def _try_deleting_ranker(self, ranker_id):
config = load_config()
url, user_id, password = get_rnr_credentials(config)
response = requests.get(path.join(url, 'v1/rankers', ranker_id),
auth=(user_id, password),
headers={
'x-global-transaction-id': 'Rishavs app',
'Content-type': 'application/json'
})
response_text = json.dumps(response.json(), indent=4, sort_keys=True)
if response.status_code == 200:
self.logger.info('Found a test ranker that needs cleanup: %s' %
response_text)
response = requests.delete(path.join(url, 'v1/rankers', ranker_id),
auth=(user_id, password),
headers={
'x-global-transaction-id':
'Rishavs app',
'Content-type': 'application/json'
})
response.raise_for_status()
self.logger.info("Successfully deleted test ranker: %s" %
ranker_id)
else:
self.logger.info(
'No cleanup required for ranker id: %s (got response: %s)' %
(ranker_id, response_text))
def __init__(self, config=load_config(), logger=initialize_logger(logging.INFO, 'BluemixServiceProxy')):
"""
:param ConfigParser config: for access to credentials
:param logging.Logger logger: for logging
"""
self.logger = logger
self.config = config
self.bluemix_url, self.bluemix_user, self.bluemix_password = get_rnr_credentials(config)
self.bluemix_connection = RetrieveAndRankV1(url=self.bluemix_url, username=self.bluemix_user,
password=self.bluemix_password)
def get_discovery_credentials(config=load_config()):
url = config.get('Discovery', 'url', fallback=discovery_v1.default_url)
user = config.get('Discovery', 'user')
password = config.get('Discovery', 'password')
version = config.get('Discovery',
'version',
fallback=discovery_v1.latest_version)
return {
'url': url,
'version': version,
'username': user,
'password': password
}
def __init__(self,
config=load_config(),
logger=initialize_logger(logging.INFO, 'DiscoveryProxy')):
"""
Initialize the connection to Bluemix
:param CofigParser config: An initialized config with user, password and environment id
"""
self.config = config
self.logger = logger
self.discovery = initialize_discovery_service(config)
self.environment_id = config.get(
'Discovery',
'environment_id',
fallback=search_for_byod_environment_id(self.discovery))
def _delete_test_cluster(self, cluster_id):
self.logger.info("Attempting to clean up the test cluster that was spun up for the unit test: %s" % cluster_id)
config = load_config()
url, user_id, password = get_rnr_credentials(config)
response = requests.get('%s/v1/solr_clusters/%s' % (url, cluster_id),
auth=(user_id, password),
headers={'x-global-transaction-id': 'Rishavs app',
'Content-type': 'application/json'})
response_text = json.dumps(response.json(), indent=4, sort_keys=True)
if response.status_code == 200:
self.logger.info('Found a test cluster that needs cleanup: %s' % response_text)
response = requests.delete('%s/v1/solr_clusters/%s' % (url, cluster_id),
auth=(user_id, password),
headers={'x-global-transaction-id': 'Rishavs app',
'Content-type': 'application/json'})
response.raise_for_status()
self.logger.info("Successfully deleted test cluster: %s" % cluster_id)
else:
self.logger.info('No cleanup required for cluster id: %s (got response: %s)' % (cluster_id, response_text))
import logging
from argparse import ArgumentParser, ArgumentDefaultsHelpFormatter
from os import path
from rnr_debug_helpers.utils.rnr_wrappers import RetrieveAndRankProxy
from rnr_debug_helpers.utils.io_helpers import initialize_logger, load_config
LOGGER = initialize_logger(logging.INFO, path.basename(__file__))
CONFIG = load_config()
# TODO: make the cluster size an optional parameter
_CLUSTER_SIZE = 2
def main(args):
LOGGER.info('Start Script')
solr_cluster = RetrieveAndRankProxy(solr_cluster_id=args.cluster_id,
cluster_name='TestCluster',
cluster_size=_CLUSTER_SIZE)
LOGGER.info('Initialized bluemix connection to solr cluster: %s' %
solr_cluster.solr_cluster_id)
solr_cluster.setup_cluster_and_collection(
config_id=args.config_id,
config_zip=args.config_path,
collection_id=args.collection_name)
LOGGER.info(
'Initialized a document collection: %s (there are %d docs in the collection)'
% (args.collection_name,
solr_cluster.get_num_docs_in_collection(args.collection_name)))
def generate_rnr_features(in_query_stream, outfile, collection_id, cluster_id, num_rows=30, config=load_config()):
"""
Iterates over a labelled query stream and generates a feature file with the columns:
<query_num>,<answer_id>,<fea_0>,<fea_1>,...,<fea_n>,<relevance_label>
:param rnr_debug_helpers.queries.LabelledQueryStream in_query_stream:
:param File outfile: where the feature file contents will be written to
:param str collection_id: the RnR solr collection to use for finding search results
:param str cluster_id: the RnR solr cluster id to use for finding search results
:param int or None num_rows: The number of search results that will be retrieved for each query. Defaults to 30
similar to RnR Web UI/Tooling
:param ConfigParser config: A config loaded with the credentials to use
"""
rnr_cluster = RetrieveAndRankProxy(solr_cluster_id=cluster_id, config=config)
writer = csv.writer(outfile)
# Iterate over queries and generate feature vectors
stats = defaultdict(int)
is_first_row = True
for qid, query in enumerate(in_query_stream):
labels_for_relevant_answer_ids = _parse_correct_answer_ids_from_query(query)
_collect_stats(stats, labels_for_relevant_answer_ids)
LOGGER.debug("Getting feature vectors for query:<<%s>>" % query.get_qid())
rnr_search_results = rnr_cluster.get_fcselect_features(query_text=query.get_qid(), collection_id=collection_id,
generate_header=is_first_row,
num_results_to_return=num_rows)
if len(rnr_search_results) == 0:
stats["num_queries_with_zero_rnr_results"] += 1
else:
if is_first_row:
writer.writerow([_QID_COLUMN_NAME] + rnr_search_results.pop(0) + [_GT_COLUMN_NAME])
is_first_row = False
stats["num_queries_with_atleast_one_search_result"] += 1
stats['num_search_results_retrieved'] += len(rnr_search_results)
num_possible_correct, num_correct_answers_in_search_results = \
_print_feature_vectors_and_check_for_correct_answers(writer, rnr_search_results, '%d' % (qid + 1),
labels_for_relevant_answer_ids)
if num_possible_correct != num_correct_answers_in_search_results:
stats['num_queries_where_at_least_correct_answer_didnt_appear_in_rnr'] += 1
stats["num_correct_in_search_result"] += num_correct_answers_in_search_results
if stats["num_queries"] % 100 == 0:
LOGGER.info("Processed %d queries from input file" % stats['num_queries'])
_average_stats_across_collection(stats)
LOGGER.info("Finished processing %d queries from input file" % stats['num_queries'])
return stats
def initialize_discovery_service(config=load_config()):
return discovery_v1.DiscoveryV1(**get_discovery_credentials(config))
