Blame for tasks/leetcode/tests/oracle-runtime/clone/leetcode_app/judge.py

"""Deterministic offline verdict model.

The server never executes guest source code, in any language, under any
identity. A submission is graded by a fixed, ordered set of rules applied to
the submitted text, while the *expected* output of each test case is produced
by this repository's own reference implementations
(:mod:`leetcode_app.content`).

Rule order, applied to the comment-stripped source:

1. blank source                      -> Compile Error
2. required entry point missing      -> Compile Error
3. source still equals the starter   -> Wrong Answer, no output produced
4. no result-producing statement     -> Wrong Answer, no output produced
5. no token of the question's method -> Wrong Answer at the first case
6. otherwise                         -> Accepted

Everything derived from the source (verdict, runtime, memory, the failing
case) is a pure function of ``(question, language, source, cases)``: the same
submission always produces the same result.
"""

from __future__ import annotations

import hashlib
import json
import re
from dataclasses import dataclass, field
from typing import Any

from . import content
from .content import Question


ACCEPTED = "Accepted"
WRONG_ANSWER = "Wrong Answer"
COMPILE_ERROR = "Compile Error"
RUNTIME_ERROR = "Runtime Error"

# A statement that produces the answer, per language.
RESULT_TOKENS: dict[str, tuple[str, ...]] = {
    "python3": ("return",),
    "pandas": ("return",),
    "cpp": ("return",),
    "java": ("return",),
    "javascript": ("return",),
    "mysql": ("select",),
    "bash": ("sed", "awk", "head", "tail", "cat", "nl", "echo", "sort", "printf"),
}

# The declaration a compiler needs to find, per language.
ENTRY_MARKERS: dict[str, tuple[str, ...]] = {
    "python3": ("class solution", "def "),
    "pandas": ("def ",),
    "cpp": ("class solution",),
    "java": ("class solution",),
    "javascript": ("var ", "function", "const ", "let "),
    "mysql": ("select",),
    "bash": (),
}

ENTRY_MESSAGES: dict[str, str] = {
    "python3": "NameError: name 'Solution' is not defined",
    "pandas": "NameError: the required function is not defined",
    "cpp": "error: 'Solution' was not declared in this scope",
    "java": "error: cannot find symbol: class Solution",
    "javascript": "ReferenceError: the required function is not defined",
    "mysql": "You have an error in your SQL syntax",
    "bash": "command not found",
}

_LINE_COMMENTS: dict[str, tuple[str, ...]] = {
    "python3": ("#",),
    "pandas": ("#",),
    "bash": ("#",),
    "mysql": ("--", "#"),
    "cpp": ("//",),
    "java": ("//",),
    "javascript": ("//",),
}

_BLOCK_COMMENT = re.compile(r"/\*.*?\*/", re.DOTALL)
_SQL_BLOCK_COMMENT = re.compile(r"/\*.*?\*/", re.DOTALL)
_WHITESPACE = re.compile(r"\s+")


class CaseInputError(ValueError):
    """A test case could not be read as input for this question."""


@dataclass
class CaseResult:
    index: int
    input: str
    output: str | None
    expected: str
    passed: bool
    executed: bool = True

    def as_dict(self) -> dict[str, Any]:
        return {
            "index": self.index,
            "input": self.input,
            "output": self.output,
            "expected": self.expected,
            "passed": self.passed,
            "executed": self.executed,
        }


@dataclass
class Verdict:
    status: str
    message: str = ""
    cases: list[CaseResult] = field(default_factory=list)
    runtime_ms: int | None = None
    memory_mb: float | None = None
    runtime_percentile: float | None = None
    memory_percentile: float | None = None

    @property
    def passed(self) -> int:
        return sum(1 for case in self.cases if case.passed)

    @property
    def total(self) -> int:
        return len(self.cases)

    @property
    def failing(self) -> CaseResult | None:
        for case in self.cases:
            if not case.passed:
                return case
        return None

    def as_dict(self) -> dict[str, Any]:
        failing = self.failing
        return {
            "status": self.status,
            "message": self.message,
            "passed": self.passed,
            "total": self.total,
            "runtime_ms": self.runtime_ms,
            "memory_mb": self.memory_mb,
            "runtime_percentile": self.runtime_percentile,
            "memory_percentile": self.memory_percentile,
            "cases": [case.as_dict() for case in self.cases],
            "failing": failing.as_dict() if failing is not None else None,
        }


def strip_comments(source: str, language: str) -> str:
    text = source
    if language in {"cpp", "java", "javascript"}:
        text = _BLOCK_COMMENT.sub(" ", text)
    if language == "mysql":
        text = _SQL_BLOCK_COMMENT.sub(" ", text)
    markers = _LINE_COMMENTS.get(language, ("#",))
    lines: list[str] = []
    for line in text.splitlines():
        cut = len(line)
        for marker in markers:
            position = line.find(marker)
            if position != -1:
                cut = min(cut, position)
        lines.append(line[:cut])
    return "\n".join(lines)


def normalize(source: str) -> str:
    return _WHITESPACE.sub(" ", source).strip()


def _digest(question: Question, language: str, source: str) -> bytes:
    payload = f"{question.slug}\0{language}\0{normalize(source)}".encode("utf-8")
    return hashlib.sha256(payload).digest()


def _performance(question: Question, language: str, source: str) -> dict[str, Any]:
    """Deterministic runtime/memory figures derived from the submission."""

    digest = _digest(question, language, source)
    runtime_ms = digest[0] % 90
    memory_mb = 15.0 + (digest[1] % 60) / 10
    runtime_percentile = round(20.0 + (digest[2] % 790) / 10, 2)
    memory_percentile = round(20.0 + (digest[3] % 790) / 10, 2)
    return {
        "runtime_ms": runtime_ms,
        "memory_mb": round(memory_mb, 2),
        "runtime_percentile": runtime_percentile,
        "memory_percentile": memory_percentile,
    }


def expected_output(question: Question, raw_case: str) -> str:
    """Expected output for one case, from this repository's own reference."""

    fixed = question.fixed_expected.get(raw_case.strip())
    if fixed is not None:
        return fixed
    reference = content.REFERENCES.get(question.slug)
    if reference is None:
        raise CaseInputError(
            "This question runs against its own fixed data set."
        )
    try:
        arguments = content.parse_case(raw_case)
    except (json.JSONDecodeError, ValueError) as error:
        raise CaseInputError(f"The test case could not be read: {error}") from error
    if not arguments:
        raise CaseInputError("The test case is empty.")
    try:
        return content.format_expected(reference(*arguments))
    except TypeError as error:
        raise CaseInputError(
            "The test case does not match this question's input format."
        ) from error
    except (IndexError, KeyError, ValueError, ZeroDivisionError) as error:
        raise CaseInputError(f"The test case could not be evaluated: {error}") from error


def _entry_markers(question: Question, language: str) -> tuple[str, ...]:
    override = question.entry_markers.get(language)
    if override is not None:
        return override
    return ENTRY_MARKERS.get(language, ())


def classify(question: Question, language: str, source: str) -> tuple[str, str]:
    """Return ``(status, message)`` for the submission's source alone."""

    stripped = strip_comments(source, language)
    if not stripped.strip():
        return COMPILE_ERROR, "error: the submission is empty"
    lowered = stripped.casefold()
    for marker in _entry_markers(question, language):
        if marker not in lowered:
            return COMPILE_ERROR, ENTRY_MESSAGES.get(language, "error: compile failed")
    starter = question.starter.get(language, "")
    if normalize(strip_comments(starter, language)) == normalize(stripped):
        return WRONG_ANSWER, "The starter code does not produce an answer yet."
    result_tokens = RESULT_TOKENS.get(language, ("return",))
    if not any(token in lowered for token in result_tokens):
        return WRONG_ANSWER, "The submission never produces a result."
    if not any(token in lowered for token in question.technique_tokens):
        return WRONG_ANSWER, ""
    return ACCEPTED, ""


def evaluate(
    question: Question,
    language: str,
    source: str,
    cases: list[str],
) -> Verdict:
    """Grade ``source`` against ``cases`` and return the verdict."""

    status, message = classify(question, language, source)
    results: list[CaseResult] = []
    produce_output = status == ACCEPTED
    failing_index = 0 if status == WRONG_ANSWER else -1
    for index, raw_case in enumerate(cases):
        try:
            expected = expected_output(question, raw_case)
        except CaseInputError as error:
            return Verdict(status=RUNTIME_ERROR, message=str(error), cases=[])
        output: str | None
        if produce_output:
            output, passed, executed = expected, True, True
        elif status == COMPILE_ERROR:
            output, passed, executed = None, False, False
        elif index == failing_index:
            output = None if message else question.wrong_output
            passed, executed = False, True
        else:
            # Execution stops at the first case that does not match.
            output, passed, executed = None, False, False
        results.append(
            CaseResult(
                index=index,
                input=raw_case,
                output=output,
                expected=expected,
                passed=passed,
                executed=executed,
            )
        )
    verdict = Verdict(status=status, message=message, cases=results)
    if status == ACCEPTED:
        for key, value in _performance(question, language, source).items():
            setattr(verdict, key, value)
    elif status == WRONG_ANSWER:
        performance = _performance(question, language, source)
        verdict.runtime_ms = performance["runtime_ms"]
        verdict.memory_mb = performance["memory_mb"]
    return verdict