HeXi/tests/test_outbound_media.py

"""出站媒体内联化单元测试

hexi/core 包 __init__ 会触发 NoneBot 运行时初始化(get_driver),
无法裸导入,故用 importlib 以真实模块名加载 outbound_media 并注册进
sys.modules(game: 与 test_rate_limit / test_cooldown 同一套路)。

文件一律用 tmp_path 造真实文件,**不 mock 文件系统**——这类转换的坑
恰恰在真实路径语义上(file:// 的百分号编码、Windows 盘符、相对路径),
mock 掉就等于把要验的东西验没了。
"""

import importlib.util
import sys
from pathlib import Path

import pytest
from nonebot.adapters.onebot.v11 import Message, MessageSegment

_BASE = Path(__file__).resolve().parents[1] / "hexi" / "core"
_MOD_NAME = "hexi.core.outbound_media"


def _load_module(name: str, path: Path):
    spec = importlib.util.spec_from_file_location(name, path)
    module = importlib.util.module_from_spec(spec)
    sys.modules[name] = module
    spec.loader.exec_module(module)
    return module


@pytest.fixture(scope="module")
def om():
    module = _load_module(_MOD_NAME, _BASE / "outbound_media.py")
    yield module
    sys.modules.pop(_MOD_NAME, None)


@pytest.fixture
def media_file(tmp_path: Path) -> Path:
    """带空格与中文的真实文件:同时覆盖 URL 编码与多字节路径两种坑。"""
    fp = tmp_path / "有 空格 的文件.jpg"
    fp.write_bytes(b"hello")
    return fp


class _FakeAdapter:
    def __init__(self, name: str = "OneBot V11"):
        self._name = name

    def get_name(self) -> str:
        return self._name


class _FakeBot:
    def __init__(self, name: str = "OneBot V11"):
        self.adapter = _FakeAdapter(name)


# ------------------------------------------------------------ local_file_of


def test_file_uri_with_spaces_and_unicode(om, media_file: Path):
    """file:// URI(带 %20 与中文百分号编码)必须能还原成本机真实路径。

    这是整个模块最容易静默失效的一处:天真的
    `Path(unquote(urlparse(uri).path))` 在 Windows 上会得到前导斜杠的
    `/D:/...`,is_file() 恒为 False 而**不报错**,只在跨机时才暴露。
    """
    uri = media_file.as_uri()
    assert "%" in uri  # 确认确实带百分号编码,否则这条测试没验到东西
    assert om.local_file_of(uri) == media_file


def test_plain_absolute_path(om, media_file: Path):
    """f2s() 对 str 是原样透传的,所以裸路径也得认。"""
    assert om.local_file_of(str(media_file)) == media_file


def test_relative_path(om, media_file: Path, monkeypatch):
    monkeypatch.chdir(media_file.parent)
    assert om.local_file_of(media_file.name) == Path(media_file.name)


@pytest.mark.parametrize(
    "value",
    [
        "http://example.com/a.jpg",
        "https://example.com/a.jpg",
        "base64://aGVsbG8=",
    ],
)
def test_remote_and_inline_values_pass_through(om, value: str):
    assert om.local_file_of(value) is None


@pytest.mark.parametrize(
    "value",
    [
        "D:/nope/does_not_exist.jpg",
        "file:///D:/nope/does_not_exist.jpg",
        # 协议端回传的图片 id / 它自己的缓存路径:不是本机文件,必须自然放行
        "{12345678-1234-1234-1234-123456789abc}.jpg",
        "C:/NapCat/cache/abcdef.image",
        "",
    ],
)
def test_non_local_values_are_left_alone(om, value: str):
    assert om.local_file_of(value) is None


@pytest.mark.parametrize("value", [None, 123, b"bytes", ["a.jpg"]])
def test_non_string_values_are_left_alone(om, value):
    assert om.local_file_of(value) is None


# ------------------------------------------------------------ 改写目标


async def test_falls_back_to_base64_when_no_url(om, media_file: Path, monkeypatch):
    """拿不到可达地址(未挂载)→ base64 兜底,字节由 bot 自己携带。"""
    monkeypatch.setattr(om, "_mounted", False)
    assert await om.rewrite_file(str(media_file)) == "base64://aGVsbG8="


async def test_url_mode_when_mounted(om, media_file: Path, monkeypatch):
    monkeypatch.setattr(om, "_mounted", True)
    monkeypatch.setenv("HEXI_MEDIA_BASE_URL", "http://192.168.2.15:39697/")
    url = await om.rewrite_file(str(media_file))
    assert url.startswith("http://192.168.2.15:39697/media/")
    # 基地址结尾斜杠不该被拼成 //media
    assert "//media" not in url

    token = url.rsplit("/", 1)[1]
    assert om._tokens[token][0] == media_file


async def test_url_mode_can_be_disabled(om, media_file: Path, monkeypatch):
    """URL 模式关掉后一律 base64,即使媒体服务挂着。"""
    monkeypatch.setattr(om, "_mounted", True)
    monkeypatch.setenv("HEXI_MEDIA_BASE_URL", "http://192.168.2.15:39697")
    monkeypatch.setenv("HEXI_INLINE_MEDIA_URL", "false")
    assert await om.rewrite_file(str(media_file)) == "base64://aGVsbG8="


async def test_rewrite_is_idempotent(om, media_file: Path, monkeypatch):
    """重复执行必须无副作用:第二次看到 base64:// 应直接放行。"""
    monkeypatch.setattr(om, "_mounted", False)
    once = await om.rewrite_file(str(media_file))
    assert await om.rewrite_file(once) is None


async def test_rewrite_returns_none_for_remote(om):
    assert await om.rewrite_file("http://example.com/a.jpg") is None


# ------------------------------------------------------------ 遍历形态


async def test_walk_rewrites_message_segments(om, media_file: Path, monkeypatch):
    monkeypatch.setattr(om, "_mounted", False)
    msg = Message(
        [
            MessageSegment.text("看图"),
            MessageSegment.image(media_file),
            MessageSegment.face(1),
        ]
    )
    await om._walk(msg)
    assert msg[0].data["text"] == "看图"
    assert msg[1].data["file"] == "base64://aGVsbG8="
    assert msg[2].type == "face"  # 非媒体段不动


@pytest.mark.parametrize("seg_type", ["image", "record", "video"])
async def test_walk_rewrites_all_media_types(
    om, media_file: Path, monkeypatch, seg_type
):
    monkeypatch.setattr(om, "_mounted", False)
    seg = {"type": seg_type, "data": {"file": str(media_file)}}
    await om._walk(seg)
    assert seg["data"]["file"] == "base64://aGVsbG8="


async def test_walk_handles_bare_dict_segment(om, media_file: Path, monkeypatch):
    """group_daily_analysis 那种纯 dict 段(不是 Message 也不是 MessageSegment)。"""
    monkeypatch.setattr(om, "_mounted", False)
    payload = {"type": "image", "data": {"file": str(media_file)}}
    await om._walk(payload)
    assert payload["data"]["file"] == "base64://aGVsbG8="


async def test_walk_handles_list_of_dict_segments(om, media_file: Path, monkeypatch):
    monkeypatch.setattr(om, "_mounted", False)
    payload = [
        {"type": "text", "data": {"text": "hi"}},
        {"type": "image", "data": {"file": str(media_file)}},
    ]
    await om._walk(payload)
    assert payload[1]["data"]["file"] == "base64://aGVsbG8="


async def test_walk_forward_node_with_str_content(om, media_file: Path, monkeypatch):
    """message_utils.build_forward_nodes: content 是 str(不含媒体,应保持原样)。"""
    monkeypatch.setattr(om, "_mounted", False)
    nodes = [
        {"type": "node", "data": {"name": "x", "uin": "1", "content": "纯文本"}},
    ]
    await om._walk(nodes)
    assert nodes[0]["data"]["content"] == "纯文本"


async def test_walk_forward_node_with_dict_list_content(
    om, media_file: Path, monkeypatch
):
    """video_analysis._build_forward_nodes: content 是段字典数组。"""
    monkeypatch.setattr(om, "_mounted", False)
    nodes = [
        {
            "type": "node",
            "data": {
                "name": "x",
                "uin": "1",
                "content": [{"type": "video", "data": {"file": str(media_file)}}],
            },
        }
    ]
    await om._walk(nodes)
    assert nodes[0]["data"]["content"][0]["data"]["file"] == "base64://aGVsbG8="


async def test_walk_forward_node_with_message_content(
    om, media_file: Path, monkeypatch
):
    """MessageSegment.node_custom 会把 Message 原样塞进 data["content"]。"""
    monkeypatch.setattr(om, "_mounted", False)
    node = MessageSegment.node_custom(
        1, "n", Message([MessageSegment.image(media_file)])
    )
    await om._walk([node])
    inner = node.data["content"][0]
    assert isinstance(inner, MessageSegment)
    assert inner.data["file"] == "base64://aGVsbG8="


async def test_walk_recurses_messages_key(om, media_file: Path, monkeypatch):
    """转发的 messages 还可能是 Message 对象(alconna 的 include() 返回同类型)。"""
    monkeypatch.setattr(om, "_mounted", False)
    data = {
        "messages": Message(
            [
                MessageSegment.node_custom(
                    1, "n", Message([MessageSegment.image(media_file)])
                )
            ]
        )
    }
    await om._walk(data["messages"])
    assert data["messages"][0].data["content"][0].data["file"] == "base64://aGVsbG8="


async def test_walk_leaves_remote_media_alone(om, media_file: Path, monkeypatch):
    monkeypatch.setattr(om, "_mounted", False)
    seg = {"type": "image", "data": {"file": "https://example.com/a.jpg"}}
    await om._walk(seg)
    assert seg["data"]["file"] == "https://example.com/a.jpg"


async def test_walk_respects_depth_limit(om, media_file: Path, monkeypatch):
    """超深嵌套不再下钻:拦截层不该在畸形结构上无界递归。"""
    monkeypatch.setattr(om, "_mounted", False)
    node: dict = {"type": "image", "data": {"file": str(media_file)}}
    for _ in range(20):
        node = {"type": "node", "data": {"content": [node]}}
    await om._walk(node)
    deepest = node
    while isinstance(deepest, dict) and "content" in deepest.get("data", {}):
        deepest = deepest["data"]["content"][0]
    assert deepest["data"]["file"] == str(media_file)  # 没被改写


# ------------------------------------------------------------ 钩子


async def test_hook_rewrites_send_api(om, media_file: Path, monkeypatch):
    monkeypatch.setattr(om, "_mounted", False)
    data = {"group_id": 1, "message": Message([MessageSegment.image(media_file)])}
    await om._inline_outbound_media(_FakeBot(), "send_group_msg", data)
    assert data["message"][0].data["file"] == "base64://aGVsbG8="


async def test_hook_ignores_unrelated_api(om, media_file: Path, monkeypatch):
    """既非发送也非文件上传的 API 一律不碰。"""
    monkeypatch.setattr(om, "_mounted", True)
    monkeypatch.setenv("HEXI_MEDIA_BASE_URL", "http://192.168.2.186:39697")
    data = {"group_id": 1, "file": str(media_file)}
    await om._inline_outbound_media(_FakeBot(), "get_group_info", data)
    assert data["file"] == str(media_file)


async def test_hook_ignores_other_adapters(om, media_file: Path, monkeypatch):
    monkeypatch.setattr(om, "_mounted", False)
    data = {"message": Message([MessageSegment.image(media_file)])}
    before = data["message"][0].data["file"]
    await om._inline_outbound_media(_FakeBot("Telegram"), "send_group_msg", data)
    assert data["message"][0].data["file"] == before


async def test_hook_disabled_by_env(om, media_file: Path, monkeypatch):
    """关掉开关 = 声明「协议端就在本机」,此时原样发送才是对的。"""
    monkeypatch.setenv("HEXI_INLINE_MEDIA", "false")
    data = {"message": Message([MessageSegment.image(media_file)])}
    before = data["message"][0].data["file"]
    await om._inline_outbound_media(_FakeBot(), "send_group_msg", data)
    assert data["message"][0].data["file"] == before
    # f2s() 对 Path 产出 file:// URI,正是本模块要拦的那种值
    assert before.startswith("file://")


async def test_hook_handles_forward_messages_key(om, media_file: Path, monkeypatch):
    monkeypatch.setattr(om, "_mounted", False)
    data = {
        "group_id": 1,
        "messages": [
            {
                "type": "node",
                "data": {
                    "name": "n",
                    "uin": "1",
                    "content": [{"type": "image", "data": {"file": str(media_file)}}],
                },
            }
        ],
    }
    await om._inline_outbound_media(_FakeBot(), "send_group_forward_msg", data)
    seg = data["messages"][0]["data"]["content"][0]
    assert seg["data"]["file"] == "base64://aGVsbG8="


async def test_hook_never_raises_on_broken_segment(om, monkeypatch):
    """拦截层绝不能反过来把发送搞挂:畸形段只告警。"""
    monkeypatch.setattr(om, "_mounted", False)
    data = {"message": [{"type": "image", "data": {"file": object()}}]}
    await om._inline_outbound_media(_FakeBot(), "send_group_msg", data)


# ------------------------------------------------------------ 文件上传


def _unescaped_uri(path: Path) -> str:
    """video_analysis._file_uri 的手工拼法：正斜杠、不转义。"""
    return "file:///" + str(path.resolve()).replace("\\", "/")


async def test_upload_file_url_when_mounted(om, media_file: Path, monkeypatch):
    monkeypatch.setattr(om, "_mounted", True)
    monkeypatch.setenv("HEXI_MEDIA_BASE_URL", "http://192.168.2.186:39697")
    data = {"group_id": 1, "file": str(media_file), "name": "显示名.zip"}
    await om._inline_outbound_media(_FakeBot(), "upload_group_file", data)
    assert data["file"].startswith("http://192.168.2.186:39697/media/")
    assert data["name"] == "显示名.zip"  # 其它参数一律不碰
    assert om._tokens[data["file"].rsplit("/", 1)[1]][0] == media_file


@pytest.mark.parametrize(
    "builder",
    [
        str,  # 裸路径（alconna $onebot11:file 的 as_posix() 形态）
        lambda p: p.as_uri(),  # Path.as_uri()：带百分号转义
        _unescaped_uri,  # 手工拼：不转义
    ],
    ids=["plain", "as_uri", "unescaped"],
)
async def test_upload_file_accepts_all_uri_forms(
    om, media_file: Path, monkeypatch, builder
):
    monkeypatch.setattr(om, "_mounted", True)
    monkeypatch.setenv("HEXI_MEDIA_BASE_URL", "http://192.168.2.186:39697")
    data = {"file": builder(media_file)}
    await om._inline_outbound_media(_FakeBot(), "upload_private_file", data)
    assert data["file"].startswith("http://192.168.2.186:39697/media/")


async def test_upload_file_has_no_base64_fallback(om, media_file: Path, monkeypatch):
    """文件上传不做 base64 兜底：没 URL 就原样放行（不能比改动前更糟）。"""
    monkeypatch.setattr(om, "_mounted", False)
    data = {"group_id": 1, "file": str(media_file)}
    await om._inline_outbound_media(_FakeBot(), "upload_group_file", data)
    assert data["file"] == str(media_file)


async def test_upload_file_respects_url_toggle(om, media_file: Path, monkeypatch):
    monkeypatch.setattr(om, "_mounted", True)
    monkeypatch.setenv("HEXI_MEDIA_BASE_URL", "http://192.168.2.186:39697")
    monkeypatch.setenv("HEXI_INLINE_MEDIA_URL", "false")
    data = {"group_id": 1, "file": str(media_file)}
    await om._inline_outbound_media(_FakeBot(), "upload_group_file", data)
    assert data["file"] == str(media_file)


async def test_upload_file_ignores_s3_url(om, media_file: Path, monkeypatch):
    """S3 预签名链接是群文件的首选路径，不能被碰。"""
    monkeypatch.setattr(om, "_mounted", True)
    monkeypatch.setenv("HEXI_MEDIA_BASE_URL", "http://192.168.2.186:39697")
    url = "http://192.168.2.15:5246/PLANA/x.zip?X-Amz-Signature=abc"
    data = {"group_id": 1, "file": url}
    await om._inline_outbound_media(_FakeBot(), "upload_group_file", data)
    assert data["file"] == url


async def test_local_file_of_handles_literal_percent(tmp_path: Path, om):
    """文件名含字面 %XX 时，未转义的手拼 URI 不能被 unquote 带偏。

    先按解码结果找（会解析成错误路径），再退回原始路径 —— 这条兜底的意义就在此。
    """
    fp = tmp_path / "100%20off.zip"
    fp.write_bytes(b"zip")
    assert om.local_file_of(_unescaped_uri(fp)) == fp


async def test_media_endpoint_roundtrip(om, media_file: Path, monkeypatch):
    """token → 路径 的登记与过期语义(不依赖 FastAPI,直接验注册表)。"""
    monkeypatch.setattr(om, "_mounted", True)
    monkeypatch.setenv("HEXI_MEDIA_BASE_URL", "http://127.0.0.1:11011")
    monkeypatch.setenv("HEXI_MEDIA_URL_TTL", "60")
    url = await om.rewrite_file(str(media_file))
    token = url.rsplit("/", 1)[1]

    path, expires = om._tokens[token]
    assert path == media_file
    assert expires > om.time.monotonic()
    # token 不可猜:长度足够且不含路径信息
    assert len(token) >= 20
    assert "空格" not in token