diff options
Diffstat (limited to 'osdf/adapters/policy/utils.py')
-rw-r--r-- | osdf/adapters/policy/utils.py | 46 |
1 files changed, 32 insertions, 14 deletions
diff --git a/osdf/adapters/policy/utils.py b/osdf/adapters/policy/utils.py index 27885a5..9acfa2a 100644 --- a/osdf/adapters/policy/utils.py +++ b/osdf/adapters/policy/utils.py @@ -15,9 +15,13 @@ # # ------------------------------------------------------------------------- # +import copy +import json from collections import defaultdict +from osdf.utils.programming_utils import dot_notation, list_flatten + def group_policies(flat_policies): """Filter policies using the following steps: @@ -27,29 +31,30 @@ def group_policies(flat_policies): :param flat_policies: list of flat policies :return: Filtered policies """ - aggregated_policies = {} - filter_policies = defaultdict(list) + filtered_policies = defaultdict(list) policy_name = [] - for policy in flat_policies: + policies = [x for x in flat_policies if x['content'].get('policy_type')] # drop ones without 'policy_type' + policy_types = set([x['content'].get('policyType') for x in policies]) + aggregated_policies = dict((x, defaultdict(list)) for x in policy_types) + + for policy in policies: policy_type = policy['content'].get('policyType') - if not policy_type: - continue - if policy_type not in aggregated_policies: - aggregated_policies[policy_type] = defaultdict(list) for resource in policy['content'].get('resourceInstanceType', []): aggregated_policies[policy_type][resource].append(policy) + for policy_type in aggregated_policies: for resource in aggregated_policies[policy_type]: - if len(aggregated_policies[policy_type][resource]) > 0: + if aggregated_policies[policy_type][resource]: aggregated_policies[policy_type][resource].sort(key=lambda x: x['priority'], reverse=True) - policy = aggregated_policies[policy_type][resource][0] - if policy['policyName'] not in policy_name: - filter_policies[policy['content']['policyType']].append(policy) - policy_name.append(policy['policyName']) - return filter_policies + prioritized_policy = aggregated_policies[policy_type][resource][0] + if prioritized_policy['policyName'] not in policy_name: + # TODO: Check logic here... should policy appear only once across all groups? + filtered_policies[prioritized_policy['content']['policyType']].append(prioritized_policy) + policy_name.append(prioritized_policy['policyName']) + return filtered_policies -def _regex_policy_name(policy_name): +def policy_name_as_regex(policy_name): """Get the correct policy name as a regex (e.g. OOF_HAS_vCPE.cloudAttributePolicy ends up in policy as OOF_HAS_vCPE.Config_MS_cloudAttributePolicy.1.xml So, for now, we query it as OOF_HAS_vCPE..*aicAttributePolicy.*) @@ -58,3 +63,16 @@ def _regex_policy_name(policy_name): """ p = policy_name.partition('.') return p[0] + p[1] + ".*" + p[2] + ".*" + + +def retrieve_node(req_json, reference): + """ + Get the child node(s) from the dot-notation [reference] and parent [req_json]. + For placement and other requests, there are encoded JSONs inside the request or policy, + so we need to expand it and then do a search over the parent plus expanded JSON. + """ + req_json_copy = copy.deepcopy(req_json) # since we expand the JSON in place, we work on a copy + if 'orderInfo' in req_json_copy['placementInfo']: + req_json_copy['placementInfo']['orderInfo'] = json.loads(req_json_copy['placementInfo']['orderInfo']) + info = dot_notation(req_json_copy, reference) + return list_flatten(info) if isinstance(info, list) else info
\ No newline at end of file |