File size: 13,356 Bytes
6759906 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 |
#!/usr/bin/env python3
"""
Code Completion Benchmark for CodeReality-1T Dataset
Evaluates code completion models using Pass@k metrics
"""
import json
import os
import re
import random
from typing import Dict, List, Tuple, Optional
from collections import defaultdict
def load_dataset_sample(data_dir: str, sample_size: int = 200) -> List[Dict]:
"""Load sample of repositories with code files."""
print(f"🔍 Loading sample of {sample_size} repositories with code files...")
repositories = []
files = [f for f in os.listdir(data_dir) if f.endswith('.jsonl')]
random.shuffle(files)
for filename in files[:15]: # Sample from first 15 files
file_path = os.path.join(data_dir, filename)
try:
with open(file_path, 'r', encoding='utf-8', errors='ignore') as f:
for line in f:
if len(repositories) >= sample_size:
break
try:
repo_data = json.loads(line)
# Filter repositories with code files
if has_code_files(repo_data):
repositories.append(repo_data)
except json.JSONDecodeError:
continue
except Exception as e:
continue
if len(repositories) >= sample_size:
break
print(f"✅ Loaded {len(repositories)} repositories with code files")
return repositories
def has_code_files(repo: Dict) -> bool:
"""Check if repository contains code files."""
code_extensions = {'.py', '.js', '.java', '.cpp', '.c', '.go', '.rs', '.ts'}
files = repo.get('files', [])
for file_obj in files:
if isinstance(file_obj, dict):
file_path = file_obj.get('path', '')
if any(file_path.endswith(ext) for ext in code_extensions):
return True
return False
def extract_function_snippets(repo: Dict, language: str = 'python') -> List[Dict]:
"""Extract function definitions for completion tasks."""
snippets = []
# Language-specific patterns
patterns = {
'python': r'def\s+(\w+)\s*\([^)]*\):\s*',
'javascript': r'function\s+(\w+)\s*\([^)]*\)\s*{',
'java': r'(?:public|private|protected)?\s*(?:static)?\s*\w+\s+(\w+)\s*\([^)]*\)\s*{',
'cpp': r'\w+\s+(\w+)\s*\([^)]*\)\s*{',
}
if language not in patterns:
return snippets
pattern = patterns[language]
extension_map = {
'python': '.py',
'javascript': '.js',
'java': '.java',
'cpp': '.cpp'
}
target_ext = extension_map[language]
files = repo.get('files', [])
for file_obj in files:
if isinstance(file_obj, dict):
file_path = file_obj.get('path', '')
content = file_obj.get('content', '')
if file_path.endswith(target_ext) and content:
matches = list(re.finditer(pattern, content, re.MULTILINE))
for match in matches:
start_pos = match.start()
function_name = match.group(1)
# Get context before function
lines_before = content[:start_pos].split('\n')
context_lines = lines_before[-5:] if len(lines_before) >= 5 else lines_before
context = '\n'.join(context_lines)
# Get function body (simplified - until next function or end)
remaining_content = content[start_pos:]
lines = remaining_content.split('\n')
function_lines = []
indent_level = None
for i, line in enumerate(lines):
if i == 0:
function_lines.append(line)
continue
# Detect indentation level
if indent_level is None and line.strip():
indent_level = len(line) - len(line.lstrip())
# Stop if we hit same or lower indentation level (end of function)
if line.strip() and indent_level is not None:
current_indent = len(line) - len(line.lstrip())
if current_indent <= indent_level and not line.strip().startswith(('if', 'for', 'while', 'try', 'except', 'else', 'elif')):
break
function_lines.append(line)
# Limit function length
if len(function_lines) > 20:
break
function_body = '\n'.join(function_lines)
# Create completion task: provide function signature, expect body
if len(function_lines) > 3: # Only meaningful functions
snippets.append({
'function_name': function_name,
'context': context,
'prompt': function_lines[0], # Function signature
'completion': '\n'.join(function_lines[1:]), # Function body
'file_path': file_path,
'language': language
})
return snippets
def simple_code_completion_model(prompt: str, language: str) -> List[str]:
"""Simple rule-based code completion for demonstration."""
completions = []
# Generate multiple completions (for Pass@k evaluation)
templates = {
'python': [
" pass",
" return None",
" # TODO: implement this function\n pass",
" result = None\n return result",
" # Implementation needed\n raise NotImplementedError()"
],
'javascript': [
" return null;",
" // TODO: implement\n return;",
" throw new Error('Not implemented');",
" var result = null;\n return result;",
" console.log('Function called');\n return;"
],
'java': [
" return null;",
" // TODO: implement this method\n return null;",
" throw new UnsupportedOperationException();",
" Object result = null;\n return result;",
" System.out.println(\"Method called\");\n return null;"
]
}
if language in templates:
# Return multiple variations for Pass@k evaluation
return templates[language]
else:
return ["// TODO: implement"]
def evaluate_completion_quality(predicted: str, actual: str) -> float:
"""Simple evaluation of completion quality."""
# Normalize strings
pred_lines = [line.strip() for line in predicted.split('\n') if line.strip()]
actual_lines = [line.strip() for line in actual.split('\n') if line.strip()]
if not actual_lines:
return 0.0
# Check for basic structural similarity
score = 0.0
# Check if both are empty implementations
empty_indicators = ['pass', 'todo', 'not implemented', 'null', 'return;', 'return null'}
pred_empty = any(indicator in predicted.lower() for indicator in empty_indicators)
actual_empty = any(indicator in actual.lower() for indicator in empty_indicators)
if pred_empty and actual_empty:
score += 0.8
elif not pred_empty and not actual_empty:
# Check for keyword similarity
pred_keywords = set(re.findall(r'\b\w+\b', predicted.lower()))
actual_keywords = set(re.findall(r'\b\w+\b', actual.lower()))
if actual_keywords:
keyword_overlap = len(pred_keywords & actual_keywords) / len(actual_keywords)
score += keyword_overlap * 0.6
# Check for similar line count
line_ratio = min(len(pred_lines), len(actual_lines)) / max(len(pred_lines), len(actual_lines))
score += line_ratio * 0.4
return min(score, 1.0)
def calculate_pass_at_k(completion_results: List[Tuple[List[str], str]], k: int = 1) -> float:
"""Calculate Pass@k metric."""
if k <= 0:
return 0.0
total_passed = 0
for completions, ground_truth in completion_results:
# Take top k completions
top_k_completions = completions[:k]
# Check if any completion passes
passed = False
for completion in top_k_completions:
quality_score = evaluate_completion_quality(completion, ground_truth)
if quality_score > 0.5: # Threshold for "passing"
passed = True
break
if passed:
total_passed += 1
return total_passed / len(completion_results) if completion_results else 0.0
def run_completion_benchmark(repositories: List[Dict]) -> Dict:
"""Run code completion benchmark."""
print("🧮 Running code completion benchmark...")
results = {
'total_repositories': len(repositories),
'completion_tasks': [],
'language_stats': defaultdict(int),
'pass_at_1': 0.0,
'pass_at_3': 0.0,
'pass_at_5': 0.0,
'average_quality': 0.0
}
completion_results = []
quality_scores = []
# Extract function snippets from repositories
for repo in repositories:
for language in ['python', 'javascript', 'java']:
snippets = extract_function_snippets(repo, language)
for snippet in snippets[:2]: # Limit per repo for performance
results['language_stats'][language] += 1
# Generate completions
completions = simple_code_completion_model(snippet['prompt'], language)
ground_truth = snippet['completion']
completion_results.append((completions, ground_truth))
# Calculate quality for first completion
if completions:
quality = evaluate_completion_quality(completions[0], ground_truth)
quality_scores.append(quality)
results['completion_tasks'].append({
'function_name': snippet['function_name'],
'language': language,
'prompt_length': len(snippet['prompt']),
'completion_length': len(ground_truth)
})
# Calculate metrics
results['pass_at_1'] = calculate_pass_at_k(completion_results, 1)
results['pass_at_3'] = calculate_pass_at_k(completion_results, 3)
results['pass_at_5'] = calculate_pass_at_k(completion_results, 5)
results['average_quality'] = sum(quality_scores) / len(quality_scores) if quality_scores else 0.0
return results
def print_benchmark_results(results: Dict):
"""Print formatted benchmark results."""
print("=" * 60)
print("🎯 CODE COMPLETION BENCHMARK RESULTS")
print("=" * 60)
print(f"Total repositories: {results['total_repositories']}")
print(f"Completion tasks: {len(results['completion_tasks'])}")
print(f"\n📊 Pass@k Metrics:")
print(f" Pass@1: {results['pass_at_1']:.3f}")
print(f" Pass@3: {results['pass_at_3']:.3f}")
print(f" Pass@5: {results['pass_at_5']:.3f}")
print(f" Average Quality: {results['average_quality']:.3f}")
print(f"\n🔤 Language Distribution:")
for language, count in sorted(results['language_stats'].items(), key=lambda x: x[1], reverse=True):
print(f" {language}: {count} functions")
print(f"\n💡 Insights:")
print("- This is a simplified demonstration benchmark")
print("- Real evaluation requires more sophisticated code execution")
print("- CodeReality-1T provides diverse, noisy code for robust testing")
print("- Consider functional correctness testing for production models")
def main():
"""Run code completion benchmark."""
print("🚀 CodeReality-1T Code Completion Benchmark")
print("=" * 60)
# Configuration
data_dir = "/mnt/z/CodeReality_Final/unified_dataset"
sample_size = 100
if not os.path.exists(data_dir):
print(f"❌ Dataset directory not found: {data_dir}")
print("Please update the data_dir path to point to your CodeReality-1T dataset")
return
# Load dataset sample
repositories = load_dataset_sample(data_dir, sample_size)
if not repositories:
print("❌ No repositories loaded. Check dataset path.")
return
# Run benchmark
results = run_completion_benchmark(repositories)
# Print results
print_benchmark_results(results)
# Save results
output_file = "code_completion_results.json"
with open(output_file, 'w') as f:
# Convert defaultdict to regular dict for JSON serialization
results_json = {
'total_repositories': results['total_repositories'],
'completion_tasks': results['completion_tasks'],
'language_stats': dict(results['language_stats']),
'pass_at_1': results['pass_at_1'],
'pass_at_3': results['pass_at_3'],
'pass_at_5': results['pass_at_5'],
'average_quality': results['average_quality']
}
json.dump(results_json, f, indent=2)
print(f"\n💾 Results saved to: {output_file}")
if __name__ == "__main__":
main() |