Improve simhash performance

build.py

@@ -5,6 +5,7 @@
 The shared library can also be built manually using the command:
 $ cythonize -X language_level=3 -a -i ./iscc_core/cdc.py
 $ cythonize -X language_level=3 -a -i ./iscc_core/minhash.py
+$ cythonize -X language_level=3 -a -i ./iscc_core/simhash.py
 """
 try:
     from Cython.Build import cythonize, build_ext
@@ -43,10 +44,7 @@ def build(setup_kwargs):
             setup_kwargs.update(
                 dict(
                     ext_modules=cythonize(
-                        [
-                            "iscc_core/cdc.py",
-                            "iscc_core/minhash.py",
-                        ]
+                        ["iscc_core/cdc.py", "iscc_core/minhash.py", "iscc_core/simhash.py"]
                     ),
                     cmdclass=dict(build_ext=build_ext_gracefull),
                 )

iscc_core/simhash.pxd

@@ -0,0 +1 @@
+cpdef bytes alg_simhash(list[bytes] hash_digests)

iscc_core/simhash.py

@@ -1,4 +1,5 @@
 # -*- coding: utf-8 -*-
+from bitarray import bitarray
 
 
 def alg_simhash(hash_digests):
@@ -16,16 +17,16 @@ def alg_simhash(hash_digests):
     vector = [0] * n_bits
 
     for digest in hash_digests:
-        h = int.from_bytes(digest, "big", signed=False)
-
+        h = bitarray()
+        h.frombytes(digest)
         for i in range(n_bits):
-            vector[i] += h & 1
-            h >>= 1
+            vector[i] += h[i]
 
-    minfeatures = len(hash_digests) * 1.0 / 2
+    minfeatures = len(hash_digests) / 2
     shash = 0
 
     for i in range(n_bits):
-        shash |= int(vector[i] >= minfeatures) << i
+        if vector[i] >= minfeatures:
+            shash |= 1 << (n_bits - 1 - i)
 
-    return shash.to_bytes(n_bytes, "big", signed=False)
+    return shash.to_bytes(n_bytes, "big")