Address review feedback

- Fix sys.path to use os.path for reliable execution from any directory
- Fix ShortType test data to use modulo for large vector sizes
- Expand Cython deserializer benchmark docstring
- Remove unnecessary blank lines between test_configs groups
- Fix NotImplemented -> NotImplementedError in integration test

Author: mykaul

benchmarks/vector_deserialize.py

@@ -25,12 +25,13 @@
 Run with: python benchmarks/vector_deserialize.py
 """
 
+import os
 import sys
 import time
 import struct
 
-# Add parent directory to path
-sys.path.insert(0, '.')
+# Add project root to path so the benchmark can be run from any directory
+sys.path.insert(0, os.path.join(os.path.dirname(os.path.abspath(__file__)), ".."))
 
 from cassandra.cqltypes import FloatType, DoubleType, Int32Type, LongType, ShortType
 from cassandra.marshal import float_pack, double_pack, int32_pack, int64_pack, int16_pack
@@ -51,7 +52,7 @@ def create_test_data(vector_size, element_type):
         values = list(range(vector_size))
         pack_fn = int64_pack
     elif element_type == ShortType:
-        values = list(range(min(vector_size, 32767)))
+        values = [i % 32767 for i in range(vector_size)]
         pack_fn = int16_pack
     else:
         raise ValueError(f"Unsupported element type: {element_type}")
@@ -144,7 +145,12 @@ def benchmark_numpy_optimization(vector_type, serialized_data, iterations=10000)
 
 
 def benchmark_cython_deserializer(vector_type, serialized_data, iterations=10000):
-    """Benchmark Cython DesVectorType deserializer."""
+    """Benchmark Cython DesVectorType deserializer.
+
+    This benchmark requires the Cython deserializers extension to be compiled.
+    When the extension is not available, or the type does not have a dedicated
+    DesVectorType deserializer, the benchmark is silently skipped (returns None).
+    """
     try:
         from cassandra.deserializers import find_deserializer
     except ImportError:
@@ -284,15 +290,12 @@ def main():
         # Small vectors
         (3, FloatType, "float", 50000),
         (4, FloatType, "float", 50000),
-
         # Medium vectors (common in ML)
         (128, FloatType, "float", 10000),
         (384, FloatType, "float", 10000),
-
         # Large vectors (embeddings)
         (768, FloatType, "float", 5000),
         (1536, FloatType, "float", 2000),
-
         # Other types (smaller iteration counts)
         (128, DoubleType, "double", 10000),
         (768, DoubleType, "double", 5000),

tests/integration/standard/test_types.py

@@ -583,7 +583,7 @@ def test_can_insert_tuples_all_collection_datatypes(self):
 
         # make sure we're testing all non primitive data types in the future
         if set(COLLECTION_TYPES) != set(['tuple', 'list', 'map', 'set']):
-            raise NotImplemented('Missing datatype not implemented: {}'.format(
+            raise NotImplementedError('Missing datatype not implemented: {}'.format(
                 set(COLLECTION_TYPES) - set(['tuple', 'list', 'map', 'set'])
             ))