avoid value 0 in preshmap and helpful user warnings

spacy/kb.pxd

@@ -112,4 +112,23 @@ cdef class KnowledgeBase:
         self._alias_index[alias_key] = alias_index
         return alias_index
 
+    cdef inline create_empty_vectors(self):
+        """ 
+        Making sure the first element of each vector is a dummy,
+        because the PreshMap maps pointing to indices in these vectors can not contain 0 as value
+        cf. https://github.com/explosion/preshed/issues/17
+        """
+        cdef int32_t dummy_value = 0
+        self._entries.push_back(
+            _EntryC(
+                vector_rows=&dummy_value,
+                feats_row=dummy_value,
+                prob=dummy_value
+            ))
+        self._aliases_table.push_back(
+            _AliasC(
+                entry_indices=[dummy_value],
+                probs=[dummy_value]
+            ))
+
 

spacy/kb.pyx

@@ -1,5 +1,6 @@
 # cython: profile=True
 # coding: utf8
+from spacy.errors import user_warning
 
 cdef class KnowledgeBase:
 
@@ -8,6 +9,7 @@ cdef class KnowledgeBase:
         self._alias_index = PreshMap()
         self.mem = Pool()
         self.strings = StringStore()
+        self.create_empty_vectors()
 
     def __len__(self):
         return self.get_size_entities()
@@ -21,8 +23,9 @@ cdef class KnowledgeBase:
     def add_entity(self, unicode entity_id, float prob, vectors=None, features=None):
         cdef hash_t id_hash = self.strings.add(entity_id)
 
-        # TODO: more friendly check for non-unique name
+        # Return if this entity was added before
         if id_hash in self._entry_index:
+            user_warning("Entity " + entity_id + " already exists in the KB")
             return
 
         cdef int32_t dummy_value = 342
@@ -33,6 +36,12 @@ cdef class KnowledgeBase:
     def add_alias(self, unicode alias, entities, probabilities):
         """For a given alias, add its potential entities and prior probabilies to the KB."""
         cdef hash_t alias_hash = self.strings.add(alias)
+
+        # Return if this alias was added before
+        if alias_hash in self._alias_index:
+            user_warning("Alias " + alias + " already exists in the KB")
+            return
+
         cdef hash_t entity_hash
 
         cdef vector[int64_t] entry_indices
@@ -47,12 +56,12 @@ cdef class KnowledgeBase:
             entry_indices.push_back(int(entry_index))
             probs.push_back(float(prob))
 
-        # TODO: check that alias hadn't been defined before
         # TODO: check sum(probabilities) <= 1
         # TODO: check len(entities) == len(probabilities)
 
         self.c_add_aliases(alias_key=alias_hash, entry_indices=entry_indices, probs=probs)
 
+
     def get_candidates(self, unicode alias):
         cdef hash_t alias_hash = self.strings.add(alias)
         alias_index = <int64_t>self._alias_index.get(alias_hash)

spacy/sandbox_test_sofie/testing_el.py

@@ -1,23 +1,28 @@
+# coding: utf-8
 import spacy
 from spacy.kb import KnowledgeBase
 
 
 def create_kb():
     mykb = KnowledgeBase()
+
     print("kb size", len(mykb), mykb.get_size_entities(), mykb.get_size_aliases())
 
     # adding entities
     entity_0 = "Q0"  # douglas adams
-    mykb.add_entity(entity_id=entity_0, prob=0.5)
     print(" adding entity", entity_0)
+    mykb.add_entity(entity_id=entity_0, prob=0.5)
 
     entity_42 = "Q42"   # douglas adams
-    mykb.add_entity(entity_id=entity_42, prob=0.5)
     print(" adding entity", entity_42)
+    mykb.add_entity(entity_id=entity_42, prob=0.5)
 
     entity_5301561 = "Q5301561"
-    mykb.add_entity(entity_id=entity_5301561, prob=0.5)
     print(" adding entity", entity_5301561)
+    mykb.add_entity(entity_id=entity_5301561, prob=0.5)
+
+    print(" adding entity", entity_5301561)
+    mykb.add_entity(entity_id=entity_5301561, prob=0.5)
 
     print("kb size", len(mykb), mykb.get_size_entities(), mykb.get_size_aliases())
 
@@ -32,6 +37,15 @@ def create_kb():
     candidates = mykb.get_candidates(alias)
     print(" ", candidates)
 
+    print(" adding alias", alias)
+    mykb.add_alias(alias=alias, entities=["Q42"], probabilities=[0.9])
+
+    print("kb size", len(mykb), mykb.get_size_entities(), mykb.get_size_aliases())
+
+    print("candidates for", alias)
+    candidates = mykb.get_candidates(alias)
+    print(" ", candidates)
+
 
 def add_el():
     nlp = spacy.load('en_core_web_sm')