⚡️ Speed up method _PQEncoderConfigUpdate.merge_with_existing by 22%

[codeflash-ai]

📄 22% (0.22x) speedup for _PQEncoderConfigUpdate.merge_with_existing in weaviate/collections/classes/config_vector_index.py

⏱️ Runtime : 2.37 microseconds → 1.95 microsecondss (best of 79 runs)

📝 Explanation and details

The optimization caches attribute lookups by storing self.type_ and self.distribution in local variables at the beginning of the method. This eliminates repeated attribute access overhead during the conditional checks.

Key changes:

• Added type_ = self.type_ and distribution = self.distribution assignments
• Modified conditionals to use local variables instead of self.attribute access

Why this improves performance:
In Python, attribute access involves dictionary lookups and potential descriptor protocol calls. By caching these values as local variables, the method avoids redundant attribute resolution during the is not None checks. Local variable access is significantly faster than attribute access because it uses direct array indexing in the local namespace rather than hash table lookups.

Test case performance:
The optimization shows consistent 20-24% speedups across all test scenarios, particularly effective for cases where both attributes are None (most common path). The speedup is most pronounced in simple cases like empty schemas with None fields, demonstrating that even basic attribute caching can yield meaningful performance gains in frequently called methods.

✅ Correctness verification report:

Test	Status
⚙️ Existing Unit Tests	🔘 None Found
🌀 Generated Regression Tests	✅ 6 Passed
⏪ Replay Tests	🔘 None Found
🔎 Concolic Coverage Tests	🔘 None Found
📊 Tests Coverage	60.0%

🌀 Generated Regression Tests and Runtime

from enum import Enum
from typing import Any, Dict, Optional

# imports
import pytest  # used for our unit tests
from weaviate.collections.classes.config_vector_index import \
    _PQEncoderConfigUpdate


# --- Mock dependencies for standalone testing ---
class PQEncoderType(Enum):
    INT8 = "int8"
    FLOAT32 = "float32"
    CUSTOM = "custom"

class PQEncoderDistribution(Enum):
    UNIFORM = "uniform"
    GAUSSIAN = "gaussian"
    CUSTOM_DIST = "custom_dist"

class _ConfigUpdateModel:
    pass
from weaviate.collections.classes.config_vector_index import \
    _PQEncoderConfigUpdate

# --- Unit tests ---

# 1. Basic Test Cases






def test_merge_with_existing_type_none_distribution_none():
    """Test when both fields are explicitly None."""
    cfg = _PQEncoderConfigUpdate(type_=None, distribution=None)
    schema = {"type": "int8", "distribution": "uniform"}
    codeflash_output = cfg.merge_with_existing(schema.copy()); result = codeflash_output # 779ns -> 647ns (20.4% faster)







def test_merge_with_existing_schema_is_empty_and_none_fields():
    """Test empty schema and both fields None; should remain empty."""
    cfg = _PQEncoderConfigUpdate(type_=None, distribution=None)
    schema = {}
    codeflash_output = cfg.merge_with_existing(schema.copy()); result = codeflash_output # 787ns -> 633ns (24.3% faster)








#------------------------------------------------
from enum import Enum
# function to test (reconstructed for testing purposes)
from typing import Any, Dict, Optional

# imports
import pytest  # used for our unit tests
from weaviate.collections.classes.config_vector_index import \
    _PQEncoderConfigUpdate


# Dummy base class for testing
class _ConfigUpdateModel:
    pass

# Dummy Enums to simulate PQEncoderType and PQEncoderDistribution
class PQEncoderType(Enum):
    TYPE_A = "type_a"
    TYPE_B = "type_b"
    TYPE_C = "type_c"

class PQEncoderDistribution(Enum):
    DIST_UNIFORM = "uniform"
    DIST_GAUSSIAN = "gaussian"
    DIST_CUSTOM = "custom"
from weaviate.collections.classes.config_vector_index import \
    _PQEncoderConfigUpdate

# unit tests

# ----------- BASIC TEST CASES -----------










def test_merge_with_existing_type_and_distribution_none():
    """Test when both type_ and distribution are None, schema should remain unchanged."""
    config = _PQEncoderConfigUpdate(type_=None, distribution=None)
    schema = {"type": "should_stay", "distribution": "should_stay", "x": 5}
    codeflash_output = config.merge_with_existing(schema); result = codeflash_output # 801ns -> 666ns (20.3% faster)








#------------------------------------------------
from weaviate.collections.classes.config_vector_index import PQEncoderDistribution
from weaviate.collections.classes.config_vector_index import PQEncoderType
from weaviate.collections.classes.config_vector_index import _PQEncoderConfigUpdate

def test__PQEncoderConfigUpdate_merge_with_existing():
    _PQEncoderConfigUpdate.merge_with_existing(_PQEncoderConfigUpdate(type_=PQEncoderType.TILE, distribution=PQEncoderDistribution.NORMAL), {})

Timer unit: 1e-09 s

To edit these changes git checkout codeflash/optimize-_PQEncoderConfigUpdate.merge_with_existing-mh30tjhc and push.