formatted and added test

INCATools · Aug 9, 2023 · 903b816 · 903b816
1 parent 24b3294
commit 903b816
Show file tree

Hide file tree

Showing 2 changed files with 47 additions and 12 deletions.
diff --git a/src/oaklib/implementations/semsimian/semsimian_implementation.py b/src/oaklib/implementations/semsimian/semsimian_implementation.py
@@ -64,7 +64,9 @@ def __post_init__(self):
             if not any(attr.startswith(s) for s in ["class_", "_"])
         ]
 
-    def _get_semsimian_object(self, predicates: List[PRED_CURIE] = None) -> Semsimian:
+    def _get_semsimian_object(
+        self, predicates: List[PRED_CURIE] = None, attributes: List[str] = None
+    ) -> Semsimian:
         """
         Get Semsimian object from "semsimian_object_cache" or add a new one.
 
@@ -79,9 +81,7 @@ def _get_semsimian_object(self, predicates: List[PRED_CURIE] = None) -> Semsimia
                     include_entailed=True, predicates=predicates
                 )
             ]
-            self.semsimian_object_cache[predicates] = Semsimian(
-                spo, self.term_pairwise_similarity_attributes
-            )
+            self.semsimian_object_cache[predicates] = Semsimian(spo, attributes)
 
         return self.semsimian_object_cache[predicates]
 
@@ -108,7 +108,9 @@ def pairwise_similarity(
         :return:
         """
         logging.debug(f"Calculating pairwise similarity for {subject} x {object} over {predicates}")
-        semsimian = self._get_semsimian_object(predicates=predicates)
+        semsimian = self._get_semsimian_object(
+            predicates=predicates, attributes=self.term_pairwise_similarity_attributes
+        )
 
         jaccard_val = semsimian.jaccard_similarity(subject, object, set(predicates))
 
@@ -163,7 +165,9 @@ def all_by_all_pairwise_similarity(
         """
         objects = list(objects)
         logging.info(f"Calculating all-by-all pairwise similarity for {len(objects)} objects")
-        semsimian = self._get_semsimian_object(predicates=predicates)
+        semsimian = self._get_semsimian_object(
+            predicates=predicates, attributes=self.term_pairwise_similarity_attributes
+        )
         all_results = semsimian.all_by_all_pairwise_similarity(
             subject_terms=set(subjects),
             object_terms=set(objects),

diff --git a/tests/test_implementations/test_semsimian_implementation.py b/tests/test_implementations/test_semsimian_implementation.py
@@ -1,4 +1,5 @@
 import os
+import timeit
 import unittest
 
 from linkml_runtime.dumpers import yaml_dumper
@@ -45,6 +46,14 @@ def setUp(self) -> None:
         self.other_oi = comparison_oi
         self.db = db
         self.compliance_tester = ComplianceTester(self)
+        self.subject_terms = {VACUOLE, NUCLEUS, NUCLEAR_MEMBRANE}
+        self.object_terms = {ENDOMEMBRANE_SYSTEM, HUMAN, FUNGI}
+        self.predicates = {IS_A, PART_OF}
+        self.term_pairwise_similarity_attributes = [
+            attr
+            for attr in vars(TermPairwiseSimilarity)
+            if not any(attr.startswith(s) for s in ["class_", "_"])
+        ]
 
     def test_definitions(self):
         """Definitions should be delegated to the wrapped adapter."""
@@ -65,7 +74,7 @@ def test_compare_pairwise_similarity(self):
         debug = False
         for s in entities:
             for o in entities:
-                for preds in [[IS_A, PART_OF]]:
+                for preds in [self.predicates]:
                     sim = adapter.pairwise_similarity(s, o, predicates=preds)
 
                     original_sim = self.other_oi.pairwise_similarity(s, o, predicates=preds)
@@ -97,10 +106,9 @@ def test_compare_pairwise_similarity(self):
                         print(yaml_dumper.dumps(original_sim))
 
     def test_all_by_all_pairwise_similarity(self):
-        subject_terms = {VACUOLE, NUCLEUS, NUCLEAR_MEMBRANE}
-        object_terms = {ENDOMEMBRANE_SYSTEM, HUMAN, FUNGI}
-        predicates = {IS_A, PART_OF}
-        result = self.oi.all_by_all_pairwise_similarity(subject_terms, object_terms, predicates)
+        result = self.oi.all_by_all_pairwise_similarity(
+            self.subject_terms, self.object_terms, self.predicates
+        )
         sem_similarity_object: TermPairwiseSimilarity = [
             x for x in result if x.subject_id == "GO:0031965" and x.object_id == "GO:0012505"
         ][0]
@@ -110,7 +118,7 @@ def test_all_by_all_pairwise_similarity(self):
         self.assertEqual(sem_similarity_object.phenodigm_score, 1.672622556711612)
 
         result2 = self.other_oi.all_by_all_pairwise_similarity(
-            subject_terms, object_terms, predicates
+            self.subject_terms, self.object_terms, self.predicates
         )
 
         sql_similarity_object: TermPairwiseSimilarity = [
@@ -128,3 +136,26 @@ def test_all_by_all_pairwise_similarity(self):
         self.assertAlmostEqual(
             sem_similarity_object.phenodigm_score, sql_similarity_object.phenodigm_score, places=2
         )
+
+    def test_semsimian_object_cache(self):
+        start_time = timeit.default_timer()
+        _ = list(
+            self.oi.all_by_all_pairwise_similarity(
+                self.subject_terms, self.object_terms, self.predicates
+            )
+        )
+        end_time = timeit.default_timer()
+        time_taken_1 = end_time - start_time
+
+        shuffled_predicate = set(reversed(list(self.predicates)))
+        start_time = timeit.default_timer()
+        _ = list(
+            self.oi.all_by_all_pairwise_similarity(
+                self.subject_terms, self.object_terms, shuffled_predicate
+            )
+        )
+        end_time = timeit.default_timer()
+        time_taken_2 = end_time - start_time
+
+        self.assertEqual(len(self.oi.semsimian_object_cache), 1)
+        self.assertTrue(time_taken_1 > time_taken_2)