise ValueError("Weights list is incorrect")
weights.append((w["label"], float(w["value"])))
calculate_confusion_matrix.delay(test.id, weights)
except Exception as e:
return self._render({self.OBJECT_NAME: test.id,
'error': e.message})
return self._render({self.OBJECT_NAME: test.id})
def _get_server(self, kwargs):
server_id = kwargs.get('server_id')
if server_id is None:
raise NotFound('Need to specify server_id')
server = Server.query.get(server_id)
if server is None:
raise NotFound(self.MESSAGE404 % kwargs)
return server
def _get_pig_fields_action(self, **kwargs):
if 'id' not in kwargs:
raise ValueError("Specify id of the datasource")
ds = self._get_details_query({}, **kwargs)
if ds is None:
raise NotFound('DataSet not found')
fields_data = []
for key, val in ds.pig_row.iteritems():
if isint(val):
data_type = 'integer'
elif isfloat(val):
data_type = 'float'
else:
data_type = 'string'
fields_data.append({'column_name': key, 'data_type': data_type})
from ..utils import XML_FIELD_TEMPLATE
xml = "\r\n".join(
[XML_FIELD_TEMPLATE % field for field in fields_data])
return self._render({
'sample_xml': xml,
'fields': fields_data,
'pig_result_line': ds.pig_row,
})
def _get_handler(self, handler_id):
if handler_id is None:
raise ValidationError('Please specify import handler')
handler = XmlImportHandler.query.get(handler_id)
if handler is None:
raise NotFound()
return handler
def _get_exports_action(self, **kwargs):
test = self._get_details_query(None, **kwargs)
if not test:
raise NotFound('Test not found')
return self._render({self.OBJECT_NAME: test.id,
'exports': test.exports,
'db_exports': test.db_exports})
def _get_pig_fields_action(self, **kwargs):
sqoop = self._get_details_query({}, **kwargs)
if sqoop is None:
raise NotFound(self.MESSAGE404 % kwargs)
return self._render({
self.OBJECT_NAME: sqoop.id,
'pig_fields': sqoop.pig_fields
})
def _get_sample_data_action(self, **kwargs):
ds = self._get_details_query({}, **kwargs)
if ds is None:
raise NotFound('DataSet not found')
if not os.path.exists(ds.filename):
raise NotFound('DataSet file cannot be found')
_, ext = os.path.splitext(ds.filename)
open_fn = gzip.open if ext == '.gz' else open if ext == '' else None
if not open_fn:
raise ValidationError('DataSet has unknown file type')
lines = []
params = self._parse_parameters([('size', int)])
sample_size = params.get('size') or 10
with open_fn(ds.filename, 'rb') as f:
line = f.readline()
while line and len(lines) < sample_size:
lines.append(json.loads(line))
line = f.readline()
return self._render(lines)
def _get_script_string_action(self, **kwargs):
script = self._get_details_query({}, **kwargs)
if script is None:
raise NotFound(self.MESSAGE404 % kwargs)
try:
return self._render({
self.OBJECT_NAME: script.id,
'script_string': script.script_string
})
except Exception as e:
return odesk_error_response(400, ERR_INVALID_DATA, str(e), e)
def _get_details_query(self, params, **kwargs):
ver_example = super(VerificationExampleResource, self).\
_get_details_query(params, **kwargs)
if ver_example is None:
raise NotFound()
if not ver_example.example.is_weights_calculated:
ver_example.example.calc_weighted_data()
return ver_example
def _get_download_action(self, **kwargs):
model = self._get_details_query(None, None, **kwargs)
if model is None:
raise NotFound(self.MESSAGE404 % kwargs)
data = json.dumps(model.to_dict())
resp = Response(data)
resp.headers['Content-Type'] = 'text/plain'
resp.headers['Content-Disposition'] = \
'attachment; filename=%s.json' % model.name
return resp
def _get_features_download_action(self, **kwargs):
model = self._get_details_query(None, **kwargs)
if model is None:
raise NotFound(self.MESSAGE404 % kwargs)
content = model.get_features_json()
resp = Response(content)
resp.headers['Content-Type'] = 'application/json'
resp.headers['Content-Disposition'] = \
'attachment; filename="%s-features.json"' % model.name
return resp
def _get_xml_download_action(self, **kwargs):
handler = self._get_details_query(None, **kwargs)
if handler is None:
raise NotFound(self.MESSAGE404 % kwargs)
content = handler.data
resp = Response(content)
resp.headers['Content-Type'] = 'text/xml'
resp.headers['Content-Disposition'] = \
'attachment; filename="%s-importhandler.xml"' % handler.name
return resp
def post(self, action=None, **kwargs):
if action == 'get_auth_url':
auth_url, oauth_token, oauth_token_secret =\
User.get_auth_url()
# TODO: Use redis?
# app.db['auth_tokens'].insert({
# 'oauth_token': oauth_token,
# 'oauth_token_secret': oauth_token_secret,
# })
auth = AuthToken(oauth_token, oauth_token_secret)
auth.save()
logging.debug("User Auth: oauth token %s added", oauth_token)
return self._render({'auth_url': auth_url})
if action == 'authenticate':
parser = reqparse.RequestParser()
parser.add_argument('oauth_token', type=str)
parser.add_argument('oauth_verifier', type=str)
params = parser.parse_args()
oauth_token = params.get('oauth_token')
oauth_verifier = params.get('oauth_verifier')
logging.debug("User Auth: trying to authenticate with token %s",
oauth_token)
# TODO: Use redis?
auth = AuthToken.get_auth(oauth_token)
if not auth:
logging.error('User Auth: token %s not found', oauth_token)
return odesk_error_response(
500, 500, 'Wrong token: {0!s}'.format(oauth_token))
oauth_token_secret = auth.get('oauth_token_secret')
auth_token, user = User.authenticate(oauth_token,
oauth_token_secret,
oauth_verifier)
logging.debug('User Auth: Removing token %s', oauth_token)
AuthToken.delete(auth.get('oauth_token'))
return self._render({'auth_token': auth_token, 'user': user})
if action == 'get_user':
user = getattr(request, 'user', None)
if user:
return self._render({'user': user})
return odesk_error_response(401, 401, 'Unauthorized')
logging.error('User Auth: invalid action %s', action)
raise NotFound('Action not found')
def _get_details_query(self, params, **kwargs):
example = super(TestExampleResource, self)._get_details_query(
params, **kwargs)
if example is None:
raise NotFound()
fields = self._get_show_fields(params)
if 'next' in fields or 'previous' in fields:
from sqlalchemy.sql import select, func, text, bindparam
from models import db
filter_params = kwargs.copy()
filter_params.update(self._prepare_filter_params(params))
filter_params.pop('id')
sort_by = params.get('sort_by', None) or 'id'
is_desc = params.get('order', None) == 'desc'
fields_to_select = [TestExample.id]
# TODO: simplify query with specifying WINDOW w
if 'previous' in fields:
fields_to_select.append(
func.lag(TestExample.id).over(
order_by=[sort_by, 'id']).label('prev'))
if 'next' in fields:
fields_to_select.append(
func.lead(TestExample.id).over(
order_by=[sort_by, 'id']).label('next'))
tbl = select(fields_to_select)
for name, val in filter_params.iteritems():
if '->>' in name: # TODO: refactor this
try:
splitted = name.split('->>')
name = "%s->>'%s'" % (splitted[0], splitted[1])
except:
logging.warning('Invalid GET param %s', name)
tbl.append_whereclause("%s='%s'" % (name, val))
tbl = tbl.cte('tbl')
select1 = select(['id', 'prev', 'next']).where(
tbl.c.id == kwargs['id'])
res = db.engine.execute(select1, id_1=kwargs['id'])
id_, example.previous, example.next = res.fetchone()
if not example.is_weights_calculated:
example.calc_weighted_data()
example = super(TestExampleResource, self)._get_details_query(
params, **kwargs)
return example
def _put_transformers_download_action(self, **kwargs):
model = self._get_details_query(None, **kwargs)
if model is None:
raise NotFound('Model not found')
if model.status != Model.STATUS_TRAINED:
return odesk_error_response(405, ERR_INVALID_METHOD,
'Model is not trained')
form = TransformersDownloadForm(obj=model)
if not form.is_valid():
return
segment = form.cleaned_data['segment']
segment_obj = Segment.query.filter(Segment.model_id == model.id)\
.filter(Segment.name == segment).all()
if not len(segment_obj):
raise NotFound('Segment not found in trained model')
data_format = form.cleaned_data['data_format']
from api.ml_models.tasks import upload_segment_features_transformers
upload_segment_features_transformers.delay(model.id, segment_obj[0].id,
data_format)
return self._render({})
def _put_pig_fields_action(self, **kwargs):
sqoop = self._get_details_query({}, **kwargs)
if sqoop is None:
raise NotFound(self.MESSAGE404 % kwargs)
from ..forms import LoadPigFieldsForm
form = LoadPigFieldsForm(obj={})
if form.is_valid():
from api.import_handlers.tasks import load_pig_fields
params = form.cleaned_data.get('params')
load_pig_fields.delay(sqoop.id, params)
return self._render({
'result':
"Generating pig fields delayed "
"(link will appear in sqoop section)"
})
return odesk_error_response(400, 400, 'Parameters are invalid')
def _put_dataset_download_action(self, **kwargs):
model = self._get_details_query(None, **kwargs)
if model is None:
raise NotFound('Model not found')
if model.status != Model.STATUS_TRAINED:
return odesk_error_response(400, ERR_INVALID_DATA,
'Model is not trained')
form = TransformDataSetForm(obj=model)
if not form.is_valid():
return
dataset = form.cleaned_data['dataset']
from api.ml_models.tasks import transform_dataset_for_download
transform_dataset_for_download.delay(model.id, dataset.id)
return self._render({})
def _get_dataset_download_action(self, **kwargs):
model = self._get_details_query(None, **kwargs)
if model is None:
raise NotFound('Model not found')
from api.tasks import TRANSFORM_DATASET_TASK
tasks = AsyncTask.get_current_by_object(model, TRANSFORM_DATASET_TASK)
downloads = []
for task in tasks:
downloads.append({
'dataset': DataSet.query.get(task.args[1]),
'task': task
})
return self._render({
self.OBJECT_NAME: model.id,
'downloads': downloads
})
def _put_csv_task_action(self, model_id, test_result_id):
"""
Schedules a task to generate examples in CSV format
"""
test = TestResult.query.get(test_result_id)
if not test:
raise NotFound('Test not found')
form = SelectFieldsForCSVForm(obj=test)
if form.is_valid():
fields = form.cleaned_data['fields']
if isinstance(fields, list) and len(fields) > 0:
from tasks import get_csv_results
logging.info('Download examples in csv')
get_csv_results.delay(test.model_id, test.id, fields)
return self._render({})
return odesk_error_response(400, ERR_INVALID_DATA,
'Fields of the CSV export is required')
def _put_run_sql_action(self, **kwargs):
"""
Run sql query for testing
"""
from api.import_handlers.forms import QueryTestForm
model = self._get_details_query({}, **kwargs)
if model is None:
raise NotFound(self.MESSAGE404 % kwargs)
form = QueryTestForm(obj={})
if not form.is_valid():
return self._render({'error': form.error_messages})
sql = form.cleaned_data['sql']
limit = form.cleaned_data['limit']
params = form.cleaned_data.get('params', {})
datasource_name = form.cleaned_data['datasource']
try:
sql = re.sub('#{(\w+)}', '%(\\1)s', sql)
sql = sql % params
except (KeyError, ValueError) as e:
return odesk_error_response(400, ERR_INVALID_DATA,
'Wrong query parameters', e)
try:
model.check_sql(sql)
except Exception as e:
return odesk_error_response(400, ERR_INVALID_DATA, str(e), e)
# Change query LIMIT
sql = model.build_query(sql, limit=limit)
try:
data = list(model.execute_sql_iter(sql, datasource_name))
except DatabaseError as e:
return odesk_error_response(400, ERR_INVALID_DATA, str(e), e)
columns = []
if len(data) > 0:
columns = data[0].keys()
return self._render({'data': data, 'columns': columns, 'sql': sql})
def _get_weights_download_action(self, **kwargs):
model = self._get_details_query(None, **kwargs)
if model is None:
raise NotFound(self.MESSAGE404 % kwargs)
if model.status != model.STATUS_TRAINED:
raise ValidationError('Model should be trained')
if not model.weights_synchronized:
raise ValidationError('Model weights should be synchronized')
trainer = model.get_trainer(force_load=True)
result = {}
for segment in model.segments:
result[segment.name] = trainer.get_weights(segment.name)
content = json.dumps(result)
resp = Response(content)
resp.headers['Content-Type'] = 'application/json'
resp.headers['Content-Disposition'] = \
'attachment; filename=%s-weights.json' % model.name
return resp
def _put_db_task_action(self, model_id, test_result_id):
"""
Schedules a task to export examples to the specified DB
"""
test = TestResult.query.get(test_result_id)
if not test:
raise NotFound('Test not found')
form = ExportToDbForm(obj=test)
if form.is_valid():
fields = form.cleaned_data['fields']
datasource = form.cleaned_data['datasource']
tablename = form.cleaned_data['tablename']
if isinstance(fields, list) and len(fields) > 0:
from tasks import export_results_to_db
logging.info('Export examples to db')
export_results_to_db.delay(
test.model_id, test.id, datasource.id, tablename, fields)
return self._render({})
return odesk_error_response(400, ERR_INVALID_DATA,
'Fields of the DB export is required')
def _get_transformers_download_action(self, **kwargs):
model = self._get_details_query(None, **kwargs)
if model is None:
raise NotFound('Model not found')
from api.tasks import TRANSFORMERS_UPLOAD_TASK
tasks = []
for segment in model.segments:
tasks.extend(
AsyncTask.get_current_by_object(segment,
TRANSFORMERS_UPLOAD_TASK))
downloads = []
for task in sorted(tasks, key=lambda x: x.created_on, reverse=True):
downloads.append({
'segment': Segment.query.get(task.args[1]),
'task': task
})
return self._render({
self.OBJECT_NAME: model.id,
'downloads': downloads
})
def _get_folder(self, kwargs):
folder = kwargs.get('folder')
if folder not in self.ALLOWED_FOLDERS:
raise NotFound(self.MESSAGE404 % kwargs)
return folder
def _get_generate_url_action(self, **kwargs):
ds = self._get_details_query({}, **kwargs)
if ds is None:
raise NotFound('DataSet not found')
url = ds.get_s3_download_url()
return self._render({self.OBJECT_NAME: ds.id, 'url': url})
def _get_trainer_download_s3url_action(self, **kwargs):
entity = self._get_details_query(None, **kwargs)
if entity is None:
raise NotFound(self.MESSAGE404 % kwargs)
url = entity.get_trainer_s3url()
return self._render({'trainer_file_for': entity.id, 'url': url})
def _get_uid(self, kwargs):
uid = kwargs.get('id')
if uid is None:
raise NotFound(self.MESSAGE404 % kwargs)
return uid