Skip to content

JiuwenSwarm is an intelligent AI Agent built on openJiuwen. It extends the powerful capabilities of large language models directly to your fingertips through various communication apps you use daily.

Python
3.9 KB
Raw Blame
# Copyright (c) Huawei Technologies Co., Ltd. 2026. All rights reserved.

from __future__ import annotations

from pathlib import Path
from unittest.mock import MagicMock, patch

import pytest

from jiuwenswarm.extensions.hooks_context import ArtifactPostProcessHookContext
from jiuwenswarm.server.runtime.code_source_unicode import (
    _normalize_code_artifact_hook,
    is_code_unicode_readable_enabled,
    normalize_python_script_file,
    normalize_python_source_unicode_literals,
    register_code_source_unicode_hook,
)


def test_normalize_unicode_escape_to_readable() -> None:
    source = 'doc.add_paragraph("\\u6807\\u9898")\n'
    normalized, count = normalize_python_source_unicode_literals(source)
    assert count == 1
    assert '"标题"' in normalized


def test_raw_string_unchanged() -> None:
    source = 'value = r"\\u6587"\n'
    normalized, count = normalize_python_source_unicode_literals(source)
    assert count == 0
    assert normalized == source


def test_inconsistent_indent_degrades_to_original() -> None:
    """IndentationError from generate_tokens must not propagate."""
    source = 'def f():\n    if True:\n        x = "\\u6807"\n      y = 1\n'
    normalized, count = normalize_python_source_unicode_literals(source)
    assert count == 0
    assert normalized == source


def test_tab_error_degrades_to_original() -> None:
    """TabError from generate_tokens must not propagate (3.11 tokenize may not raise it)."""
    source = 'x = "\\u6807"\n'
    with patch(
        "jiuwenswarm.server.runtime.code_source_unicode.tokenize.generate_tokens",
        side_effect=TabError("inconsistent use of tabs and spaces in indentation"),
    ):
        normalized, count = normalize_python_source_unicode_literals(source)
    assert count == 0
    assert normalized == source


def test_incomplete_token_degrades_to_original() -> None:
    source = 'x = "\\u6807'
    normalized, count = normalize_python_source_unicode_literals(source)
    assert count == 0
    assert normalized == source


def test_surrogate_escape_pair_unchanged() -> None:
    source = 'x = "\\ud83d\\ude00"\n'
    normalized, count = normalize_python_source_unicode_literals(source)
    assert count == 0
    assert normalized == source


def test_normalize_python_script_file_surrogate_preserves_file(tmp_path: Path) -> None:
    source = 'x = "\\ud83d\\ude00"\n'
    script = tmp_path / "demo.py"
    script.write_text(source, encoding="utf-8")
    assert normalize_python_script_file(script) == 0
    assert script.read_text(encoding="utf-8") == source


def test_config_disabled() -> None:
    assert is_code_unicode_readable_enabled({"code_generation": {"unicode_readable": False}}) is False


@pytest.mark.asyncio
async def test_artifact_hook_normalizes_file(tmp_path: Path) -> None:
    script = tmp_path / "demo.py"
    script.write_text('x = "\\u6807\\u9898"\n', encoding="utf-8")
    ctx = ArtifactPostProcessHookContext(
        session_id="s1",
        tool_name="write_file",
        artifact_paths=[str(script)],
    )
    await _normalize_code_artifact_hook(ctx)
    assert script.read_text(encoding="utf-8") == 'x = "标题"\n'


def test_normalize_python_script_file(tmp_path: Path) -> None:
    script = tmp_path / "demo.py"
    script.write_text('x = "\\u6807\\u9898"\n', encoding="utf-8")
    assert normalize_python_script_file(script) == 1
    assert script.read_text(encoding="utf-8") == 'x = "标题"\n'


def test_register_hook_idempotent() -> None:
    registry = MagicMock()
    with pytest.MonkeyPatch.context() as mp:
        mp.setattr(
            "jiuwenswarm.extensions.registry.ExtensionRegistry.get_instance",
            lambda: registry,
        )
        mp.setattr(
            "jiuwenswarm.server.runtime.code_source_unicode._HOOK_REGISTERED",
            False,
        )
        register_code_source_unicode_hook()
        register_code_source_unicode_hook()
    assert registry.register.call_count == 1