File size: 26,175 Bytes
a100cc5 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 |
import logging
import asyncio
from tqdm import tqdm
from .RepoKnowledgeGraph import RepoKnowledgeGraph
from .ModelService import ModelService
from .utils.logger_utils import setup_logger
from .utils.chunk_utils import organize_chunks_by_file_name, join_organized_chunks, extract_filename_from_chunk
from .Node import ChunkNode
LOGGER_NAME = "QUESTION_MAKER_LOGGER"
class QuestionMaker:
"""
The QuestionMaker class is responsible for generating code comprehension questions and answers
based on code chunks and knowledge graphs. It leverages a language model service to formulate
questions and answers that test deep understanding of code, focusing on mechanisms, design decisions,
and subtle behaviors. It supports generating questions for neighboring code chunks as well as for
specific entities (e.g., functions, classes) that are both declared and called in the codebase.
"""
def __init__(self):
"""
Initializes the QuestionMaker, sets up logging, and instantiates the model service.
"""
setup_logger(LOGGER_NAME)
self.logger = logging.getLogger(LOGGER_NAME)
self.model_service = ModelService()
def generate_questions_answers(self, candidate_chunks:dict) -> list:
"""
Placeholder for generating questions and answers from candidate chunks.
Args:
candidate_chunks (dict): Dictionary mapping chunk groups to process.
Returns:
list: List of question-answer pairs.
"""
pass
def test_chunk_sensibility(self, knowledge_graph: RepoKnowledgeGraph) -> list:
"""
Placeholder for testing the sensibility of code chunks in the knowledge graph.
Args:
knowledge_graph (RepoKnowledgeGraph): The knowledge graph to test.
Returns:
list: List of results or metrics.
"""
pass
async def make_n_neighbouring_chunk_questions_async(self, knowledge_graph: RepoKnowledgeGraph) -> list:
"""
Generates questions and answers for all possible groups of n directly neighboring code chunks
in each file of the knowledge graph. This helps assess understanding of code that spans multiple
adjacent chunks, such as related functions or code blocks.
Args:
knowledge_graph (RepoKnowledgeGraph): The knowledge graph to generate questions from.
Returns:
list: A list of dictionaries, each containing a question, answer, the involved chunks, and category.
"""
file_nodes = knowledge_graph.get_all_files()
# create candidate chunks dictionary
candidate_chunks = []
for file_node in file_nodes:
self.logger.info(f"Processing file node: {file_node}")
chunks = knowledge_graph.get_chunks_of_file(file_node.id)
num_chunks = len(chunks)
# For each n, collect all n-sized tuples of directly neighbouring chunks
for n in range(2, num_chunks + 1):
for i in range(num_chunks - n + 1):
# Only directly neighbouring chunks
candidate_chunks.append(list(chunks[i:i+n]))
# generate questions and answers from candidate chunks in parallel, in batches of 15
async def process_chunk_group(chunks):
"""
Helper coroutine to generate a question and answer for a specific group of neighboring chunks.
Args:
chunks (list): The list of code chunks to generate the question and answer from.
Returns:
dict: Contains question, answer, chunks, and category.
"""
question = await self._generate_neighboring_question_from_chunks_async(chunks)
answer = await self.answer_question_about_chunks_async(chunks, question)
return {
'question': question,
'clean_question': question,
'answer': answer,
'chunks': [chunk.dict() for chunk in chunks],
'category': 'neighbouring_chunks'
}
# Batch processing in groups of 15 with tqdm
batch_size = 15
results = []
total = len(candidate_chunks)
for i in tqdm(range(0, total, batch_size), desc="Generating neighbouring chunk questions", unit="batch"):
batch = candidate_chunks[i:i+batch_size]
tasks = [process_chunk_group(chunks) for chunks in batch]
batch_results = []
for coro in tqdm(asyncio.as_completed(tasks), total=len(tasks), desc="Questions in batch", leave=False):
batch_results.append(await coro)
results.extend(batch_results)
return results
async def make_entity_declaration_call_specific_questions_async(self, knowledge_graph: RepoKnowledgeGraph) -> list:
"""
Generates questions and answers about specific entities (e.g., functions, classes) that have both
a declaration and at least one call site in the knowledge graph. Focuses on cross-file references
by default.
Args:
knowledge_graph (RepoKnowledgeGraph): The knowledge graph to generate questions from.
Returns:
list: A list of dictionaries, each containing a question, answer, entity, involved chunks, and category.
"""
self.logger.info("Generating entity-specific questions.")
candidate_pairs = self.get_entities_with_declaration_and_calling(knowledge_graph)
async def process_entity_pair(pair):
"""
Helper coroutine to generate a question and answer for a specific entity's declaration and call site.
Args:
pair (dict): Contains entity name, declaring_chunk_id, and calling_chunk_id.
Returns:
dict: Contains question, answer, entity, chunks, and category.
"""
entity_name = pair['entity']
chunks = [knowledge_graph[pair['declaring_chunk_id']], knowledge_graph[pair['calling_chunk_id']]]
question = await self.make_entity_specific_question_async(chunks, entity_name)
answer = await self.answer_question_about_chunks_async(chunks, question)
return {
'question': question,
'clean_question': question,
'answer': answer,
'entity': entity_name,
'chunks': [chunk.dict() for chunk in chunks],
'category': 'entity_declaration_call_specific'
}
# Batch processing with tqdm
batch_size = 15
results = []
total = len(candidate_pairs)
for i in tqdm(range(0, total, batch_size), desc="Generating entity-specific questions", unit="batch"):
batch = candidate_pairs[i:i+batch_size]
tasks = [process_entity_pair(pair) for pair in batch]
batch_results = []
for coro in tqdm(asyncio.as_completed(tasks), total=len(tasks), desc="Questions in batch", leave=False):
batch_results.append(await coro)
results.extend(batch_results)
return results
async def make_interacting_entities_specific_questions_async(self, entity_A:str, entity_B:str,
decl_chunk_A: ChunkNode, decl_chunk_B: ChunkNode,
call_chunk: ChunkNode) -> str:
"""
Generates a question and answer about two entities that interact in the same chunk.
Each entity has a declaration and at least one call site, and the question focuses on their interaction.
Args:
entity_A (str): Name of the first entity.
entity_B (str): Name of the second entity.
decl_chunk_A (str): Chunk of the declaration of entity A.
decl_chunk_B (str): Chunk of the declaration of entity B.
call_chunk (str): Chunk ID where both entities interact.
Returns:
str: the generated question as plain text.
"""
entity_A_definition_code = decl_chunk_A.content
entity_B_definition_code = decl_chunk_B.content
entity_interaction_code = call_chunk.content
prompt = f"""You are given two code entities, {entity_A} and {entity_B}, along with a snippet where they interact.
Your task is to write **one clear and concise question** about their relationship.
### Input:
* {entity_A} Definition Code:
{entity_A_definition_code}
* {entity_B} Definition Code:
{entity_B_definition_code}
* Interaction Code (where they interact):
{entity_interaction_code}
### Guidelines:
* Ask about design, abstraction, dependencies, or side effects.
* The question should highlight something a developer might consider when reviewing or improving the code.
* Keep the question short and direct so it can be answered briefly.
* Do not explain the code or provide answers.
### Output:
**Question**: <your question here>
"""
initial_question = await self.model_service.query_async(prompt=prompt)
return await self.extract_question_from_generated_text_async(generated_text=initial_question)
def get_all_candidate_pairs_and_triplets(self, knowledge_graph: RepoKnowledgeGraph) -> list:
candidate_triplets = []
candidate_pairs = []
interacting_entity_triplets = self.get_interacting_entity_triplets(knowledge_graph)
for triplet in interacting_entity_triplets:
chunks = [
knowledge_graph[triplet['decl_chunk_A']],
knowledge_graph[triplet['decl_chunk_B']],
knowledge_graph[triplet['call_chunk']]
]
candidate_triplets.append({
'entities': (triplet['entity_A'], triplet['entity_B']),
'chunks': [chunk.dict() for chunk in chunks],
'category': 'interacting_entities'
})
declaration_calling_pairs = self.get_entities_with_declaration_and_calling(knowledge_graph)
for pair in declaration_calling_pairs:
chunks = [knowledge_graph[pair['declaring_chunk_id']], knowledge_graph[pair['calling_chunk_id']]]
candidate_pairs.append({
'entity': pair['entity'],
'chunks': [chunk.dict() for chunk in chunks],
'category': 'entity_declaration_call_specific'
})
return candidate_pairs, candidate_triplets
async def make_interacting_entity_questions_async(self, knowledge_graph: RepoKnowledgeGraph) -> list:
"""
Generates questions and answers about pairs of entities that interact in the same chunk.
Each entity has a declaration and at least one call site, and the question focuses on their interaction.
Args:
knowledge_graph (RepoKnowledgeGraph): The knowledge graph to generate questions from.
Returns:
list: A list of dictionaries, each containing a question, answer, entities, involved chunks, and category.
"""
self.logger.info("Generating interacting entity questions.")
triplets = self.get_interacting_entity_triplets(knowledge_graph)
async def process_triplet(triplet):
"""
Helper coroutine to generate a question and answer for a specific interacting entity triplet.
Args:
triplet (dict): Contains entity_A, entity_B, decl_chunk_A, decl_chunk_B, and call_chunk.
Returns:
dict: Contains question, answer, entities, chunks, and category.
"""
chunks = [
knowledge_graph[triplet['decl_chunk_A']],
knowledge_graph[triplet['decl_chunk_B']],
knowledge_graph[triplet['call_chunk']]
]
question = await self.make_interacting_entities_specific_questions_async(entity_A=triplet['entity_A'],
entity_B=triplet['entity_B'],
decl_chunk_A=knowledge_graph[triplet['decl_chunk_A']],
decl_chunk_B=knowledge_graph[triplet['decl_chunk_B']],
call_chunk=knowledge_graph[triplet['call_chunk']])
answer = await self.answer_question_about_chunks_async(chunks, question)
return {
'question': question,
'clean_question': question,
'answer': answer,
'entities': (triplet['entity_A'], triplet['entity_B']),
'chunks': [chunk.dict() for chunk in chunks],
'category': 'interacting_entities'
}
# Batch processing with tqdm
batch_size = 15
results = []
total = len(triplets)
for i in tqdm(range(0, total, batch_size), desc="Generating interacting entity questions", unit="batch"):
batch = triplets[i:i+batch_size]
tasks = [process_triplet(triplet) for triplet in batch]
batch_results = []
for coro in tqdm(asyncio.as_completed(tasks), total=len(tasks), desc="Questions in batch", leave=False):
batch_results.append(await coro)
results.extend(batch_results)
return results
async def _generate_neighboring_question_from_chunks_async(self, chunks: list) -> str:
"""
Generates a single code comprehension question for a group of code chunks using the model service.
The question is designed to probe deep understanding of the code's mechanisms, design, or pitfalls.
Args:
chunks (list): The list of code chunks to generate the question from.
Returns:
str: The generated question as plain text.
"""
organized_chunks = organize_chunks_by_file_name(chunks)
joined_chunks = join_organized_chunks(organized_chunks)
system_prompt = """You are an expert in evaluating code comprehension. The user will provide, in the next message, the content of a code submission (in any programming language). Your goal is to analyze this code, identify its critical, subtle, or obscure aspects, and generate **one relevant question in English** to ask someone in order to assess their understanding of the code.
This question should focus on:
* essential mechanisms of how the code works,
* important design decisions,
* potential pitfalls or unexpected behaviors,
* or any aspect that requires deep comprehension.
The goal is to test whether the person has **truly understood** the code—not just skimmed through it.
Respond with **only one question**, in plain text. Do not include any explanation, comment, or wrapper (e.g., no dictionaries, no lists).
"""
initial_question = await self.model_service.query_with_instructions_async(instructions=system_prompt, prompt=joined_chunks)
return await self.extract_question_from_generated_text_async(generated_text=initial_question)
async def answer_question_about_chunks_async(self, chunks: list, question: str) -> str:
"""
Generates an answer to a code comprehension question about a group of code chunks using the model service.
The answer should demonstrate deep understanding and cover mechanisms, design, and pitfalls.
Args:
chunks (list): The list of code chunks to answer the question about.
question (str): The question to answer.
Returns:
str: The generated answer as plain text.
"""
organized_chunks = organize_chunks_by_file_name(chunks)
joined_chunks = join_organized_chunks(organized_chunks)
system_prompt = """You are an expert in evaluating code comprehension. The user will provide, in the next message, the content of a code submission (in any programming language) and a question about it. Your goal is to analyze this code, identify its critical, subtle, or obscure aspects, and generate **one relevant answer in English** to the question.
This answer should focus on:
* essential mechanisms of how the code works,
* important design decisions,
* potential pitfalls or unexpected behaviors,
* or any aspect that requires deep comprehension.
The goal is to provide a clear and thorough answer that demonstrates a deep understanding of the code.
"""
return await self.model_service.query_with_instructions_async(instructions=system_prompt, prompt=joined_chunks + "\n\n" + question)
async def make_entity_specific_question_async(self, chunks: list, entity_name:str):
"""
Generates a question about a specific entity (e.g., function, class) in the context of the provided code chunks.
The question is designed to probe understanding of the entity's purpose, behavior, and interactions.
Args:
chunks (list): The list of code chunks to generate the question from.
entity_name (str): The name of the entity to focus on.
Returns:
str: The generated question as plain text.
"""
organized_chunks = organize_chunks_by_file_name(chunks)
joined_chunks = join_organized_chunks(organized_chunks)
system_prompt = f"""You will be given one or more code snippets, possibly from multiple files.
A specific entity (such as a class, function, or variable) will be identified.
---
## Entity of Focus: {entity_name}
### Task:
* Write **one clear and concise question** about this entity.
* The question should highlight something a developer might consider, such as its purpose, behavior, interactions, or potential improvements.
### Guidelines:
* Keep the question short and direct.
* Do not explain the code or give an answer.
### Output:
**Question**: <your question here>
"""
initial_question= await self.model_service.query_with_instructions_async(instructions=system_prompt, prompt=joined_chunks)
return await self.extract_question_from_generated_text_async(generated_text=initial_question)
def get_entities_with_declaration_and_calling(self, knowledge_graph: RepoKnowledgeGraph, cross_file_only: bool = True) -> list:
"""
Finds all entities in the knowledge graph that have both a declaration and at least one call site.
Optionally restricts to cases where the declaration and call are in different files (cross-file).
Args:
knowledge_graph (RepoKnowledgeGraph): The knowledge graph to search in.
cross_file_only (bool): If True, only consider cross-file declaration/call pairs.
Returns:
list: List of dictionaries with 'entity', 'declaring_chunk_id', and 'calling_chunk_id'.
"""
candidate_pairs = []
entities = knowledge_graph.entities
for entity_name in entities:
entity = entities[entity_name]
if len(entity['declaring_chunk_ids']) and len(entity['calling_chunk_ids']):
found = False
for declaring_chunk_id in entity['declaring_chunk_ids']:
for calling_chunk_id in entity['calling_chunk_ids']:
if declaring_chunk_id != calling_chunk_id:
if cross_file_only and extract_filename_from_chunk(knowledge_graph[declaring_chunk_id]) == extract_filename_from_chunk(knowledge_graph[calling_chunk_id]):
continue
else:
candidate_pairs.append({'entity': entity_name, 'declaring_chunk_id' : declaring_chunk_id, 'calling_chunk_id': calling_chunk_id})
found = True
break
if found:
break
return candidate_pairs
def get_interacting_entity_triplets(self, knowledge_graph: RepoKnowledgeGraph) -> list:
"""
Finds triplets of chunk ids such that:
- Two entities (A, B) are interacting in the same chunk (call_chunk)
- Each entity has a declaring chunk (decl_chunk_A, decl_chunk_B)
- Both entities have non-empty declaring_chunk_ids and calling_chunk_ids
Returns:
list of dicts with keys:
'entity_A', 'entity_B', 'decl_chunk_A', 'decl_chunk_B', 'call_chunk'
"""
triplets = []
seen_pairs = set()
entities = knowledge_graph.entities
for entity_A_name, entity_A in entities.items():
if not entity_A['declaring_chunk_ids'] or not entity_A['calling_chunk_ids']:
continue
for entity_B_name, entity_B in entities.items():
if entity_A_name == entity_B_name:
continue
if not entity_B['declaring_chunk_ids'] or not entity_B['calling_chunk_ids']:
continue
pair_key = (entity_A_name, entity_B_name)
if pair_key in seen_pairs:
continue
# Find intersection of calling_chunk_ids
call_chunks = set(entity_A['calling_chunk_ids']) & set(entity_B['calling_chunk_ids'])
found = False
for call_chunk in call_chunks:
for decl_chunk_A in entity_A['declaring_chunk_ids']:
for decl_chunk_B in entity_B['declaring_chunk_ids']:
triplets.append({
'entity_A': entity_A_name,
'entity_B': entity_B_name,
'decl_chunk_A': decl_chunk_A,
'decl_chunk_B': decl_chunk_B,
'call_chunk': call_chunk
})
seen_pairs.add(pair_key)
found = True
break
if found:
break
if found:
break
return triplets
async def extract_question_from_generated_text_async(self, generated_text: str) -> str:
"""
Extracts the question from the generated text. The question is expected to be the last line of the text.
Args:
generated_text (str): The text generated by the model.
Returns:
str: The extracted question.
"""
prompt = f"Extract only the question from the following text. Return the question exactly, with no extra words or labels:\n\n{generated_text}\n\n"
return await self.model_service.query_async(prompt=prompt)
def select_diverse_candidates(self, candidate_pairs, candidate_triplets, max_pairs=20, max_triplets=20):
"""
Selects a limited number of pairs and triplets with maximum diversity in entity representation.
Args:
candidate_pairs (list): List of candidate pairs (dicts with 'entity', ...).
candidate_triplets (list): List of candidate triplets (dicts with 'entities', ...).
max_pairs (int): Maximum number of pairs to select.
max_triplets (int): Maximum number of triplets to select.
Returns:
(list, list): Selected pairs and triplets.
"""
# Select pairs
selected_pairs = []
used_entities = set()
for pair in candidate_pairs:
entity = pair['entity']
if entity not in used_entities:
selected_pairs.append(pair)
used_entities.add(entity)
if len(selected_pairs) >= max_pairs:
break
# Select triplets
selected_triplets = []
used_entities_triplets = set()
for triplet in candidate_triplets:
entities = set(triplet['entities'])
if not entities & used_entities_triplets:
selected_triplets.append(triplet)
used_entities_triplets.update(entities)
if len(selected_triplets) >= max_triplets:
break
return selected_pairs, selected_triplets
async def transform_answser_into_mcq_answer_async(self, question, answer, chunks):
"""
Transforms the question and answer into a format suitable for MCQ generation.
"""
code = join_organized_chunks(organize_chunks_by_file_name(chunks))
prompt = f"""
You are an expert Python developer and technical writer. I will give you:
1. A Python code snippet
2. A question about that code
3. A detailed answer to the question
Your task is to **sanitize** the answer. That means:
- Strip away all fluff, filler, and redundant explanation
- Focus only on what directly answers the question
- Make it **short, clear, and direct**, as if it were a correct MCQ answer
- Prefer concise phrases or a single clear sentence over paragraph explanations
- Keep any necessary technical detail, but no more than needed
Do **not** repeat the question. Do **not** rephrase the code. Just give the concise, final answer.
- **Input Code**:
{code}
- **Question**:
{question}
- **Original Answer**:
{answer}
- **Sanitized Answer**:
"""
return await self.model_service.query_async(prompt)
|