From 5f180883434ef6a0305e3df3bb5535e70c955ef2 Mon Sep 17 00:00:00 2001
From: semio <semio@posteo.de>
Date: Mon, 11 Nov 2019 17:17:04 +0800
Subject: [PATCH] new constructor for entity domain

related issue #119
---
 ddf_utils/model/ddf.py     | 26 +++++++++++++++++++++++---
 ddf_utils/model/package.py |  5 ++++-
 2 files changed, 27 insertions(+), 4 deletions(-)

diff --git a/ddf_utils/model/ddf.py b/ddf_utils/model/ddf.py
index 5a84767..865bb9c 100644
--- a/ddf_utils/model/ddf.py
+++ b/ddf_utils/model/ddf.py
@@ -80,7 +80,7 @@ class EntityDomain:
 
     @entities.validator
     def _check_entities_identity(self, attribute, value):
-        entities_id_list = [x.id for x in value]
+        entities_id_list = self.entity_ids
         counter = Counter(entities_id_list)
         error = False
         for k, v in counter.items():
@@ -90,6 +90,27 @@ def _check_entities_identity(self, attribute, value):
         if error:
             raise ValueError("duplicated entity detected")
 
+    @classmethod
+    def from_entity_list(cls, domain_id, entities, allow_duplicated=True, **kwargs):
+        if not allow_duplicated:
+            return cls(id=domain_id, entities=entities, props=kwargs)
+        # if there are duplicates, we need to combine all duplicates
+        # now construct a new entities list without duplicates
+        entity_ids = [x.id for x in entities]
+        entities_new = dict((i, Entity(id=i, domain=domain_id, sets=[], props={})) for i in entity_ids)
+        for x in entities:
+            en = entities_new[x.id]
+            for s in x.sets:
+                if s not in en.sets:
+                    en.sets.append(s)
+            en.props.update(x.props)
+            entities_new[x.id] = en
+        return cls(id=domain_id, entities=list(entities_new.values()), props=kwargs)
+
+    @property
+    def entity_ids(self):
+        return [x.id for x in self.entities]
+
     @property
     def entity_sets(self):
         sets = set()
@@ -102,8 +123,7 @@ def get_entity_set(self, s):
         return [e for e in self.entities if s in e.sets]
 
     def has_entity(self, sid):
-        all_ids = [e.id for e in self.entities]
-        return sid in all_ids
+        return sid in self.entity_ids
 
     def to_dict(self, eset=None):
         if eset:
diff --git a/ddf_utils/model/package.py b/ddf_utils/model/package.py
index 6220e6f..74a4e58 100644
--- a/ddf_utils/model/package.py
+++ b/ddf_utils/model/package.py
@@ -252,7 +252,10 @@ def load_ddf(self):
             domains_tmp[domain].append(entity)
 
         for domain, entities_ in domains_tmp.items():
-            domains[domain] = EntityDomain(id=domain, entities=entities_)
+            # TODO: maybe get properties from concepts table
+            # Allow duplicated entity because they may be defined in multiple resources
+            # i.e. multiple entity sets in separated files.
+            domains[domain] = EntityDomain.from_entity_list(domain_id=domain, entities=entities_, allow_duplicated=True)
 
         # load datapoints. Here we will use Dask for all
         # 1. create categories for entity domains