Merge branch 'main' into fix-benchmark-prompt

Author: aseembits93

.github/workflows/end-to-end-test-init-optim.yaml

@@ -20,7 +20,7 @@ jobs:
       COLUMNS: 110
       MAX_RETRIES: 3
       RETRY_DELAY: 5
-      EXPECTED_IMPROVEMENT_PCT: 300
+      EXPECTED_IMPROVEMENT_PCT: 30
       CODEFLASH_END_TO_END: 1
     steps:
       - name: 🛎️ Checkout

codeflash/api/aiservice.py

@@ -1,5 +1,7 @@
 from __future__ import annotations
 
+import time
+
 import json
 import os
 import platform
@@ -95,6 +97,7 @@ def optimize_python_code(
         - List[OptimizationCandidate]: A list of Optimization Candidates.
 
         """
+        start_time = time.perf_counter()
         payload = {
             "source_code": source_code,
             "dependency_code": dependency_code,
@@ -118,6 +121,8 @@ def optimize_python_code(
             optimizations_json = response.json()["optimizations"]
             logger.info(f"Generated {len(optimizations_json)} candidates.")
             console.rule()
+            end_time = time.perf_counter()
+            logger.debug(f"Optimization took {end_time - start_time:.2f} seconds.")
             return [
                 OptimizedCandidate(
                     source_code=opt["source_code"],

codeflash/verification/concolic_testing.py

@@ -1,5 +1,7 @@
 from __future__ import annotations
 
+import time
+
 import ast
 import subprocess
 import tempfile
@@ -20,6 +22,7 @@
 def generate_concolic_tests(
     test_cfg: TestConfig, args: Namespace, function_to_optimize: FunctionToOptimize, function_to_optimize_ast: ast.AST
 ) -> tuple[dict[str, list[FunctionCalledInTest]], str]:
+    start_time = time.perf_counter()
     function_to_concolic_tests = {}
     concolic_test_suite_code = ""
     if (
@@ -84,4 +87,6 @@ def generate_concolic_tests(
         else:
             logger.debug(f"Error running CrossHair Cover {': ' + cover_result.stderr if cover_result.stderr else '.'}")
             console.rule()
+    end_time = time.perf_counter()
+    logger.debug(f"Generated concolic tests in {end_time - start_time:.2f} seconds")
     return function_to_concolic_tests, concolic_test_suite_code

codeflash/verification/verifier.py

@@ -1,5 +1,7 @@
 from __future__ import annotations
 
+import time
+
 import ast
 from pathlib import Path
 from typing import TYPE_CHECKING
@@ -29,6 +31,7 @@ def generate_tests(
 ) -> tuple[str, str, Path] | None:
     # TODO: Sometimes this recreates the original Class definition. This overrides and messes up the original
     #  class import. Remove the recreation of the class definition
+    start_time = time.perf_counter()
     test_module_path = Path(module_name_from_file_path(test_path, test_cfg.tests_project_rootdir))
     response = aiservice_client.generate_regression_tests(
         source_code_being_tested=source_code_being_tested,
@@ -54,7 +57,8 @@ def generate_tests(
     else:
         logger.warning(f"Failed to generate and instrument tests for {function_to_optimize.function_name}")
         return None
-
+    end_time = time.perf_counter()
+    logger.debug(f"Generated tests in {end_time - start_time:.2f} seconds")
     return (
         generated_test_source,
         instrumented_behavior_test_source,

tests/scripts/end_to_end_test_init_optimization.py

@@ -9,7 +9,7 @@ def run_test(expected_improvement_pct: int) -> bool:
         file_path="remove_control_chars.py",
         function_name="CharacterRemover.remove_control_characters",
         test_framework="pytest",
-        min_improvement_x=1.0,
+        min_improvement_x=0.3,
         coverage_expectations=[
             CoverageExpectation(
                 function_name="CharacterRemover.remove_control_characters", expected_coverage=100.0, expected_lines=[14]