Remove the dependency on scipy #1754

pyproject.toml

@@ -100,7 +100,6 @@ dependencies = [
   "aboutcode.hashid==0.2.0",
   # AboutCode pipeline
   "aboutcode.pipeline==0.2.1",
-  "scipy==1.15.3",  # 1.16.x requires Python >=3.11
   # ScoreCode
   "scorecode==0.0.4",
   # Workaround issue https://github.com/aboutcode-org/scancode.io/issues/1885

scanpipe/pipes/symbolmap.py

@@ -20,12 +20,11 @@
 # ScanCode.io is a free software code scanning tool from nexB Inc. and others.
 # Visit https://github.com/aboutcode-org/scancode.io for support and download.
 
+import math
 from collections import Counter
 from dataclasses import dataclass
 from dataclasses import field
 
-from scipy.spatial.distance import jensenshannon
-
 from aboutcode.pipeline import LoopProgress
 from scanpipe.models import CodebaseRelation
 from scanpipe.pipes import flag
@@ -292,6 +291,21 @@ def get_symbols_probability_distribution(symbols, unique_symbols):
     return probability_dist
 
 
+def jensenshannon(p, q):
+    """Compute the Jensen-Shannon distance between two probability distributions."""
+    m = [(pi + qi) / 2.0 for pi, qi in zip(p, q)]
+
+    left = 0.0
+    right = 0.0
+    for pi, qi, mi in zip(p, q, m):
+        if pi > 0:
+            left += pi * math.log(pi / mi)
+        if qi > 0:
+            right += qi * math.log(qi / mi)
+
+    return math.sqrt((left + right) / 2.0)
+
+
 def get_similarity_between_source_and_deployed_symbols(
     source_symbols,
     deployed_symbols,

scanpipe/tests/pipes/test_symbolmap.py

@@ -277,6 +277,28 @@ def test_scanpipe_pipes_symbolmap_get_symbols_probability_distribution(self):
         # print(result_prob_dist)
         self.assertListEqual(result_prob_dist, expected_prob_dist)
 
+    def test_jensenshannon(self):
+        # Identical distributions -> distance is 0
+        self.assertEqual(symbolmap.jensenshannon([1.0, 0.0, 0.0], [1.0, 0.0, 0.0]), 0.0)
+
+        # Completely different distributions -> maximum distance
+        self.assertAlmostEqual(
+            symbolmap.jensenshannon([1.0, 0.0], [0.0, 1.0]), 0.8325546, places=5
+        )
+
+        # Partial overlap
+        self.assertAlmostEqual(
+            symbolmap.jensenshannon([1.0, 0.0], [0.5, 0.5]),
+            0.46450140402245893,
+            places=5,
+        )
+
+        # Uniform distributions -> distance is 0
+        self.assertEqual(
+            symbolmap.jensenshannon([0.25, 0.25, 0.25, 0.25], [0.25, 0.25, 0.25, 0.25]),
+            0.0,
+        )
+
     def test_get_similarity_between_source_and_deployed_symbols(
         self,
     ):