import base64 from collections import namedtuple from itertools import count from textwrap import indent import httpx from httpx import HTTPStatusError from backend.constants import SNEKBOX_URL from backend.models import FormResponse, Form with open("resources/unittest_template.py") as file: TEST_TEMPLATE = file.read() UnittestResult = namedtuple("UnittestResult", "question_id return_code passed result") def filter_unittests(form: Form) -> Form: """ Replace the unittest data section of code questions with the number of test cases. This is used to redact the exact tests when sending the form back to the frontend. """ for question in form.questions: if question.type == "code" and "unittests" in question.data: question.data["unittests"]["tests"] = len(question.data["unittests"]["tests"]) return form def _make_unit_code(units: dict[str, str]) -> str: """Compose a dict mapping unit names to their code into an actual class body.""" result = "" for unit_name, unit_code in units.items(): result += ( f"\ndef test_{unit_name.lstrip('#')}(unit):" # Function definition f"\n{indent(unit_code, ' ')}" # Unit code ) return indent(result, " ") def _make_user_code(code: str) -> str: """Compose the user code into an actual base64-encoded string variable.""" code = base64.b64encode(code.encode("utf8")).decode("utf8") return f'USER_CODE = b"{code}"' async def _post_eval(code: str) -> dict[str, str]: """Post the eval to snekbox and return the response.""" async with httpx.AsyncClient() as client: data = {"input": code} response = await client.post(SNEKBOX_URL, json=data, timeout=10) response.raise_for_status() return response.json() async def execute_unittest(form_response: FormResponse, form: Form) -> list[UnittestResult]: """Execute all the unittests in this form and return the results.""" unittest_results = [] for index, question in enumerate(form.questions): if question.type == "code" and "unittests" in question.data: passed = False # Tests starting with an hashtag should have censored names. hidden_test_counter = count(1) hidden_tests = { test.lstrip("#").lstrip("test_"): next(hidden_test_counter) for test in question.data["unittests"]["tests"].keys() if test.startswith("#") } # Compose runner code unit_code = _make_unit_code(question.data["unittests"]["tests"]) user_code = _make_user_code(form_response.response[question.id]) code = TEST_TEMPLATE.replace("### USER CODE", user_code) code = code.replace("### UNIT CODE", unit_code) try: response = await _post_eval(code) except HTTPStatusError: return_code = 99 result = "Unable to contact code runner." else: return_code = int(response["returncode"]) # Parse the stdout if the tests ran successfully if return_code == 0: stdout = response["stdout"] passed = bool(int(stdout[0])) # If the test failed, we have to populate the result string. if not passed: failed_tests = stdout[1:].strip().split(";") # Redact failed hidden tests for i, failed_test in enumerate(failed_tests.copy()): if failed_test in hidden_tests: failed_tests[i] = f"hidden_test_{hidden_tests[failed_test]}" result = ";".join(failed_tests) else: result = "" elif return_code in (5, 6, 99): result = response["stdout"] # Killed by NsJail elif return_code == 137: return_code = 7 result = "Timed out or ran out of memory." # Another code has been returned by CPython because of another failure. else: return_code = 99 result = "Internal error." unittest_results.append(UnittestResult( question_id=question.id, return_code=return_code, passed=passed, result=result )) return unittest_results